High School Philosophy & Ethics

Three arguments, and one of them is broken

Read these three. Each has two supporting lines and a conclusion. Decide which ones you would accept before you read on.

A. (1) Every student caught with a phone in class loses it until Friday. (2) Ravi was caught with a phone in class on Tuesday. (3) So Ravi loses his phone until Friday.

B. (1) Every teenager in this country owns a smartphone. (2) Ravi is a teenager in this country. (3) So Ravi owns a smartphone.

C. (1) Everyone who cheated on the test got a zero. (2) Ravi got a zero. (3) So Ravi cheated on the test.

Most readers accept A, hesitate over B, and only notice on a second pass that C has a hole in it. The interesting part is that the three reactions have three different causes, and philosophy begins when you can name them.

A holds together, and its two supporting lines are true of Ravi's school, so its conclusion is forced. B also holds together: if line 1 and line 2 were both true, line 3 could not be false. But line 1 is not true. Some teenagers do not own a smartphone, so B proves nothing about Ravi even though nothing is wrong with its shape. C is different again, and worse. Its supporting lines could both be true while its conclusion is false, because there are other ways to get a zero. Ravi may have been absent. He may have handed in a blank page. The rule says cheating leads to a zero; it does not say a zero can only come from cheating.

Key idea: There are two separate questions to ask about any argument. Does it hold together? Are its parts true? Mixing them is the single most common mistake in reasoning, and the rest of this lesson is about keeping them apart.

What an argument is, in this subject

Outside philosophy, an argument is two people raising their voices. Inside it, an argument is a set of statements offered as grounds for another statement. The grounds are the premises. The statement they are meant to support is the conclusion. That is all. No shouting required, and no disagreement either: you can build an argument for something everyone already believes.

A statement is something that can be true or false. "The bus is late" is a statement. "Is the bus late?" is not, and neither is "Catch the bus." Questions and commands cannot be premises, which is why the first job in reading any messy paragraph is to rewrite its claims as flat statements you could mark true or false.

Ordinary speech hides the structure. Someone says: "Obviously we should start school later, teenagers are exhausted." Two statements are in there, and one of them is doing the supporting. Rewritten:

(1) Teenagers at this school are exhausted in the morning.
(2) So school should start later.

Now you can see that something is missing. Being exhausted does not by itself entail a change to the timetable; you need a bridge, something like "If students are exhausted in the morning, the school day should start later." Philosophers call that a suppressed premise, and finding it is not nitpicking. The suppressed premise is usually the controversial one, which is exactly why it was left unsaid.

The point: Set an argument out in numbered lines before you judge it. Half the work of criticism is discovering what the arguer never wrote down.

Validity is a claim about shape, not about the world

An argument is valid when it is impossible for its premises to be true and its conclusion false at the same time. The Internet Encyclopedia of Philosophy puts it in one line: a deductive argument is valid if and only if it takes a form that makes it impossible for the premises to be true and the conclusion nevertheless to be false. Notice the word "form". Validity is not about whether the premises are actually true. It is about whether truth would have to flow through.

This is the part that feels wrong at first, so here is the discomfort head on. The following argument is valid:

(1) All fish are mammals.
(2) All whales are fish.
(3) So all whales are mammals.

Both premises are false. The conclusion happens to be true. And the argument is valid, because if everything in the fish class were a mammal, and everything in the whale class were a fish, then whales would have to be mammals. The shape guarantees it. Validity is a conditional promise: if the premises, then the conclusion. A conditional promise can be kept by someone whose premises are nonsense.

The four combinations are worth having in front of you.

Read the second row again. A valid argument with a false premise tells you nothing at all about its conclusion. That is why "but it is logical" is never a defence on its own.

There is only one combination that cannot happen: a valid argument with all true premises and a false conclusion. That is ruled out by the definition of validity, and it is the whole reason logic is useful.

Soundness: valid, and the premises are actually true

An argument is sound when it is valid and all of its premises are in fact true. Sound arguments have true conclusions, guaranteed. That is the only guarantee in the subject, and it is bought at a price: you now have to defend every premise, and premises about the world are defended with evidence, not with logic.

So arguments fail in two very different ways, and the repair is different for each. If an argument is invalid, no amount of research on the premises will help; the structure has to change. If an argument is valid but a premise is false, the structure is fine and the fight moves to the facts. Saying which of the two is wrong with an argument is the most useful single sentence you can produce in a philosophy class.

What matters here: Validity is about form and is settled by thinking. Truth of premises is about the world and is settled by evidence. Soundness needs both.

How to test for validity: try to break it

Here is a procedure you can run on any argument, without notation. Ask: can I describe a situation, however strange, in which every premise is true and the conclusion is false? If you can, the argument is invalid and your description is a counterexample. If you honestly cannot, that is evidence the argument is valid.

Run it on argument C from the opening.

(1) Everyone who cheated on the test got a zero. (2) Ravi got a zero. (3) So Ravi cheated.

Situation: nine students cheated and each got a zero, so premise 1 holds. Ravi was off sick, missed the test, and was recorded as zero under a different rule. Premise 2 holds. Conclusion false. Argument broken in one sentence. Notice what the counterexample did not do: it did not claim Ravi is innocent. It showed that these premises do not settle the question either way.

Now run the same test on argument A. You would need a situation where every student caught with a phone loses it until Friday, and Ravi was caught with a phone, and yet Ravi does not lose his phone until Friday. Try to build it. Every attempt requires quietly breaking premise 1, by adding an exception the premise says does not exist. No counterexample is available, and that is what validity feels like from the inside.

One warning about the test. Failing to think of a counterexample is not proof of validity, only evidence. Students often miss counterexamples because they imagine only realistic situations. Validity does not care about realism. Whales can be fish in a counterexample; the question is whether the described situation is coherent, not whether it is likely.

Two forms worth memorising, and one impostor

A handful of shapes carry most everyday reasoning. Two of them are valid and look almost identical to one that is not.

Affirming the antecedent, or modus ponens: If P then Q. P. Therefore Q. "If the test is on Friday, we revise Thursday. The test is on Friday. So we revise Thursday." Valid.

Denying the consequent, or modus tollens: If P then Q. Not Q. Therefore not P. "If the test is on Friday, we revise Thursday. We did not revise Thursday. So the test is not on Friday." Valid, and this is the shape that does most of the work in science, where a prediction fails and the hypothesis behind it takes the damage.

The impostor, affirming the consequent: If P then Q. Q. Therefore P. "If the test is on Friday, we revise Thursday. We revised Thursday. So the test is on Friday." Invalid. We might have revised Thursday out of habit, or for a different subject. Q can be true for reasons that have nothing to do with P.

Argument C at the top of this lesson was an instance of the impostor in disguise: cheating leads to a zero, Ravi has a zero, therefore cheating. Once you can see that shape you will find it everywhere, including in your own first drafts.

The oldest catalogue of such shapes is the syllogism, worked out by Aristotle in the fourth century BCE. "All men are mortal; Socrates is a man; so Socrates is mortal" is the example every logic book uses, and it is the same machinery as argument A.

In short: If P then Q is not the same claim as if Q then P. Almost every everyday reasoning error is some version of forgetting that.

Deduction, induction, and what each one can promise

Not every good argument is valid, and demanding validity everywhere would leave you unable to reason about the future. Compare:

Deductive: (1) All mammals have lungs. (2) A dolphin is a mammal. (3) So a dolphin has lungs. If the premises are true, the conclusion cannot fail. That is deductive reasoning: the conclusion adds nothing that was not already contained in the premises, which is why it is safe and why it never tells you anything new about the world.

Inductive: (1) Every one of the four hundred swans recorded in this valley since 1990 has been white. (2) So the next swan recorded here will be white. If the premise is true the conclusion is probable, not guaranteed. That is inductive reasoning: it reaches beyond the evidence, which is why it can teach you something and why it can be wrong without anyone making a mistake.

Inductive arguments are not graded valid or invalid. They are graded strong or weak, and strength depends on things validity never cares about: how many cases, how varied, how they were collected, whether anyone looked for exceptions. Four hundred swans from one valley is weaker than four hundred from forty countries. A sample of friends who agree with you is weaker still.

Around 1740 David Hume asked what justifies the move from observed cases to unobserved ones, and could not find an answer that did not already assume the thing in question. Why think the future will resemble the past? Because it always has. But "it always has, so it will" is itself the inductive move being questioned. This is the problem of induction, and no agreed solution exists nearly three centuries later. Science works anyway. Philosophy is allowed to record both facts at once.

Where the neat distinction gets untidy

Textbooks present deduction and induction as two clean boxes. The specialist literature is less sure. Timothy Shanahan's survey for the Internet Encyclopedia of Philosophy walks through proposal after proposal for drawing the line, including intention, evidential completeness, and necessity against probability, and shows each one running into trouble. Consider: "The best explanation of the broken window and the football in the hallway is that someone kicked a ball indoors." What box is that in? It is not valid, so not deductive. It does not generalise from many cases either, so calling it induction stretches the word.

That untidiness is not a reason to drop the distinction, which is genuinely useful most of the time. It is a first lesson in how philosophy works: a distinction can be worth teaching and still be under repair. When you meet a confident textbook claim in this course, it is fair to ask whose account it is and what the objection to it was.

Common misconceptions

• "Valid means true." It does not. "All fish are mammals, all whales are fish, so all whales are mammals" is valid and has two false premises. Valid means the premises could not be true while the conclusion is false.
• "If I disagree with the conclusion, the argument is invalid." Validity is about the link from premises to conclusion. A valid argument for a conclusion you reject is a signal to find the false premise, which is real work and often changes your mind.
• "Finding a counterexample shows the conclusion is false." It shows the premises do not establish the conclusion. The conclusion may still be true for other reasons.
• "Inductive arguments are just weak deductive ones." They are a different kind of support with different standards. A well-designed study is a strong inductive argument and is not defective for failing to be valid.
• "Logic settles arguments about politics or morality." Logic checks the link. Almost every real dispute turns on a premise, and premises are defended with evidence, definitions or moral claims, not with logic alone.

Where this leaves us

• An argument is premises offered in support of a conclusion, and rewriting one in numbered lines usually exposes a premise nobody stated.
• Valid means the premises could not be true with the conclusion false. It is a fact about form, and it can hold when every premise is false.
• Sound means valid plus all premises true. Only sound arguments guarantee their conclusions.
• To test validity, try to describe a coherent case where the premises hold and the conclusion fails. Success means invalid; repeated failure is evidence of validity, not proof.
• Modus ponens and modus tollens are valid. Affirming the consequent is not, and it is the error hiding in most everyday reasoning mistakes.
• Inductive arguments are graded strong or weak rather than valid or invalid, and Hume's problem of induction is still open.
• Say which failure you are alleging: a broken link, or a false premise. The repairs are different.

Sources

1. Internet Encyclopedia of Philosophy. (n.d.). Validity and soundness. University of Tennessee at Martin. iep.utm.edu
2. Shanahan, T. (n.d.). Deductive and inductive arguments. Internet Encyclopedia of Philosophy. iep.utm.edu
3. Beall, J., Restall, G., & Sagi, G. (2024). Logical consequence. In E. N. Zalta & U. Nodelman (Eds.), Stanford Encyclopedia of Philosophy. plato.stanford.edu
4. Internet Encyclopedia of Philosophy. (n.d.). Argument. University of Tennessee at Martin. iep.utm.edu

A published correlation, and the argument it invites

In October 2012 the New England Journal of Medicine printed a short paper by the cardiologist Franz Messerli reporting that, across a set of countries, chocolate consumed per person and Nobel laureates per head of population rose together, with a correlation coefficient of 0.791. The paper wondered aloud whether the flavanols in cocoa might improve thinking.

Here is the argument the finding invites, written out honestly.

(1) Countries where people eat more chocolate produce more Nobel laureates per head.
(2) So eating chocolate raises a population's chance of producing Nobel laureates.

Where exactly does this fail? Not at premise 1, which is a measured fact about the data and may well be true. The failure is in the step. To get from 1 to 2 you need a further claim: that nothing else explains both numbers at once. And something obvious does. Rich countries buy more of most things, including chocolate, and they also fund universities, laboratories and the careers that lead to Nobel prizes. National wealth can raise both figures without a single cocoa bean affecting a single thought.

Notice how the diagnosis was made. Not by disliking the conclusion, and not by shouting "correlation is not causation" as though the phrase were a spell. It was made by naming the missing premise and then producing a specific rival explanation that makes the missing premise false. That is what all fallacy-hunting is, underneath. The names are just filing labels for mistakes that keep recurring.

The upshot: To debug an argument, find the step where the support does not reach, then say what would have to be true for it to reach. The label comes last.

What a fallacy is, and what it is not

A fallacy is a pattern of bad reasoning common enough to be worth a name. Two things it is not. It is not a false statement: "the Earth has two moons" is false but not a fallacy, because it is not an inference. And it is not merely an argument you dislike. If you cannot say which step fails and why, you have not found a fallacy; you have found a disagreement.

Traditionally fallacies are sorted into two families. Formal fallacies are broken shapes: affirming the consequent, from the last lesson, is one, and it is invalid no matter what you plug in. Informal fallacies are failures of relevance, evidence or language, where the shape may be fine but something else has gone wrong. Most of what follows is informal, which is also why it is harder: whether an informal fallacy has been committed often depends on the content and the context, not on the pattern alone.

The catalogue, with a real case for each

Read the middle column first and try to spot the trouble before reading the right-hand column.

Remember: Each row has the same structure: a premise that would be relevant to something, attached to a conclusion it does not reach. Fix on the gap, not the vocabulary.

Two of these are often wrongly accused

Fallacy labels are thrown around more confidently than they deserve. Take two.

Slippery slopes are sometimes good arguments. "If the school allows this one exception to the deadline, it will have to allow the next, because the reason given applies equally to every student with a part-time job" is not a fallacy. It supplies the mechanism: the justification generalises. A slope argument is fallacious when the steps are asserted, not when they are defended. So the right criticism is never "that is a slippery slope"; it is "why does step two follow from step one?"

Attacking the source is sometimes legitimate. If a claim rests on someone's testimony, their reliability is part of the evidence. Pointing out that a witness was paid by one side is relevant. It becomes an ad hominem when the argument stands on its own and the attack is used to avoid it. The test: does the criticism target the support, or the person offering it?

There is also a wider unease among specialists about the whole enterprise. Hans Hansen's survey of fallacy theory for the Stanford Encyclopedia points out that the standard lists are inherited from Aristotle by way of centuries of textbooks, that the usual definition of a fallacy as an argument that seems valid but is not fits some entries badly, and that agreement on how to individuate fallacies is still missing. You should learn the list, because it is genuinely useful shorthand. You should not treat it as a finished science.

The fallacy fallacy

Here is a trap that catches good students. Suppose someone argues for a conclusion badly:

(1) Vaping is bad for you, because everyone knows it is.
(2) So vaping is bad for you.

The appeal to popularity is real, and the argument fails. What follows about vaping? Nothing. The conclusion may well be true on other grounds, and the medical evidence is where you would go next. Treating a bad argument for P as though it established not-P is itself an error, sometimes called the fallacy fallacy.

This matters for how you write. When you identify a fallacy, say exactly what you have shown, which is that this route to the conclusion is blocked. Then, if you are being honest, ask whether a better route exists. Some of the strongest philosophy papers spend their first page dismantling a bad argument for a view and their second page building a good one for the same view.

Worth holding on to: Refuting an argument is not the same as refuting a claim. Say which one you have done.

A worked debug, step by step

A letter to a local paper, of the kind that appears every spring: "The council wants to spend 200,000 on a new skate park. But we already have a youth centre that nobody uses. Young people today are not interested in facilities, they are interested in their phones. Anyone who claims otherwise has not been near a bus stop lately. And if we say yes to this, every interest group in town will be queueing up for their own building."

Four moves, four diagnoses.

1. The youth centre nobody uses. This could be relevant evidence: if a similar facility failed, that is data about demand. But it is offered as decisive, and it is not. Skateboarding and a youth centre draw different people. The arguer needs the premise that usage of one predicts usage of the other, and that premise is doing all the work while remaining unstated.
2. Young people are only interested in phones. A generalisation from casual observation to a claim about a whole age group. What would test it? Attendance figures at the existing skate park two towns over, or a survey with a real sample. Bus stops are where you see people waiting, which is not a random sample of anything.
3. Anyone who claims otherwise has not been near a bus stop. The disagreement is reframed as ignorance on the opponent's part. No evidence has been added. This is the ad hominem move in its politest form.
4. Every interest group will queue up. A slope, and an unsupported one. It could be repaired: if the council has no criteria for funding facilities, then granting this one really does create a precedent with no stopping point. Supply that mechanism and the fourth move becomes the letter's best argument. Leave it asserted and it is the weakest.

Notice the letter is not stupid. Three of its four moves point at something real. The failure is that each stops one premise short of doing its job, and the fourth could be rescued by an argument the writer did not make.

Common misconceptions

• "Naming the fallacy wins the argument." It shows one route to the conclusion is blocked. The claim itself may still be true, and a serious opponent will simply offer a better argument.
• "Any appeal to an expert is an appeal to authority fallacy." Relying on relevant expertise is normally reasonable and often unavoidable. The fallacy is using authority outside its field, or treating it as proof rather than as evidence.
• "Slippery slope arguments are automatically fallacious." They are fallacious when the links are asserted without reason. A slope with a stated mechanism is an ordinary causal argument to be judged on its evidence.
• "Correlation never tells you anything about causes." It is real evidence, and often the only evidence available. It does not by itself distinguish between a cause, a reverse cause, and a third factor behind both, which is why the extra premise has to be argued for.
• "If someone is biased, their argument fails." Bias gives you a reason to check the work, not a reason to skip it. An argument's premises and structure are visible to anyone, whatever the arguer wants.

What to carry forward

• Debug an argument by locating the step where support stops reaching, then naming the premise that would close the gap.
• A fallacy is a recurring pattern of bad inference, not a false statement and not an unpopular view.
• Formal fallacies are broken shapes. Informal ones depend on content and context, which is why they are contested.
• The standard catalogue is useful shorthand: ad hominem, straw man, false dilemma, slippery slope, begging the question, misused authority, post hoc, equivocation, hasty generalisation, appeal to popularity.
• Slippery slope and source criticism are legitimate when a mechanism or a reliability claim is actually supplied.
• Fallacy theory itself is unsettled among specialists, so use the labels as shorthand rather than as verdicts.
• The fallacy fallacy: a bad argument for a claim leaves the claim exactly where it was.

Sources

1. Hansen, H. (2024). Fallacies. In E. N. Zalta & U. Nodelman (Eds.), Stanford Encyclopedia of Philosophy. plato.stanford.edu
2. Internet Encyclopedia of Philosophy. (n.d.). Fallacies. University of Tennessee at Martin. iep.utm.edu
3. Messerli, F. H. (2012). Chocolate consumption, cognitive function, and Nobel laureates. New England Journal of Medicine, 367(16), 1562-1564. Correlation coefficient of 0.791 reported for chocolate consumption against laureates per capita; paywalled, bibliographic details confirmed on PubMed.

Two men outside a courthouse in Athens, 399 BCE

A man is waiting outside the office of the king-archon in Athens because he has been indicted. His name is Socrates, he is about seventy, and within weeks he will be sentenced to death. While he waits, he meets an acquaintance called Euthyphro, who is also there on legal business. Euthyphro is prosecuting his own father. A labourer on the family estate killed a slave; Euthyphro's father bound the man, threw him in a ditch, and sent to Athens for a ruling, and the man died in the ditch before the answer came. The family thinks Euthyphro is monstrous for taking his father to court. Euthyphro thinks he is being pious.

That is the setup of the Euthyphro, a dialogue of about twenty pages written by Plato, who was in his twenties when Socrates died. Everything quoted in this lesson comes from the translation made by Benjamin Jowett in the nineteenth century, which is in the public domain and free to read in full.

Two facts about the scene are worth holding on to, because they explain the whole dialogue. Socrates is about to be tried for impiety. And here is a man who claims to know exactly what piety is.

What Socrates actually asks for

Watch the request. Socrates does not ask Euthyphro for examples of pious actions, and he does not ask whether prosecuting his father is pious. He asks for something harder.

I may have a standard to which I may look, and by which I may measure actions, whether yours or those of any one else, and then I shall be able to say that such and such an action is pious, such another impious.

A standard by which to measure actions. Not a list, a criterion: something that tells you, for any action you like, whether it belongs in the pious pile. This is the shape of a huge number of philosophical questions. What makes an action right? What makes a sentence true? What makes a person the same person over time? Each asks for the standard rather than the list, and the reason is practical. A list runs out. A standard does not.

Key idea: When a philosopher asks what something is, they are usually asking for the criterion that decides every case, not for examples.

How to read a page of this, slowly

Primary texts are slower to read than summaries and worth the time, because a summary tells you the conclusion while the text shows you the argument. Here is a four-step pass you can use on any philosophical passage, including the ones later in this course.

1. Find the question. Write it as a question, in your own words, in one sentence. In the Euthyphro: what is piety?
2. Find the answer on offer. Who proposes it, and in what words? Euthyphro proposes that piety is what is dear to the gods.
3. Find the argument against it. This is where most first readings fail. Do not summarise the objection as a mood; track the steps. Which premise does the objection attack, and what does it offer in place of it?
4. Find what is left standing. Many dialogues end without an answer. Ask what has been ruled out, which is real progress, and what question the failure opens.

Now run the pass on the dialogue.

Euthyphro's answers, and how each one breaks

Euthyphro offers three definitions in turn, and they get better.

First: piety is doing as he is doing, prosecuting a wrongdoer. Socrates points out this is an example, not a standard. Marked as failed on step 1 of the request.

Second: piety is what is dear to the gods. Socrates observes that the Greek gods quarrel constantly, and about exactly the things people quarrel about, which are questions of justice and honour. If Athena approves of an act and Ares hates it, the definition makes the same act both pious and impious.

Third: Euthyphro repairs it. Piety is what all the gods love; impiety is what they all hate. That closes the leak. And it is at this point, with a genuinely improved definition on the table, that Socrates asks the question the dialogue is remembered for.

The point which I should first wish to understand is whether the pious or holy is beloved by the gods because it is holy, or holy because it is beloved of the gods.

Read it twice. It is an either-or about the direction of explanation. Does the holiness come first, so that the gods love it for a reason? Or does the loving come first, so that being loved is what holiness consists in? Euthyphro says he does not understand the question, which is honest of him, so Socrates explains it with an analogy about things carried and things seen: a thing is not seen because it is visible, but is visible because it is seen. The state of being loved follows the act of loving, not the reverse.

Then he asks Euthyphro to choose, and Euthyphro chooses the first option.

SOCRATES: It is loved because it is holy, not holy because it is loved?
EUTHYPHRO: Yes.

And with that, Socrates shows, the definition has collapsed. If the gods love holy things because they are holy, then holiness is a property those things already have, and divine love is a consequence of it, not the thing itself. Being loved by all the gods is then a reliable sign of piety, like a thermometer reading is a reliable sign of a fever, and no more the same thing as piety than the reading is the same thing as the fever. In Jowett's rendering: "when I ask you what is the essence of holiness, to offer an attribute only, and not the essence, the attribute of being loved by all the gods."

Euthyphro's reply is the best line in the dialogue and every student recognises the feeling.

I really do not know, Socrates, how to express what I mean. For somehow or other our arguments, on whatever ground we rest them, seem to turn round and walk away from us.

He then remembers an appointment and leaves. The dialogue ends with no definition of piety.

Why the dilemma still bites

Strip out the Greek gods and the Euthyphro dilemma becomes a live question for anyone who grounds morality in a divine will, a doctrine usually called divine command theory. Ask: is an action right because God commands it, or does God command it because it is right? Both answers cost something, and the course takes no position on which cost is worth paying.

Take the first horn, that the commanding makes it right. Then the content of morality is settled by the command and nothing else, so had the command been different, the morality would have been different. Cruelty to children would have been right if commanded. Defenders reply that a perfectly good God would never command such a thing, and that the objection smuggles in a standard of goodness independent of God in order to complain. That reply has force, and it is also close to conceding the second horn.

Take the second horn, that God commands it because it is right. Then rightness is what it is independently, and God recognises it rather than creating it, which many theists find perfectly acceptable and others see as demoting God from author to expert witness. Some philosophers and theologians take a third route, identifying goodness with God's nature rather than with God's commands, so that the question of whether God could have commanded cruelty does not arise.

What you should take from this is a technique, not a verdict. Socrates built a two-sided trap by asking about the direction of a dependence. You can build the same trap elsewhere. Is an action illegal because the legislature forbade it, or did the legislature forbid it because it was already wrong? Is a painting good because experts admire it, or do they admire it because it is good? The dilemma is one of philosophy's reusable tools.

The point: Ask which way the explanation runs. Many confident definitions turn out to have picked a symptom and called it the cause.

Weeks later, the same courthouse

The trial happens. Plato's Apology is his version of the defence speech, and the word means defence, not regret. The charge, as the dialogue reports it, is that Socrates is "an evil-doer and corrupter of the youth, who does not receive the gods whom the state receives, but introduces other new divinities."

Socrates spends much of the speech on the source of his reputation. He says a friend once asked the oracle at Delphi whether anyone was wiser than Socrates, and was told that no one was. Socrates, who insists he knows nothing worth knowing, treats this as a puzzle to be solved by investigation: find a wiser man and the oracle is refuted. So he questions the people with reputations for wisdom, starting with a politician.

Well, although I do not suppose that either of us knows anything really beautiful and good, I am better off than he is, for he knows nothing, and thinks that he knows; I neither know nor think that I know. In this latter particular, then, I seem to have slightly the advantage of him.

This is the famous claim, and notice how modest it actually is. Socrates does not claim to know more facts. He claims an accurate view of his own ignorance, which he treats as the only wisdom available to him. The word for the practice built on it is the Socratic method: question a confident person about what they claim to know until either the claim survives or it walks away, as Euthyphro's did.

This also explains the trial. Going around a small city proving that its respected men cannot define what they profess is not a way to be popular. The Apology records, in Socrates' own account, that each examination made him enemies.

In the penalty phase, after the vote goes against him, someone suggests he could simply leave Athens and keep quiet. He says he cannot, and gives his reason.

if I say again that daily to discourse about virtue, and of those other things about which you hear me examining myself and others, is the greatest good of man, and that the unexamined life is not worth living, you are still less likely to believe me.

That sentence, usually quoted as five words on a poster, is in the text an argument about a penalty. He is explaining why exile with silence is not a lesser punishment than death but a greater one, given what he takes a human life to be for. Whether he is right is a question this course returns to in the lesson on the good life. Notice, though, how much of the sentence the poster version loses.

Common misconceptions

• "Socrates wrote the dialogues." He wrote nothing we have. Plato, Xenophon and Aristophanes are our sources, and they disagree, so scholars distinguish the historical Socrates from Plato's character.
• "Socrates says he knows nothing, so he is being modest or ironic." The claim in the Apology is specific: he does not claim knowledge he lacks, unlike those he questions. That is a claim about self-knowledge, and he acts on it consistently enough to die for it.
• "The Euthyphro proves there is no God, or that religion cannot ground morality." It proves neither. It shows that one definition of piety fails, and it sets a dilemma that theists have answered in several ways, including by locating goodness in God's nature rather than in commands.
• "A dialogue that reaches no definition has failed." Ruling out answers is progress, and knowing why an attractive answer fails is often more useful than a definition you cannot defend.
• "Apology means he apologised." It translates a Greek word for a legal defence. He does not concede the charges, and the speech is not conciliatory.

The short version

• Socrates asks for a standard that decides every case, not for examples. Most philosophical what-is questions have that form.
• Read a primary text by finding the question, the answer on offer, the argument against it, and what remains standing.
• Euthyphro's definitions fail in order: an example, then a definition the quarrelling gods break, then a repaired version that the dilemma dismantles.
• The dilemma asks whether the pious is loved because it is pious or pious because it is loved. Each horn has a cost, and both are still defended.
• Dependence questions of that form are a reusable tool: law and morality, expert taste and artistic value, popularity and truth.
• In the Apology, Socrates claims only that he does not think he knows what he does not know, and treats that as the wisdom the oracle meant.
• The line about the unexamined life is part of an argument about why exile would be a worse penalty than death, not a slogan.

Sources

1. Plato. (1999). Euthyphro (B. Jowett, Trans.; Project Gutenberg eBook No. 1642). Project Gutenberg. gutenberg.org
2. Plato. (1999). Apology (B. Jowett, Trans.; Project Gutenberg eBook No. 1656). Project Gutenberg. gutenberg.org
3. Nails, D., & Monoson, S. S. (2022). Socrates. In E. N. Zalta & U. Nodelman (Eds.), Stanford Encyclopedia of Philosophy. plato.stanford.edu
4. Woodruff, P. (2022). Plato's shorter ethical works. In E. N. Zalta & U. Nodelman (Eds.), Stanford Encyclopedia of Philosophy. plato.stanford.edu

A procedure for destroying everything you believe

In 1641 a French soldier turned mathematician, living in the Netherlands, published a short book in Latin: Meditationes de prima philosophia. Rene Descartes was forty-four. The Meditations is written in the first person, as six days of thinking, and the first day is an act of demolition. All quotations here come from John Veitch's translation, made in 1853 and long out of copyright.

Descartes sets out the plan in the first paragraphs. He is not going to examine his beliefs one at a time, because there are too many. He is going to go after the foundations, on the principle that if the bottom of a wall goes, the wall goes with it.

as the removal from below of the foundation necessarily involves the downfall of the whole edifice, I will at once approach the criticism of the principles on which all my former beliefs rested.

He also sets the bar deliberately high. He will reject a belief not only if it is false but if there is any ground for doubting it at all. That is a strange standard for ordinary life and a perfectly sensible one for his purpose, which is to find out whether anything at all is certain. Treat what follows as a procedure with stages, and watch what each stage removes.

Stage one: the senses have caught me out before

All that I have, up to this moment, accepted as possessed of the highest truth and certainty, I received either from or through the senses. I observed, however, that these sometimes misled us; and it is the part of prudence not to place absolute confidence in that by which we have even once been deceived.

As an argument:

(1) Everything I count as most certain came through the senses.
(2) The senses have deceived me at least once.
(3) It is imprudent to trust completely anything that has deceived you once.
(4) So I should not trust the senses completely.

This stage is weaker than it looks, and Descartes says so himself. He immediately raises the objection: the senses mislead about small or distant things, but surely not about whether I am sitting here by the fire in a winter dressing gown holding this piece of paper. To deny that, he says, would put him in the company of people who believe their heads are made of clay. Stage one removes the far and the faint. It leaves the near and the obvious.

Stage two: the dream argument

So he escalates, and here the argument gets its teeth.

How often have I dreamt that I was in these familiar circumstances, that I was dressed, and occupied this place by the fire, when I was lying undressed in bed? At the present moment, however, I certainly look upon this paper with eyes wide awake; the head which I now move is not asleep; I extend this hand consciously and with express purpose, and I perceive it; the occurrences in sleep are not so distinct as all this. But I cannot forget that, at other times I have been deceived in sleep by similar illusions; and, attentively considering those cases, I perceive so clearly that there exist no certain marks by which the state of waking can ever be distinguished from sleep, that I feel greatly astonished; and in amazement I almost persuade myself that I am now dreaming.

Laid out, the dream argument runs:

(1) I have had dreams in which I seemed to be doing exactly what I now seem to be doing.
(2) In those dreams I could not tell that I was dreaming.
(3) There is no mark present in waking experience and absent from dreaming.
(4) So no experience I am having can establish that I am awake.
(5) So beliefs based on present experience are not certain.

Where would you attack it? Premise 3 is the one most philosophers press. Dreams are, arguably, less coherent, less continuous, less detailed under scrutiny. Try reading a page of text in a dream. But notice the difficulty in using that reply: to know that the present experience has the coherence that waking has, you would have to already trust the present experience. The reply seems to need what it sets out to prove.

What survives stage two? Descartes thinks something does. Even in a dream, three plus two is five, and a square has four sides. The dream can put a false scene in front of you, but not, apparently, a false sum.

What matters here: Each stage of the doubt is aimed at a specific source of belief. Stage two removes what the senses report. It leaves reasoning untouched.

Stage three: the malignant demon

So Descartes builds a doubt aimed at reasoning itself.

I will suppose, then, not that Deity, who is sovereignly good and the fountain of truth, but that some malignant demon, who is at once exceedingly potent and deceitful, has employed all his artifice to deceive me; I will suppose that the sky, the air, the earth, colors, figures, sounds, and all external things, are nothing better than the illusions of dreams, by means of which this being has laid snares for my credulity; I will consider myself as without hands, eyes, flesh, blood, or any of the senses, and as falsely believing that I am possessed of these.

The evil demon is not a claim that such a being exists. It is a device: a hypothesis designed to be consistent with every piece of evidence you have, so that no evidence can rule it out. If a sufficiently powerful deceiver is arranging your experiences and interfering with your calculations, then even the sum might be wrong every time you check it, and you would notice nothing.

That is why the demon is stronger than the dream. The dream attacks what you see. The demon attacks what you work out. After stage three, Descartes has no senses he trusts, no body he is sure of, and no arithmetic he can lean on.

The one thing the demon cannot arrange

Then, in the second Meditation, the demon hypothesis turns on itself.

Doubtless, then, I exist, since I am deceived; and, let him deceive me as he may, he can never bring it about that I am nothing, so long as I shall be conscious that I am something. So that it must, in fine, be maintained, all things being maturely and carefully considered, that this proposition, I am, I exist, is necessarily true each time it is expressed by me, or conceived in my mind.

The structure is simple and worth admiring. Deception needs someone deceived. Doubt needs someone doubting. So the more thoroughly the demon works, the more certain it is that there is something being worked on. The certainty is not general and permanent, and Descartes is careful about this: the proposition is necessarily true "each time it is expressed by me, or conceived in my mind". It is certain while you are thinking it. It does not license a claim about yourself last Tuesday.

What is this thing that is certain to exist? Not a body, since bodies are still in doubt. Descartes' answer, at this point in the argument, is deliberately thin.

What is a thinking thing? It is a thing that doubts, understands, conceives, affirms, denies, wills, refuses; that imagines also, and perceives.

Notice how much is not claimed. Nothing yet about the soul surviving death, nothing about brains, nothing about other people. Just: there is thinking going on, and that makes something that thinks. Whether even that much follows is the question we come to below.

The core of it: The cogito is not a proof that a self exists forever. It is the observation that doubting cannot get started without a doubter, and it holds only while the doubting is happening.

The wax, and why Descartes trusts the mind over the eye

The second Meditation ends with an example so good it has outlived the argument it serves. Descartes picks up a piece of wax, fresh from the hive: it smells of flowers, it is hard and cold, it has a colour and a shape and it makes a sound when tapped. Then he puts it near the fire.

what remained of the taste exhales, the smell evaporates, the color changes, its figure is destroyed, its size increases, it becomes liquid, it grows hot, it can hardly be handled, and, although struck upon, it emits no sound. Does the same wax still remain after this change? It must be admitted that it does remain; no one doubts it.

Every property the senses reported has changed. The judgement that it is the same wax has not. So whatever you were tracking when you knew it was wax, it was not the smell, the colour, the hardness or the sound. Descartes concludes that what remains is "something extended, flexible, and movable", and that this is grasped not by the eye or the hand but by the mind alone.

You can accept the observation and resist the conclusion. A modern critic might say: what you are tracking is a continuous physical object through a change of state, and children learn to do that from experience, not by pure reason. Descartes would reply that experience of this piece of wax cannot teach you that it could take an infinity of other shapes, and that this is exactly what your concept of wax already includes. The disagreement is real and the argument is a fair one to have.

Running the procedure on a new input

Here is the value of treating this as a procedure rather than a story. Change the input and see what the stages do.

Input: a video of a politician saying something outrageous. Stage one asks whether the medium has deceived you before, and video now certainly has. Stage two asks whether there is a mark that separates real footage from synthetic footage, and for a careful forgery there may be none available to your eyes. Stage three asks whether there is any evidence at all that a well-resourced deceiver could not also produce. That is a serious question about provenance, and it explains why the practical answer is not better eyes but a chain of custody: who recorded it, on what device, who published it, what else corroborates it.

Notice what changed. Descartes' demon was unfalsifiable by design, so his doubt could only be answered from inside his own mind. A deepfake is produced by people, which leaves traces, so the doubt is answerable by evidence outside the video. That difference, between a doubt nothing could settle and a doubt that better evidence could settle, is the distinction the next lesson turns on.

Where the argument is attacked

Three standard objections, each worth knowing.

The I is smuggled in. All that the reasoning strictly delivers is that thinking is occurring. Getting from "there is thinking" to "there is an I that thinks" adds something: that thoughts require a single continuing owner. The eighteenth-century physicist and aphorist Georg Lichtenberg made the point by suggesting the honest report would be closer to "it thinks", in the way one says "it is raining".

The circle. Later in the Meditations Descartes argues that God exists and would not allow him to be deceived about what he perceives clearly and distinctly, which is how he escapes the demon and recovers the external world. Critics from his own century onward have objected that he uses clear and distinct perception to prove God, and God to validate clear and distinct perception. Whether the charge sticks depends on details of what he claims at each step, and it remains a live scholarly dispute.

The doubt may be unusable. If you genuinely suspend belief in everything, you cannot conduct an investigation, because investigating uses memory, language and inference. Descartes' own writing needs a shared language and a reader. His defence is that the doubt is a method, adopted for a purpose and set aside afterwards, not a state of mind he actually lived in.

Common misconceptions

• "Descartes believed in the evil demon." He did not. It is a hypothesis constructed to be immune to evidence, in order to test which beliefs could survive the worst case.
• "Cogito ergo sum is the climax of the Meditations." The Latin phrase appears in the earlier Discourse on the Method. In the Meditations the wording is "I am, I exist", and it is qualified as true each time it is thought.
• "Descartes concluded that nothing can be known." The opposite. He doubts in order to find something indubitable and then rebuild, which is why his position is not scepticism but a response to it.
• "The dream argument shows you might be dreaming right now, so nothing matters." It shows that present experience cannot certify itself. It says nothing about what you should do, and Descartes himself continued to eat dinner.
• "The wax example proves objects are only in the mind." It argues that what you grasp about the wax is grasped by the intellect rather than the senses. The wax is still a physical thing.

Putting it together

• Descartes attacks foundations rather than individual beliefs, and rejects anything open to any doubt at all.
• Stage one removes distant and faint sensory reports. Stage two, the dream argument, removes present experience by denying any mark that separates waking from sleep.
• Stage three, the demon, is built to be consistent with all evidence, which extends the doubt to reasoning itself.
• The cogito survives because deception and doubt both require someone undergoing them. It is certain only while being thought.
• The thinking thing is defined thinly: it doubts, understands, affirms, denies, wills, imagines, perceives. Nothing about bodies or souls yet.
• The wax argues that what stays constant through total sensory change is grasped by the mind, not the senses.
• The main objections: the I may be smuggled in, the later appeal to God may be circular, and a total doubt may be unusable in practice.

Sources

1. Descartes, R. (1901). Meditations on first philosophy (J. Veitch, Trans.; first published 1853), Meditations I and II. Wikisource. wikisource.org
2. Newman, L. (2023). Descartes' epistemology. In E. N. Zalta & U. Nodelman (Eds.), Stanford Encyclopedia of Philosophy. plato.stanford.edu
3. Comesaña, J., & Klein, P. (2026). Skepticism. In E. N. Zalta & U. Nodelman (Eds.), Stanford Encyclopedia of Philosophy. plato.stanford.edu

A definition that looks finished

Here is a definition of knowledge that has been taught in schools and universities for a very long time, and that most people produce on their own if you press them. You know that something is the case when three conditions hold.

1. You believe it.
2. It is true.
3. You are justified in believing it: you have good reason, not just a hunch.

Short for justified true belief, it is often called the JTB analysis, and versions of the idea go back to Plato's Theaetetus. It has the shape philosophers want: three conditions, each necessary, together sufficient. Any case of knowledge should satisfy all three, and anything satisfying all three should be a case of knowledge.

This lesson starts from that answer, takes it seriously, and then traces exactly where it fails. The failure is famous, it took one three-page paper to produce, and sixty years later there is still no agreed replacement.

Why each condition is there

Test each condition by imagining it removed.

Drop belief. A student sits an exam, writes that the Danube flows into the Black Sea, and is right, but she wrote it because it looked like the least implausible option and she does not actually believe it. We would not say she knew. Knowing involves holding the thing to be so.

Drop truth. In 1900 a great many well-informed doctors were confident that ulcers were caused by stress and acid. They believed it and they had reasons. They did not know it, because it was wrong, and the discovery that a bacterium was often responsible won a Nobel Prize in 2005. This condition is what makes knowledge different from confidence: you cannot know something false.

Drop justification. Someone with no information guesses that the coin will land heads, and it does. True belief, no knowledge. Guessing right is not knowing, which is why a weather forecast that is correct by accident is not a forecast anyone should use again.

So all three conditions are doing real work, and the analysis is not silly. It fails anyway.

Why this matters: Testing a definition by deleting one part at a time is a technique, not a formality. It tells you what each part is for, which is what you need before you can say which part has gone wrong.

The three-page paper

In 1963 Edmund Gettier published an article in the journal Analysis under the title "Is Justified True Belief Knowledge?". His answer, in two examples and barely three pages, was no. His paper is still in copyright, so here is the first case in plain words rather than his.

Smith and Jones have both applied for the same job. Smith has excellent evidence that Jones will get it: the company president has told him so directly. Smith has also, for reasons of his own, counted the coins in Jones's pocket, and there are ten. So Smith forms a belief:

The man who will get the job has ten coins in his pocket.

That belief is well justified. It follows from two things Smith has good evidence for. Now add the twist. Smith gets the job, not Jones. And Smith, entirely unaware of it, happens to have ten coins in his own pocket.

Run the three conditions. Smith believes it: yes. It is true: yes, the man who got the job does have ten coins in his pocket, because that man is Smith. Smith is justified: yes, by the president's testimony and his own counting. All three conditions hold. And nobody thinks Smith knew. He was right by accident, in a way that had nothing to do with his reasons.

That is a Gettier case: a counterexample of exactly the kind you learned to build in the first lesson of this course. The premises of the definition are satisfied and its conclusion fails.

The obvious repair, and the case that kills it

The first thing anyone tries is to find what went wrong inside Smith's reasoning, and something did. He reached his true belief by way of a false one: that Jones would get the job. So add a fourth condition. Knowledge is justified true belief whose justification passes through no falsehood. This is the no-false-lemmas repair, and for Gettier's own two cases it works.

Now here is the case that defeats it, from the 1970s and usually credited to Carl Ginet, by way of Alvin Goldman. You are driving through the countryside and see, clearly and in good light, a barn beside the road. You form the belief that there is a barn there. There is. Your eyes work, the light is good, you have no false belief anywhere in your reasoning.

What you do not know is that this district is full of barn facades: painted fronts propped up to look like barns from the road, for reasons of local tourism. Dozens of them. The one you happened to look at is the only real barn for miles.

Justified: yes. True: yes. Believed: yes. No falsehood in the chain: correct, there is none. Knowledge? Almost everyone says no, and for the same reason as before. You would have formed the same belief looking at any of the facades, so your being right was luck.

The point: Gettier cases are not a trick about false premises. They are about a mismatch between why you believe and why the belief is true.

The diagnosis, and four ways to act on it

Line up the two cases and the common feature is visible. In each, the reason the belief is true has nothing to do with the reason the person holds it. Smith's justification concerned Jones; the truth concerned Smith. Your justification concerned how that structure looked; the truth concerned the one building in the county that was not a facade. Knowledge seems to require that the belief be true because of the way it was formed, not merely alongside it.

Four families of repair follow from that diagnosis. None commands general agreement, and each is defended by serious philosophers today.

There is also a fifth response, which is to reject the project. Timothy Williamson has argued at length that knowledge is not analysable into belief plus a list of extra conditions at all, and that it should be treated as basic, with belief and evidence explained in terms of it rather than the other way round. If sixty years of counterexamples have not produced an analysis, that is at least a reasonable inference to draw.

Notice what has happened to the state of play. Nobody thinks JTB is sufficient. Almost everyone still thinks each of its three conditions is necessary. The open question is what the fourth ingredient is, or whether there is one.

Why this is not only a seminar puzzle

Three places where the same structure shows up with something at stake.

Eyewitness identification. A witness picks the right person from a line-up. Suppose the line-up was badly constructed, so that the witness would have picked whoever stood in that position. The identification is true and the witness is sincere, and yet the procedure did not make the truth of it likely. Courts and researchers care about exactly this, which is why identification procedures are governed by rules about how the line-up is built.

A medical test with a bad base rate. A test says a patient has a condition, and the patient has it. If the test returns positive for almost everyone, the correct result was not produced by anything that tracks the condition. The belief is true; the process does not deserve credit.

An answer from a language model. Ask a chatbot for the year a treaty was signed and it returns the right year. Did it know? The question is not mystical: it is whether the process that produced the string is one that tracks the historical record, or one that produces plausible strings that are often right and sometimes confidently wrong. Gettier's structure is exactly the right tool for asking it, and the answer matters when the same system is asked about a medicine dose.

Common misconceptions

• "Gettier showed that knowledge is impossible." He showed that one definition is not sufficient. Ordinary knowledge is not in doubt; the account of what it consists in is.
• "A Gettier case is just a case of bad evidence." The evidence is good. That is what makes the cases work. If Smith had no evidence, the justification condition would fail and there would be no puzzle.
• "Justified means proved." Justification is a matter of having good reason, which is compatible with being wrong. Demanding proof would make almost nothing count as knowledge.
• "The right answer is just to add a fourth condition." Several fourth conditions have been proposed and each faces counterexamples. Some philosophers now argue the analysis cannot be completed.
• "If a belief is true and you have reasons, the reasons must be what made it true." The Gettier cases are precisely the cases where that comes apart, and they are not rare enough to ignore.

What you now know

• The traditional analysis: knowledge is justified true belief, with each condition doing separate work.
• Removing belief, truth or justification in turn produces cases nobody counts as knowledge, so all three are necessary.
• Gettier's 1963 cases satisfy all three conditions and are not knowledge, because the reasons and the truth come apart.
• The no-false-lemmas repair handles Gettier's own cases and fails on fake barns, where no falsehood appears.
• Reliabilism, defeasibility and safety each add a condition about how the belief was formed, and each faces the problem of saying which circumstances count.
• Williamson's alternative is to stop analysing knowledge and treat it as basic.
• The same structure matters in line-ups, diagnostic tests and machine answers: a true output from a process that does not track the truth.

Sources

1. Ichikawa, J., & Steup, M. (2026). The analysis of knowledge. In E. N. Zalta & U. Nodelman (Eds.), Stanford Encyclopedia of Philosophy. plato.stanford.edu
2. Hetherington, S. (n.d.). Gettier problems. Internet Encyclopedia of Philosophy. iep.utm.edu
3. Gettier, E. L. (1963). Is justified true belief knowledge? Analysis, 23(6), 121-123. The original three-page counterexample paper; in copyright and described here rather than quoted.
4. Williamson, T. (2000). Knowledge and its limits. Oxford University Press. Chapter 1 argues against analysing knowledge into belief plus further conditions.

A paragraph from 1740 that will not go away

Near the end of a section of his Treatise of Human Nature, David Hume slips in an observation that he introduces almost as an afterthought.

In every system of morality, which I have hitherto met with, I have always remarked, that the author proceeds for some time in the ordinary way of reasoning, and establishes the being of a God, or makes observations concerning human affairs; when of a sudden I am surprized to find, that instead of the usual copulations of propositions, is, and is not, I meet with no proposition that is not connected with an ought, or an ought not. This change is imperceptible; but is, however, of the last consequence. For as this ought, or ought not, expresses some new relation or affirmation, it is necessary that it should be observed and explained; and at the same time that a reason should be given, for what seems altogether inconceivable, how this new relation can be a deduction from others, which are entirely different from it.

Hume's complaint is procedural and it is the complaint from your first lesson. A conclusion has appeared that the premises do not contain. All the premises were about what is the case; the conclusion is about what ought to be. Somewhere a premise went unstated. This is now called the is-ought problem, and it sits underneath a dispute that is very much alive: how far can scientific evidence settle a question about what we should do?

Two positions, both held by serious people, and this lesson gives each its best case without choosing between them.

First, what science settles extremely well

In the early 1980s the consensus about stomach ulcers was that they were caused by stress and excess acid. Bacteria, everyone agreed, could not survive in the stomach. Two Australians, Robin Warren and Barry Marshall, kept finding curved bacteria in biopsy samples and argued the organism caused the disease. Marshall, unable to get an animal model, drank a culture of it and developed gastritis. In 2005 the pair received the Nobel Prize in Physiology or Medicine for the discovery of Helicobacter pylori and its role in gastritis and peptic ulcer disease. Treatment changed from managing a chronic condition to a course of antibiotics.

That story shows the scientific method doing what nothing else does. A claim about the world was made precise enough to be wrong, tested, and forced through a hostile consensus by evidence. Notice the sequence: a specific prediction, a check that could have gone the other way, a result, a change of mind.

Karl Popper made that sequence the mark of science itself. On his account, what distinguishes a scientific theory is not that evidence supports it, since evidence can be found for almost anything, but that it forbids something. A theory is scientific to the extent that it is falsifiable: it says what will not be observed, and sticks its neck out. Einstein's general relativity predicted that starlight would bend by a specific amount near the sun, which is a claim that could have failed. A theory that fits every possible outcome tells you nothing.

In short: Science earns its authority by making claims that could have turned out false, and then checking.

Where the boundary gets hard

Falsifiability is a clean criterion, and philosophers of science have spent decades showing that the clean version does not quite work.

A failed prediction never falsifies just one thing. Any test uses the theory plus a pile of assumptions: the instrument works, the sample is representative, the background conditions are as assumed. When a prediction fails, logic alone does not tell you which member of that pile to blame. When Uranus moved wrongly, astronomers did not abandon Newtonian gravity; they proposed an unseen planet, and found Neptune. When Mercury moved wrongly, the same move was tried, no planet was found, and eventually the theory did give way. Both responses are legitimate, which means the difference between honest adjustment and protecting a theory from refutation is a matter of judgement rather than logic.

The lines fall in odd places. As Sven Ove Hansson's survey of the demarcation problem notes, criteria strict enough to exclude astrology often exclude parts of respectable science, while criteria loose enough to include everything scientists actually do let in things nobody wants. Most philosophers of science now treat the boundary as a family of markers rather than one test: testability, openness to correction, published methods, independent replication, and a willingness to let results change conclusions.

Underdetermination. Kyle Stanford's survey of the problem sets out the harder version: for any body of evidence, more than one theory may fit it, sometimes including theories nobody has thought of yet. Scientists choose between rivals using considerations like simplicity and consistency with other theories, and those considerations are not themselves readings on an instrument. That does not make science arbitrary. It does mean that the picture of data compelling one unique answer is too simple.

The dispute: can evidence settle what we ought to do?

Position A, in its strongest form. Moral questions are questions about how things go for people and other sentient creatures. Whether a policy reduces suffering, whether a punishment deters, whether a drug works, whether a school rule helps students learn: all of these are open to measurement, and all of them are what moral disputes turn out to be about once you clear away the confusion. Defenders of this position, associated with moral naturalism in philosophy, argue that Hume's gap is narrower than it looks, because once we agree that suffering is bad, the rest is investigation. Nobody thinks health is a matter of opinion, and health is not a purely physical notion either.

They add a sharp point: people who insist on the is-ought gap almost always make moral claims anyway, and those claims turn out to depend on empirical beliefs. Someone who says corporal punishment is wrong is relying on facts about what it does to children. Change the facts and the moral verdict moves. That looks like evidence settling a moral question.

Position B, in its strongest form. Look again at the bridging premise. "Suffering is bad" is not the result of an observation; it is a value judgement, and a fairly deep one. You can measure suffering precisely and still face the question of how much weight to give it against liberty, or fairness, or the claims of people not yet born. Two honest researchers can agree on every number about a prison policy and still disagree, because one thinks deterrence matters most and the other thinks proportional punishment does.

Take a real shape of argument. Suppose a study established that a particular punishment deterred a certain crime. It does not follow that the punishment should be used, because a further question remains about what may be done to a person in order to influence others. That question is not empirical, and no better instrument answers it. Defenders of Position B also note that Position A's key move, identifying goodness with wellbeing, is a philosophical claim rather than a finding, and that arguing for it requires exactly the kind of reasoning it claims to replace.

What would move the dispute. Position A would be vindicated by a convincing derivation of a moral conclusion from purely factual premises, with no evaluative premise hidden in the definitions. Position B would be vindicated by a demonstration that every such derivation smuggles a value in, which is what critics have claimed each time one is offered. Both sides have made progress, and the argument is over the status of the bridging premise, not over whether evidence matters.

Bottom line: Evidence constrains moral conclusions tightly and rarely produces them alone. Ask which premise in the argument is doing the evaluating.

Sorting a mixed question

Most real disputes are mixtures. A useful habit is to split a question into three kinds of part before arguing about it.

Doing this has a practical payoff. It stops two people arguing past each other when they actually disagree about the fourth row while trading studies about the first. And it locates where you need evidence and where you need an argument.

Common misconceptions

• "Science proves things." Scientific claims are supported and can be revised; the strongest ones have survived serious attempts to refute them. Proof in the strict sense belongs to mathematics and logic.
• "If a theory has not been falsified, it is confirmed." Surviving tests raises confidence, and how much depends on how demanding the tests were. A theory that forbids nothing survives everything and gains nothing.
• "The is-ought gap means science is irrelevant to ethics." It means factual premises alone do not yield moral conclusions. Almost every moral argument depends on facts, and getting them wrong makes the argument wrong.
• "Philosophy of science is anti-science." The people who raised the demarcation and underdetermination problems were mostly trying to explain why science works so well, and several were practising scientists.
• "A question with no experiment attached has no answer." Whether an argument is valid, what a word means in a statute, whether a punishment is proportionate: these have defensible answers reached by reasoning rather than measurement.

Looking back

• Hume noticed that moral arguments switch from is to ought without a stated premise, which is a structural complaint, not a rejection of morality.
• Science earns its authority by forbidding outcomes and checking, as the Helicobacter pylori case shows against a hostile consensus.
• Falsifiability is the right instinct and an imperfect criterion, because a failed prediction never indicts one assumption on its own.
• Demarcation is now handled by a family of markers rather than a single test, and underdetermination means evidence can leave rival theories standing.
• Position A: moral questions are questions about measurable effects on wellbeing, so evidence largely settles them.
• Position B: the premise linking wellbeing to what ought to be done is itself evaluative, so evidence constrains without deciding.
• Split real questions into empirical, conceptual and normative parts, and argue about the right one.

Sources

1. Hume, D. (1739-1740). A treatise of human nature, Book III, Part I, Section I (Project Gutenberg eBook No. 4705). Project Gutenberg. gutenberg.org
2. Hansson, S. O. (2025). Science and pseudo-science. In E. N. Zalta & U. Nodelman (Eds.), Stanford Encyclopedia of Philosophy. plato.stanford.edu
3. Stanford, K. (2023). Underdetermination of scientific theory. In E. N. Zalta & U. Nodelman (Eds.), Stanford Encyclopedia of Philosophy. plato.stanford.edu
4. The Nobel Prize. (2005). The Nobel Prize in Physiology or Medicine 2005: Barry J. Marshall and J. Robin Warren. Nobel Prize Outreach. nobelprize.org
5. Lutz, M. (2024). Moral naturalism. In E. N. Zalta & U. Nodelman (Eds.), Stanford Encyclopedia of Philosophy. plato.stanford.edu

An iron bar through a foreman's head, Vermont, 1848

At about half past four on the afternoon of 13 September 1848, a railway foreman named Phineas Gage was setting a blast in the rock south of Cavendish, Vermont. The tamping iron in his hands was three feet seven inches long, an inch and a quarter thick, and weighed thirteen and a quarter pounds. It struck a spark from the rock and the powder exploded. The iron entered the left side of his face just in front of the angle of the jaw, passed behind his left eye, destroyed much of the left frontal lobe of his brain, and flew out through the top of his skull.

Gage lived another twelve years, and what made him famous was what happened to his character. His physician, John Harlow, reported that the steady, capable foreman had become fitful, irreverent and given to the grossest profanity, full of plans that were abandoned as soon as they were made, and friends said that he was, for a time at least, "no longer Gage". Historians now think the change has been exaggerated in retelling, and that Gage recovered much of his social skill in later years while driving a stagecoach in Chile. Even the careful version leaves a question standing. An iron bar, a purely physical object, went through a lump of tissue, and what changed was a man's temper, judgement and plans.

What is a mind, if damaging a brain can do that? Three answers have carried most of the argument since the seventeenth century: the mind is a separate, non-physical thing; the mind is the brain; the mind is what the brain does. Each of them can explain Gage, which is why the case, vivid as it is, settles nothing on its own. Start with the strongest case for the first answer, in the words of the man who made it.

Descartes' case for two things

In Lesson 4 you watched Descartes doubt everything until only a thinking thing was left. In the sixth and last of his Meditations, published in 1641, he uses that result to argue that the thinking thing is not the body. The position is called substance dualism: there are two basic kinds of thing, matter, which takes up space, and mind, which thinks and takes up none. Here is the argument in John Veitch's translation.

And, firstly, because I know that all which I clearly and distinctly conceive can be produced by God exactly as I conceive it, it is sufficient that I am able clearly and distinctly to conceive one thing apart from another, in order to be certain that the one is different from the other, seeing they may at least be made to exist separately, by the omnipotence of God ... because, on the one hand, I have a clear and distinct idea of myself, in as far as I am only a thinking and unextended thing, and as, on the other hand, I possess a distinct idea of body, in as far as it is only an extended and unthinking thing, it is certain that I, [that is, my mind, by which I am what I am], is entirely and truly distinct from my body, and may exist without it.

The square brackets appear in Veitch's translation. Set out in numbered lines, the argument runs like this.

(1) Whatever I can clearly and distinctly conceive existing apart could be made to exist apart, by God if by nothing else.
(2) I can clearly and distinctly conceive my mind existing apart from my body, as a thinking thing that takes up no space.
(3) So my mind could exist apart from my body.
(4) If one thing could exist without another, they are not one and the same thing.
(5) So my mind is not my body.

Check the form first, as Lesson 1 taught you. It is valid: if (1) and (2) are true, (3) follows, and (3) with (4) gives (5). Premise (4) is close to a truth of logic; if this pen could exist while that pen did not, they are two pens. So everything rests on (1) and (2), and above all on the step from "I can clearly conceive it" to "it could really be so".

Key idea: The conceivability argument is valid. The whole dispute is about whether being able to conceive the mind without the body shows that the two could really come apart.

The water objection, and Descartes' second argument

The standard objection is that what you can conceive depends on what you know. A chemist in 1700 could clearly conceive of water that was not made of hydrogen and oxygen, since nobody yet knew what water was made of. Water is nonetheless H₂O, and following arguments that Saul Kripke gave in lectures of 1970, many philosophers hold that it could not have been anything else. If conceiving water without H₂O does not show that they are two things, conceiving a mind without a brain does not show that either.

Defenders of the argument have a reply. Water has a hidden nature, which chemistry found by looking beneath its appearance. A pain, they say, has no appearance to look beneath: the way it feels is what it is. So the gap between what can be conceived and what is possible, which is real for water, may not exist for experience. Whether that reply works is still argued over, and the Stanford Encyclopedia's entry on dualism sets out how far each side has pressed it.

A few paragraphs later Descartes gives a second, independent argument.

I here remark, in the first place, that there is a vast difference between mind and body, in respect that body, from its nature, is always divisible, and that mind is entirely indivisible. For in truth, when I consider the mind, that is, when I consider myself in so far only as I am a thinking thing, I can distinguish in myself no parts, but I very clearly discern that I am somewhat absolutely one and entire; and although the whole mind seems to be united to the whole body, yet, when a foot, an arm, or any other part is cut off, I am conscious that nothing has been taken from my mind ...

(1) Every body can be divided into parts.
(2) My mind cannot be divided into parts.
(3) If two things differ in any property, they are not one and the same thing.
(4) So my mind is not a body, and in particular it is not my brain.

Premise (3) is secure: if x and y were one thing, nothing could be true of x that is false of y. Premise (2) is where the evidence of the last two centuries presses. Gage lost much of a lobe and, if the reports are right, part of his character with it. In split-brain patients, whose corpus callosum, the band of fibres joining the two halves of the brain, has been cut to control severe epilepsy, each hemisphere can be shown a different picture. Asked what they saw, such patients describe only the picture sent to the hemisphere that controls speech, while the left hand, run by the other hemisphere, picks out the object from the other picture. One person seems, for a moment, to know and not know the same thing.

A dualist can answer that surgery divides the brain's supply of information to the mind, not the mind itself. But premise (2) was supposed to be something Descartes could see simply by attending to himself, and these cases suggest that attending to yourself is not a reliable witness on the question.

The letter Descartes could not answer

In 1643 Elisabeth of Bohemia, a twenty-four-year-old princess living with her exiled family in The Hague, began a correspondence with Descartes by asking the question his theory has never shaken off. If the mind is a thinking thing with no extension, how does it move the body? In the mechanical science of the day, one thing moved another by pushing it, and pushing needs a surface. Descartes' answers did not satisfy her, and the letters that followed, which ran until his death in 1650, are still the classic statement of the interaction problem.

The objection has a modern form. Every physical event that neuroscience has traced, including the firing of the neurons that move your arm, has turned out to have physical causes, and nothing has shown up that needs a non-physical push. A dualist replies that this assumes what it needs to prove, since nobody has traced every cause of a single decision. A physicalist answers that the dualist is betting on a gap in the causal story that research keeps narrowing. Notice that this part of the dispute is about evidence and about which explanation is simpler, which is one reason it has lasted.

So what?: Dualism's strength is how the mind seems from the inside. Its weakness is explaining how two utterly different kinds of thing act on each other.

Three answers, side by side

Physicalism is the view that everything that exists is physical, or is fixed by what is physical. Two physicalist theories of mind have mattered most. The identity theory, set out by U. T. Place in 1956 and J. J. C. Smart in 1959, says that each kind of mental state simply is a kind of brain process, in the way that lightning simply is an electrical discharge. Functionalism says that a mental state is defined by its job. Pain is whatever state is caused by damage to the body, causes distress and the wish to be rid of it, and produces wincing and avoidance, whatever that state happens to be made of.

Read across the third row first. All three theories can accommodate Gage. Physicalists argue that the dualist's instrument story adds nothing, since every effect of the injury is explained by the injury. Dualists reply that their theory was never meant to explain brain damage. It is meant to explain experience, and on that question they think it is the physicalist columns that are in trouble.

Now read the bottom row. The identity theory's problem was pressed by Hilary Putnam in the 1960s. If pain is one particular kind of human brain process, then an octopus, say, whose nervous system is built very differently, could not be in pain, and neither could a Martian or a machine. That seemed wrong. The point that one mental state can be realised in many physical systems is called multiple realisability, and functionalism was built to accommodate it. A role can be played by many different actors, and on this view pain is a role.

Functionalism's own weakness sits in the same row, and it is the subject of the next section: could a system play every role that pain plays and still feel nothing at all?

The upshot: Each theory handles the cases that motivated it and struggles with one problem that the others handle better. That is why the argument has run for nearly four centuries.

What it is like: the problem of experience

In 1714 Gottfried Wilhelm Leibniz put the problem in an image. He was not defending Descartes, and he was certainly not a physicalist; he was arguing that perception cannot be mechanical. Robert Latta's 1898 translation of the Monadology gives it like this.

Moreover, it must be confessed that perception and that which depends upon it are inexplicable on mechanical grounds, that is to say, by means of figures and motions. And supposing there were a machine, so constructed as to think, feel, and have perception, it might be conceived as increased in size, while keeping the same proportions, so that one might go into it as into a mill. That being so, we should, on examining its interior, find only parts which work one upon another, and never anything by which to explain a perception.

Walk through the mill, or scan a living brain, and you find parts acting on parts. You do not find the redness of red. In 1974 Thomas Nagel gave the modern version in a paper called "What Is It Like to Be a Bat?". A bat finds moths by echolocation, a sense no human has. Suppose we knew every physical fact about the bat's brain and its sonar. Nagel argued that we would still not know what echolocating is like for the bat, because that is a fact tied to a point of view, and physical science describes the world from no point of view in particular. On his account, a creature has conscious experience exactly when there is something it is like to be that creature.

In 1982 Frank Jackson turned the thought into an argument with premises, now called the knowledge argument. Mary is a brilliant scientist who has lived all her life in a black-and-white room, and has learned, from black-and-white books and screens, every physical fact there is about colour and colour vision. One day she walks out and sees a ripe tomato.

(1) Before her release, Mary knows every physical fact about seeing red.
(2) On her release, she learns something new about seeing red: what it is like.
(3) So some facts about seeing red are not physical facts.
(4) So physicalism is false.

The argument is valid, so a physicalist has to deny a premise. There are two main ways of doing it.

The ability reply. Mary learns no new fact. She gains new abilities: to recognise red on sight, to imagine it, to remember it. Knowing how is not knowing that, so "learns something new" in premise (2) shifts its meaning between the premises, the fallacy of equivocation from Lesson 2. Dualists answer that what Mary gains is plainly knowledge of what something is like, and that imagining red is already a kind of experience of red, so the reply only moves the problem somewhere else.

The new-concept reply. Mary learns an old fact in a new way. She already knew the fact about her visual system; now she can think about it through a concept that only having the experience can give her, much as you might know a fact about Venus under the name "the evening star" without knowing it under the name "the morning star". Dualists answer that the new way of thinking must itself involve something the physical story leaves out.

The argument's author changed sides. By 2003 Jackson had given up the property dualism the argument supports and accepted physicalism, arguing that it rests on a mistaken picture of what an experience is. The argument did not go with him. Property dualists still defend it, holding that experiences have properties that are not physical even though there is no separate thinking substance.

Why this matters: Physicalism has an account of what a mind does. The disputed question is whether any physical account captures what having a mind is like.

Could a machine think? Turing's test

In 1950 the mathematician Alan Turing published a paper called "Computing Machinery and Intelligence", which opens: "I propose to consider the question, 'Can machines think?'" He then declined to answer it directly. The word "think" was too vague to test, so he replaced the question with a game. An interrogator types questions to two hidden players, one human and one machine, and has to say which is which. If interrogators cannot reliably tell, Turing argued, we have the same reason to credit the machine with thought that we have for crediting other people with it, since with other people too behaviour is all we ever have to go on. This is the Turing test.

Turing made a prediction: that in about fifty years an average interrogator would have no more than a 70 percent chance of making the right identification after five minutes of questioning. For decades that looked optimistic. Then in March 2025 Cameron Jones and Benjamin Bergen posted a report of two randomised, pre-registered three-party tests, in which each interrogator talked for five minutes with a person and a machine at the same time. GPT-4.5, told to adopt a humanlike persona, was judged to be the human 73 percent of the time, more often than the real people it was paired with. ELIZA, a program from the 1960s, was chosen 23 percent of the time.

Does that mean GPT-4.5 thinks? Go back to the table. The functionalist column says a pass is serious evidence, provided the states producing the conversation play the right roles. The identity theorist says it depends what the machine is made of and how it works. The dualist says no behaviour is ever enough on its own. And one argument from 1980 was aimed squarely at Turing's idea.

John Searle asked you to imagine him locked in a room with a rule book, written in English, for manipulating Chinese characters he cannot read. Questions in Chinese come in under the door; he follows the rules and passes back strings of characters, and the answers are good enough that the people outside assume a Chinese speaker is inside. Searle understands no Chinese at all. He is doing what a computer does, manipulating symbols by their shapes, so, he argued, running a program can never be enough for understanding. This is the Chinese Room.

(1) If running the right program were sufficient for understanding, Searle in the room would understand Chinese.
(2) Searle in the room does not understand Chinese.
(3) So running the right program is not sufficient for understanding.

That is modus tollens, and it is valid. The best-known reply, the systems reply, attacks premise (1). Of course the man does not understand, since he is only one part of a larger system, just as a single neuron in your head does not understand English. The whole system, man plus rule book plus room, is what understands. Searle answered by imagining that he memorises the rule book and works in the open air: now the whole system is inside him, and he still understands nothing. Defenders of the systems reply say that is what you would expect if a second system were running on his brain, one he has no access to. The exchange has not been settled, and it is no longer only a seminar question now that people talk to machines every day.

Remember: The Turing test measures whether a machine's conversation can be told from a person's. Whether that is evidence of thinking is the question, not the answer.

Common misconceptions

• "Neuroscience has proved that the mind is the brain." It has shown that mental life depends closely on the brain, which Descartes' dualism already allows. The dispute is over whether dependence is identity, and a brain scan cannot show that.
• "Dualism is a religious position." Descartes' arguments for it make no appeal to scripture, and the property dualists who defend the knowledge argument today argue from consciousness alone. Many dualists make no religious claim at all.
• "If you can imagine something, it is possible." Someone ignorant of chemistry can imagine water that is not H₂O. What you can imagine depends on what you know, which is the main objection to Descartes' first argument.
• "Physicalists think experiences are not real." Most physicalists hold that experiences are entirely real and are physical processes. The view that some mental states do not exist at all, eliminativism, is a separate and far more radical position.
• "Passing the Turing test proves a machine is conscious." Turing designed the test to set aside words like "conscious". Passing shows that the conversation cannot be told apart from a person's; whether that shows thought, let alone feeling, is exactly what Searle disputes.

Summing up

• Gage's injury shows that the mind depends closely on the brain. All three theories accept that, so the case alone decides nothing.
• Descartes' conceivability argument is valid; critics attack the step from what can be conceived to what is possible, using cases like water and H₂O.
• His divisibility argument depends on the mind having no parts, a premise that split-brain cases put under pressure.
• The interaction problem, first pressed by Elisabeth of Bohemia in 1643, asks how something with no extension could move a body.
• The identity theory says mental states are brain processes; multiple realisability pushed many physicalists toward functionalism, which defines mental states by their roles.
• Leibniz's mill, Nagel's bat and Jackson's Mary press one point: a physical description seems to leave out what experience is like. Physicalists answer with the ability and new-concept replies.
• The Turing test replaces "can machines think?" with a test of conversation. A 2025 study found a chatbot judged human 73 percent of the time; Searle's Chinese Room argues that no such result could show understanding, and the systems reply disputes that.

Sources

1. Descartes, R. (1901). Meditations on first philosophy (J. Veitch, Trans.; first published 1853), Meditation VI. Wikisource. wikisource.org
2. Leibniz, G. W. (1898). The monadology (R. Latta, Trans.), section 17. Wikisource. wikisource.org
3. Robinson, H., & Weir, R. (2025). Dualism. In E. N. Zalta & U. Nodelman (Eds.), Stanford Encyclopedia of Philosophy. plato.stanford.edu
4. Oppy, G., & Dowe, D. (2021). The Turing test. In E. N. Zalta & U. Nodelman (Eds.), Stanford Encyclopedia of Philosophy. plato.stanford.edu
5. Jones, C. R., & Bergen, B. K. (2025). Large language models pass the Turing test (arXiv:2503.23674). arXiv preprint. arxiv.org

A dot on a clock face, 1983

In a study published in 1983, the American neuroscientist Benjamin Libet and three colleagues sat volunteers in front of a screen on which a single dot swept round a clock face. Electrodes on the scalp recorded brain activity; electrodes on the forearm recorded the movement of a muscle. The instruction was simple. Flex your wrist whenever you feel like it, and note where the dot was at the moment you first became aware of the urge to move.

Averaged over many trials, the results came out like this. The volunteers reported first feeling the urge about 200 milliseconds before the wrist moved. But a slow build-up of electrical activity over the motor areas of the brain, called the readiness potential, had begun about 500 milliseconds before the movement. The brain seemed to be preparing the act some 300 milliseconds before the person knew they had decided to act.

The headline many people drew was that your brain decides before you do, and that the feeling of choosing is a report on a decision already made. Libet himself did not draw it. In later runs he found that people could still veto a movement after becoming aware of the urge, and he regarded his results as compatible with free will. Critics asked whether flicking a wrist for no reason is a fair model of choosing a university or deciding whether to tell the truth. And in 2012 Aaron Schurger, Jacobo Sitt and Stanislas Dehaene argued in the Proceedings of the National Academy of Sciences that the readiness potential may be the build-up of random background fluctuations in neural activity, which sooner or later crosses a threshold, rather than the signature of a decision.

So the experiment settles less than its headline. But it makes vivid a question that needs no laboratory. Suppose every decision you make, including the next one, is the result of earlier causes: brain states, which came from genes, upbringing and circumstance, which came from events before you were born. Are you free? Can you deserve praise or blame for anything? Philosophers have split three ways over this, and the split has lasted three centuries.

The intelligence that knows your next move

In 1814 the mathematician Pierre-Simon Laplace put the idea of a fully fixed world in its most famous form. This is Frederick Truscott and Frederick Emory's translation of 1902.

We ought then to regard the present state of the universe as the effect of its anterior state and as the cause of the one which is to follow. Given for one instant an intelligence which could comprehend all the forces by which nature is animated and the respective situation of the beings who compose it, an intelligence sufficiently vast to submit these data to analysis, it would embrace in the same formula the movements of the greatest bodies of the universe and those of the lightest atom; for it, nothing would be uncertain and the future, as the past, would be present to its eyes.

The translators' dashes appear here as commas. Later writers called this imagined intelligence Laplace's demon, though Laplace never used the word. What it expresses is determinism: the thesis that the state of the world at any one time, together with the laws of nature, fixes everything that happens afterwards. If it is true, an intelligence given the state of the universe in 1814 could in principle have calculated what you will eat for breakfast tomorrow.

Is determinism true? That is a question for physics, and physics has not settled it. Quantum mechanics is standardly read as saying that some events, such as the moment a particular atom decays, are not fixed by anything before them, although deterministic interpretations of the theory also exist. You will see shortly that for the question of free will this matters less than you might expect.

Notice what determinism is not. It is not fatalism, the idea that what will happen will happen whatever you do, as in the old story of the servant who flees to Samarra to escape Death and finds Death waiting for him there. Determinism says the future happens because of what you do. Your deliberation is one of the links in the chain, not a spectator watching it.

The point: Determinism is a claim about causes, not about outcomes arriving whatever you do. The worry it raises is not that your choices are idle, but that they are themselves the products of things you never chose.

The argument that determinism leaves no room for you

Here is the central argument against free will, in a simplified form of what Peter van Inwagen in 1983 called the Consequence Argument.

(1) If determinism is true, every choice I make is fixed by the laws of nature together with the state of the world before I was born.
(2) I have no control over the laws of nature.
(3) I have no control over the state of the world before I was born.
(4) If something is fixed by things over which I have no control, I have no control over it either.
(5) So if determinism is true, I have no control over any choice I make.

It is valid, and premises (2) and (3) look undeniable. That puts the weight on premise (4), a principle about how a lack of control is passed along a chain of causes, and on what "control" means. Anyone who holds that free will and determinism cannot both be true is an incompatibilist, and incompatibilists then split over whether determinism is true. That gives three positions.

The word "libertarian" here has nothing to do with politics. It names the view that free choices are real and are not fixed by prior causes. A modern cousin of hard determinism, hard incompatibilism, adds that free will of the kind blame requires would be ruled out even if determinism were false, for reasons you will meet below under the heading of luck.

The compatibilist case, in Hume's words

David Hume thought the whole dispute rested on a confusion about a word. Section VIII of his Enquiry Concerning Human Understanding (1748) begins by arguing that everyone already believes in necessity in human affairs, because everyone predicts people.

It is universally acknowledged that there is a great uniformity among the actions of men, in all nations and ages, and that human nature remains still the same, in its principles and operations. The same motives always produce the same actions. The same events follow from the same causes.

His illustration is grim and exact.

A prisoner who has neither money nor interest, discovers the impossibility of his escape, as well when he considers the obstinacy of the gaoler, as the walls and bars with which he is surrounded; and, in all attempts for his freedom, chooses rather to work upon the stone and iron of the one, than upon the inflexible nature of the other.

The prisoner counts on the gaoler's character as surely as on the iron of the bars, and he is right to. Human choices, Hume is saying, are regular enough to predict, and we rely on that every day. Then he defines liberty in a way that sits comfortably with it.

By liberty, then, we can only mean a power of acting or not acting, according to the determinations of the will; that is, if we choose to remain at rest, we may; if we choose to move, we also may. Now this hypothetical liberty is universally allowed to belong to every one who is not a prisoner and in chains. Here, then, is no subject of dispute.

Freedom, on this view, is the opposite of constraint, not the opposite of causation. You are free when you do what you choose and nothing stops you; you are unfree when you are chained, locked in or forced. Hume's boldest move comes next. Responsibility does not merely survive causation; it needs it.

Actions are, by their very nature, temporary and perishing; and where they proceed not from some cause in the character and disposition of the person who performed them, they can neither redound to his honour, if good; nor infamy, if evil.

In numbered lines:

(1) To be free is to act according to your own will, without constraint.
(2) Acting according to your will is compatible with your will having causes.
(3) We hold people responsible for actions that come from their lasting character.
(4) An action with no cause would come from nothing in the person.
(5) So responsibility requires that actions be caused, and freedom in the sense of (1) is compatible with determinism.

Hume called this his "reconciling project". The general view, that freedom and determinism can both hold, is compatibilism.

In short: For Hume the opposite of free is forced, not caused. An uncaused action would not be more yours; it would not be yours at all.

Two modern reinforcements

The obvious reply to Hume is that he has changed the subject. The freedom people care about, incompatibilists say, is the ability to have done otherwise, with the past exactly as it was, and a determined person never has that. In 1969 Harry Frankfurt attacked the assumption behind the reply with what is now called a Frankfurt case. Jones is deciding whether to do something wrong. Black wants Jones to do it and has arranged things so that, if Jones shows any sign of deciding the other way, Black can intervene and make him do it. Jones decides to do it for his own reasons, and Black never lifts a finger. Jones could not have done otherwise. Yet he seems exactly as responsible as he would have been if Black had never existed. If that is right, responsibility does not require the ability to do otherwise, and the principle that it does, the principle of alternative possibilities, is false.

In a lecture of 1962 called "Freedom and Resentment", P. F. Strawson argued in a different way. Resentment when someone wrongs you, gratitude when someone helps you, indignation on behalf of a stranger: these reactive attitudes, he said, are part of the fabric of human relationships. We suspend them for particular reasons, as when we learn that someone was pushed, or is a small child, or was not in their right mind. But suspension works case by case. A general theoretical thesis like determinism gives no reason of that kind, and our commitment to the attitudes runs too deep to be given up on the strength of one.

What matters here: Compatibilists hold that what responsibility needs is that your action comes from you, from your reasons and your character, with nobody forcing it. Whether the chain of causes began before you were born is, for them, beside the point.

The case against compatibilism

Incompatibilists answer in two ways. The first is the Consequence Argument itself. If your character, your reasons and your decision were all fixed before you existed, then saying that the action "comes from you" relabels the chain without giving you any control over it.

The second is the manipulation argument, whose best-known version is due to Derk Pereboom. Imagine a team of neuroscientists who, before a man called Plum was born, engineered his brain so that at a certain moment he would decide to kill someone, by way of his own reasoning and with nobody constraining him at the time. Plum meets every one of Hume's conditions: he acts on his own will, from his own character, unforced. Most people judge that he is not responsible, because the scientists set him up. Pereboom then asks what difference it makes if the setting-up is done by the laws of nature and the state of the world long ago rather than by a team of scientists. If there is no relevant difference, determinism is manipulation without a manipulator.

Compatibilists reply that there is a difference: manipulation bypasses the ordinary ways a person comes to have and revise their values, and ordinary causation runs through them. Whether that difference can be stated without assuming the conclusion is where much of the current argument sits.

The libertarian case, and the objection from luck

Libertarians accept the incompatibilist arguments and deny that determinism holds for our free choices. Their evidence begins with experience. When you deliberate you take the options to be genuinely open, and it is hard to see how you could deliberate about something you believed was already settled. The Scottish philosopher Thomas Reid, writing in 1788, argued that in a free act the agent himself is the cause, not some event inside him; the view is now called agent causation. Robert Kane defends a different version. In what he calls self-forming actions, when you are torn between keeping a promise and walking away from it, indeterminism in the brain leaves the outcome open, and whichever way it goes, you made it go that way, because you were really trying to do both.

(1) We are sometimes morally responsible for what we do.
(2) Moral responsibility requires that we could have done otherwise, with the past as it was.
(3) If determinism were true of our choices, we could never have done otherwise with the past as it was.
(4) So determinism is not true of our responsible choices.

Notice that this argument and the hard determinist's share premises (2) and (3). They part company over (1): the libertarian keeps it and gives up determinism, and the hard determinist keeps determinism and gives up (1). That is why the table has three rows and not two.

The strongest objection to the libertarian is the argument from luck. Suppose a choice really is undetermined. Run the world back to the moment before it and play it again. On some replays you keep the promise; on others you walk away, with everything about you, your reasons, your character and your effort, exactly the same each time. What made the difference, then? Nothing about you, since you were the same in every replay. The difference looks like chance, and chance is not control. This is why the indeterminism of quantum physics does not by itself help: a random event in your brain gives you an outcome that nobody controlled, not one that you did. Libertarians reply that the agent, or the agent's effort, is exactly what settles it, and that the question why you chose A rather than B is answered by your reasons for A, even if those reasons did not guarantee it.

Bottom line: Compatibilism must answer the charge that it offers freedom without control. Libertarianism must answer the charge that indeterminism offers chance rather than control. Hard determinism accepts both charges and asks whether we can live without deserved blame.

What would settle it, and what would not

Laboratory results like Libet's test a narrower claim: whether conscious intentions come early enough to cause simple movements. Even a perfect answer would leave the main question standing, because a determinist and a libertarian can agree on every millisecond of the timing. Physics could in principle settle whether determinism is true, which would bear on the dispute between hard determinists and libertarians. It could not settle whether compatibilism is right, because that is a question about what freedom and control require, and it is argued the way Frankfurt and Pereboom argue, with cases and principles. That is why a question that looks as if it is about the brain has turned, for three centuries, into a question about what it means to say that someone could have done otherwise.

Common misconceptions

• "Determinism means your choices make no difference." That is fatalism. Determinism says outcomes happen because of your choices, and your deliberation is one of the causes.
• "Quantum randomness rescues free will." A random event in the brain would be an outcome nobody controlled. Indeterminism removes one obstacle for the libertarian and opens the objection from luck.
• "Libet proved that free will is an illusion." The timing measurements are disputed, a 2012 model reads the readiness potential as accumulated noise, the task involved no reasons at all, and Libet himself thought his veto results left room for free will.
• "Compatibilists deny that everything is caused." They allow that everything may be; they deny that being caused makes an action unfree.
• "If nobody has free will, nobody should be locked up." Free will sceptics such as Derk Pereboom and Gregg Caruso argue that dangerous people can still be kept from harming others, on the model of quarantine for a contagious disease. What changes is the justification: protection rather than deserved suffering.

Recap

• Libet's experiment found brain activity building before people reported the urge to move; its interpretation is disputed, and it tests a narrower question than free will.
• Determinism says the past and the laws fix every later event. It differs from fatalism, because on determinism your choices are among the causes.
• The Consequence Argument is valid; its weight rests on the principle that no control over the sources means no control over the result.
• Hume defines liberty as a power of acting according to the will, compatible with causation, and argues that responsibility needs actions to come from lasting character.
• Frankfurt cases challenge the principle of alternative possibilities; Strawson argues that the reactive attitudes cannot be overturned by a general thesis.
• The manipulation argument asks how determinism differs from being engineered; compatibilists answer that ordinary causation runs through a person's own capacities.
• Libertarians appeal to agent causation or to self-forming actions; the objection from luck says an undetermined choice is settled by nothing about the agent.

Sources

1. Hume, D. (1902). An enquiry concerning human understanding (L. A. Selby-Bigge, Ed.; 2nd ed.; first published 1748), Section VIII. Project Gutenberg eBook No. 9662. gutenberg.org
2. Laplace, P. S. (1902). A philosophical essay on probabilities (F. W. Truscott & F. L. Emory, Trans.; first published 1814). Project Gutenberg eBook No. 58881. gutenberg.org
3. O'Connor, T., & Franklin, C. (2022). Free will. In E. N. Zalta & U. Nodelman (Eds.), Stanford Encyclopedia of Philosophy. plato.stanford.edu
4. McKenna, M., & Coates, D. J. (2024). Compatibilism. In E. N. Zalta & U. Nodelman (Eds.), Stanford Encyclopedia of Philosophy. plato.stanford.edu
5. Libet, B., Gleason, C. A., Wright, E. W., & Pearl, D. K. (1983). Time of conscious intention to act in relation to onset of cerebral activity (readiness-potential): The unconscious initiation of a freely voluntary act. Brain, 106(3), 623-642.

A philosopher in a glass case, and his first paragraph

Inside the entrance to the Student Centre at University College London there is a glass case, and in it sits the skeleton of Jeremy Bentham, dressed in his own clothes, with a wax head fitted with some of his own hair. Bentham died in 1832 and left instructions that his body be dissected and then preserved as what he called an auto-icon. The real head, mummified with results that looked macabre, sat in the same case for years until repeated student pranks got it locked away.

Bentham's theory of morality is stated in the first paragraph of his Introduction to the Principles of Morals and Legislation, printed in 1780 and published in 1789.

Nature has placed mankind under the governance of two sovereign masters, pain and pleasure. It is for them alone to point out what we ought to do, as well as to determine what we shall do. On the one hand the standard of right and wrong, on the other the chain of causes and effects, are fastened to their throne. They govern us in all we do, in all we say, in all we think: every effort we can make to throw off our subjection, will serve but to demonstrate and confirm it.

Two claims are packed in there. One is psychological: pleasure and pain determine what we do. The other is moral: they point out what we ought to do. The second is the one that matters here, and a few lines later Bentham states it as a principle.

By utility is meant that property in any object, whereby it tends to produce benefit, advantage, pleasure, good, or happiness, (all this in the present case comes to the same thing) or (what comes again to the same thing) to prevent the happening of mischief, pain, evil, or unhappiness to the party whose interest is considered: if that party be the community in general, then the happiness of the community: if a particular individual, then the happiness of that individual.

This is utilitarianism in its founding form. An action is right to the degree that it tends to increase happiness, meaning pleasure and the absence of pain, for everyone it affects. Bentham adds that this applies "not only of every action of a private individual, but of every measure of government". He was a law reformer first, and he wanted a test a legislator could actually use.

Worth holding on to: Utilitarianism judges an action by one thing only, its consequences for the happiness of everyone affected, with each person's happiness counting equally.

Bentham's procedure, in his own words

A test a legislator can use needs a method, and Chapter IV of the Introduction supplies one, now called the felicific calculus. A pleasure or pain, Bentham says, is greater or less according to its intensity, its duration, its certainty or uncertainty, and its propinquity or remoteness, which means how soon it arrives. When we judge an act rather than a single feeling, two more circumstances count.

Its fecundity, or the chance it has of being followed by sensations of the same kind: that is, pleasures, if it be a pleasure: pains, if it be a pain. Its purity, or the chance it has of not being followed by sensations of the opposite kind: that is, pains, if it be a pleasure: pleasures, if it be a pain.

When more than one person is involved, a seventh is added.

Its extent; that is, the number of persons to whom it extends; or (in other words) who are affected by it.

He then tells you to begin with one person, total the pleasures and the pains the act produces for them, repeat for everyone affected, and sum. Bentham was not naive about this. He wrote that "it is not to be expected that this process should be strictly pursued previously to every moral judgment", only that it should "be always kept in view". So treat what follows the way he meant it: not as a machine that prints verdicts, but as a way of making sure you have asked every question that matters.

Running the calculus: the school's last six thousand pounds

A school has six thousand pounds left in this year's budget and two proposals for it. Option A pays for a trained counsellor one day a week for the rest of the year. Option B replaces the failing projectors in twelve classrooms. The numbers below are estimates made up for practice, not measurements, and inventing them is part of the exercise, because the calculus forces you to say what you actually believe about each quantity.

Step 1: list everyone affected. For A: the 36 students who would be seen, and roughly 72 family members and close friends who feel the difference. For B: the 900 students taught in those rooms, and 40 teachers.

Step 2: estimate intensity and duration for each group. Intensity on a scale of 0 to 10, duration in months. A student in real difficulty who gets help feels it strongly: say 8, lasting 12 months, since help this term carries into next year. A slightly sharper slide is a mild pleasure, say 0.5, but the projectors last 36 months.

Step 3: multiply by certainty. Counselling does not help everyone who comes: say 0.6. New projectors almost certainly work: 0.9.

Step 4: propinquity. Both options start next week, so the timing is equal and drops out of the comparison.

Step 5: fecundity and purity. A student who is helped tends to attend more and do better, which adds 6 to each student's total. Teachers lose some time learning the new system, an impurity worth minus 4.8 each.

Step 6: multiply by extent and add up.

On these numbers the projectors win by five to one. Before you object, notice what the procedure has already done. It has made you name every group affected, say how strongly and for how long, and admit how uncertain you are. Two people who disagree about this budget can now see exactly which number they disagree about.

The core of it: The calculus turns a vague sense that one option is better into a list of estimates, each of which can be challenged separately.

The same calculation with one input changed

Now suppose the local council announces that every classroom projector in the area will be replaced free of charge next September. The new projectors would then matter for 6 months, not 36. Change that one number and run the rows again.

Option A is unchanged at 3,326.4, and it now wins. One input, the duration of the projectors' benefit, flipped the verdict. That is a strength and a weakness at once. It is a strength because it tells the head teacher exactly which fact to find out before deciding. It is a weakness because every other number in the table was guessed too, and a different guess about the intensity of a clearer slide, or the certainty of counselling, could flip it back.

The first result also exposes the objection critics press hardest, the problem of aggregation. In the first run, a tiny pleasure spread across 900 people outweighed serious help for 36 students in real difficulty. Push the numbers further and the calculus will say that a trivial benefit to enough people outweighs anything at all done to a few. Many critics, and not only critics of Bentham, think some harms to a person cannot be cancelled by small gains to others, however many. Utilitarians reply that refusing to add up is itself a choice with consequences, and usually a worse one for the many.

Key idea: A changed input can flip the verdict, which shows you where to look for evidence. The deeper dispute is whether benefits to different people can simply be added together at all.

Mill's corrections

John Stuart Mill was raised on Bentham's ideas, since his father was Bentham's close ally, and in his short book Utilitarianism (1863) he restated the theory and repaired it where he thought it was weakest. His statement of the principle is the one most often quoted.

The creed which accepts as the foundation of morals, Utility, or the Greatest Happiness Principle, holds that actions are right in proportion as they tend to promote happiness, wrong as they tend to produce the reverse of happiness. By happiness is intended pleasure, and the absence of pain; by unhappiness, pain, and the privation of pleasure.

He insists that the happiness in question is everyone's, counted evenly.

... the happiness which forms the utilitarian standard of what is right in conduct, is not the agent's own happiness, but that of all concerned. As between his own happiness and that of others, utilitarianism requires him to be as strictly impartial as a disinterested and benevolent spectator.

First correction: quality as well as quantity. Critics had called utilitarianism a doctrine worthy only of swine, since pigs enjoy pleasure too. Mill answered that pleasures differ in kind, and gave a test.

Of two pleasures, if there be one to which all or almost all who have experience of both give a decided preference, irrespective of any feeling of moral obligation to prefer it, that is the more desirable pleasure.

And then the most famous lines in the book.

It is better to be a human being dissatisfied than a pig satisfied; better to be Socrates dissatisfied than a fool satisfied. And if the fool, or the pig, is of a different opinion, it is because they only know their own side of the question. The other party to the comparison knows both sides.

This changes the calculus. Bentham's scale had one dimension for strength; Mill says a pleasure of understanding or friendship can outrank any amount of a lower pleasure. The objection is that Mill has let in a standard other than pleasure, since if one pleasure is better than a more intense one, something besides pleasure is doing the ranking. Defenders reply that the ranking is still fixed by what people who have experienced both actually prefer, which keeps it inside the theory.

Second correction: you do not calculate every time. The obvious practical objection is that nobody has time to run the calculus before every action. Mill's reply uses an image from navigation.

Nobody argues that the art of navigation is not founded on astronomy, because sailors cannot wait to calculate the Nautical Almanack. Being rational creatures, they go to sea with it ready calculated; and all rational creatures go out upon the sea of life with their minds made up on the common questions of right and wrong ...

Rules like "keep your promises" and "do not steal" are, on this view, the recorded results of humanity's long experience of what tends to produce happiness. You follow them in ordinary life and fall back on the principle when they conflict. That thought separates two versions of the theory. Act utilitarianism says an act is right if it produces at least as much happiness as any alternative. Rule utilitarianism says an act is right if it follows a rule whose general acceptance would produce the most happiness. Scholars still argue about which one Mill held.

The trolley, the footbridge and the surgeon

In 1967 Philippa Foot described the driver of a runaway tram who can steer from a track where five workers will be killed onto a track where one will be. Judith Jarvis Thomson, who named the trolley problem in 1976, added variations. In the switch case, you are a bystander next to a lever that will divert the trolley from the five onto the one. In the footbridge case, you stand on a bridge beside a very heavy man, and the only way to stop the trolley is to push him onto the track, killing him and saving the five. In the transplant case, a surgeon has five patients who will die without organs and one healthy visitor whose organs would save them all.

For the act utilitarian, all three cases have the same shape: one death against five. Yet when people are asked, a majority approve of pulling the switch and disapprove of pushing the man, and almost nobody approves of the surgeon. A 2020 study by Edmond Awad and colleagues put three such dilemmas to 70,000 people in 42 countries and found that the order in which people ranked the sacrifices, from most to least acceptable, was the same in every country, even though how acceptable each one was varied a great deal from country to country. A 2009 survey of professional philosophers found 68 percent would pull the switch.

Utilitarians have three responses, and they are different theories in practice.

Accept the verdict. Some act utilitarians say the intuition against pushing is a feeling, not an argument, and that five deaths are worse than one however they come about.

Count the side effects. Others say the act utilitarian has not finished counting. A world in which surgeons may kill healthy visitors is a world in which nobody visits hospitals, and the fear and lost trust outweigh the five lives. Critics answer that you can stipulate the case to be secret and one-off, and then the calculation says operate.

Move to rules. The rule utilitarian asks which rule about killing would produce the most happiness if generally accepted, and answers that a rule forbidding doctors to kill patients for their organs beats any rule permitting it. The surgeon is wrong because the rule is right. The standard objection is that the view is unstable. If breaking the rule would do more good in this case, a rule utilitarian either breaks it, and collapses into act utilitarianism, or keeps it at a cost in happiness, which looks like valuing the rule for its own sake.

The upshot: Every version of utilitarianism owes an answer to the transplant case. Act utilitarians must accept the verdict or find costs in it; rule utilitarians must explain why keeping a rule matters when breaking it would do more good.

Common misconceptions

• "Utilitarianism says the ends justify the means." It says an action is right because of all its consequences for everyone affected. Means have consequences too, and on the rule version certain means are ruled out by the rules that do most good.
• "The greatest happiness of the greatest number means majority rule." Extent is one of seven dimensions. Intense suffering for a few can outweigh mild pleasure for many, as the second run of the school example showed.
• "Utility means usefulness." In Bentham's sense it means the tendency to produce happiness or prevent unhappiness, not practical convenience.
• "Mill thought higher pleasures are more intense." He said they are higher in quality, as judged by those who have experienced both, even when they come with more discontent.
• "A utilitarian must push the man off the footbridge." Only on one version. Rule utilitarians, and act utilitarians who count the wider effects, can reach the other verdict.

What to remember

• Bentham's principle of utility judges every action, and every law, by its tendency to increase or diminish the happiness of those affected.
• His calculus weighs intensity, duration, certainty, propinquity, fecundity, purity and extent, and he did not expect it to be run in full before every judgement.
• Running it on a real choice makes every estimate explicit, and changing a single input, here the projectors' duration, can flip the verdict.
• The aggregation objection says small benefits to many should not outweigh serious harm or help for a few.
• Mill ranks pleasures by quality, using the preferences of those who know both, and treats moral rules as a ready-calculated almanac.
• Act utilitarianism judges acts by their consequences; rule utilitarianism judges them by the rules whose general acceptance does most good.
• The switch, footbridge and transplant cases have the same numbers and draw different verdicts from most people, which every version of the theory has to explain.

Sources

1. Bentham, J. (1789). An introduction to the principles of morals and legislation, Chapters I and IV. Wikisource. wikisource.org
2. Mill, J. S. (2004). Utilitarianism, Chapter II (Project Gutenberg eBook No. 11224; first published 1863). Project Gutenberg. gutenberg.org
3. Driver, J. (2025). The history of utilitarianism. In E. N. Zalta & U. Nodelman (Eds.), Stanford Encyclopedia of Philosophy. plato.stanford.edu
4. Nathanson, S. (n.d.). Act and rule utilitarianism. Internet Encyclopedia of Philosophy. iep.utm.edu
5. Awad, E., Dsouza, S., Shariff, A., Rahwan, I., & Bonnefon, J.-F. (2020). Universals and variations in moral decisions made in 42 countries by 70,000 participants. Proceedings of the National Academy of Sciences, 117(5), 2332-2337. pubmed.ncbi.nlm.nih.gov

A knock at the door, 1797

In 1797 the Swiss-born French political writer Benjamin Constant published an objection to a German philosopher who was then seventy-three. A murderer comes to your door and asks whether your friend, whom he intends to kill, is hiding in your house. Your friend is. If the duty to tell the truth really has no exceptions, Constant argued, you must tell the murderer where your friend is, and a principle with that consequence cannot be right.

The philosopher was Immanuel Kant, and within months he replied in an essay translated as "On a Supposed Right to Tell Lies from Benevolent Motives". He accepted the inference. Lying to the murderer, Kant held, would still be wrong. Both men agreed that you could refuse to answer; the argument was about what you may do if refusing is not an option.

Most people who hear Kant's answer think it is monstrous. Yet it comes from a theory that many philosophers regard as the deepest account of morality ever given, and some of Kant's most careful defenders think he misapplied his own principle. To see which, you need the principle, in his words.

The only thing good without qualification

Kant's Groundwork of the Metaphysics of Morals (1785) opens with a claim that is the exact opposite of Bentham's starting point. This is the translation by Thomas Kingsmill Abbott, published under the title Fundamental Principles of the Metaphysic of Morals.

Nothing can possibly be conceived in the world, or even out of it, which can be called good, without qualification, except a good will. Intelligence, wit, judgement, and the other talents of the mind, however they may be named, or courage, resolution, perseverance, as qualities of temperament, are undoubtedly good and desirable in many respects; but these gifts of nature may also become extremely bad and mischievous if the will which is to make use of them, and which, therefore, constitutes what is called character, is not good.

Think of the courage of a skilled bank robber, or the intelligence of a fraudster. Those qualities make the crime worse, not better. Even happiness, Kant says a few lines later, can make people proud and presumptuous without a good will to correct it. The good will is good whatever it achieves, even if bad luck stops it achieving anything at all.

What makes a will good? Kant's answer is that it acts from duty, and he draws the line with a shopkeeper.

For example, it is always a matter of duty that a dealer should not over charge an inexperienced purchaser; and wherever there is much commerce the prudent tradesman does not overcharge, but keeps a fixed price for everyone, so that a child buys of him as well as any other. Men are thus honestly served; but this is not enough to make us believe that the tradesman has so acted from duty and from principles of honesty: his own advantage required it ...

The shopkeeper does the right thing, so his action is in accordance with duty. But he does it because honesty is good for business, so he does not act from duty, and his action, Kant says, has no moral worth. Moral worth belongs to the person who is honest because honesty is right, and who would stay honest if it stopped paying.

Key idea: For Kant, what makes an action right is not what it brings about but the principle it is done on. A good result done for a bad reason has no moral worth.

From duty to a test

Every action, Kant thinks, is done on a maxim, the principle the agent actually acts on: "when I am short of money, I will borrow and promise to repay", say. Most commands we follow are hypothetical imperatives: if you want to pass the exam, revise. They bind you only if you want the end. A moral command binds you whatever you want, so it must be a categorical imperative, and Kant argues there is only one.

There is therefore but one categorical imperative, namely, this: Act only on that maxim whereby thou canst at the same time will that it should become a universal law.

The test is not "what would happen if everyone did this?", as if Kant were counting consequences after all. It is whether your maxim could even exist as a law for everyone, and whether you could consistently want it to, without contradiction. Kant later gives a second formula, which he takes to be the same law seen from another side.

So act as to treat humanity, whether in thine own person or in that of any other, in every case as an end withal, never as means only.

Notice the word "only". You treat the bus driver as a means every morning, and there is nothing wrong with that, because the driver has agreed to the arrangement and is pursuing ends of their own through it. What the formula forbids is using a person merely as a means, in a way they could not possibly agree to.

Running the test on Kant's own case

Kant works the test on an example. Here it is, followed by the procedure it illustrates.

Another finds himself forced by necessity to borrow money. He knows that he will not be able to repay it, but sees also that nothing will be lent to him unless he promises stoutly to repay it in a definite time. He desires to make this promise, but he has still so much conscience as to ask himself: "Is it not unlawful and inconsistent with duty to get out of a difficulty in this way?" Suppose however that he resolves to do so: then the maxim of his action would be expressed thus: "When I think myself in want of money, I will borrow money and promise to repay it, although I know that I never can do so." ... For supposing it to be a universal law that everyone when he thinks himself in a difficulty should be able to promise whatever he pleases, with the purpose of not keeping his promise, the promise itself would become impossible, as well as the end that one might have in view in it, since no one would consider that anything was promised to him, but would ridicule all such statements as vain pretences.

Step 1: state the maxim. When I need money, I will borrow it by promising to repay, knowing I cannot.

Step 2: universalise it. Imagine a world in which everyone in need makes false promises to borrow, as a law of nature.

Step 3: look for a contradiction in conception. In that world, nobody would treat a promise to repay as meaning anything, so nobody would lend on the strength of one. The maxim needs lenders to believe promises; its universal version guarantees they do not. The maxim cannot even be conceived as a universal law. This marks a perfect duty: one that allows no exceptions.

Step 4: look for a contradiction in the will. Some maxims can exist as universal laws but cannot consistently be wanted as such. Kant's example is a man who resolves never to help anyone in need. A world like that could exist, but nobody could will it, since everyone will sometimes need help. That marks an imperfect duty, such as helping others, which leaves room for judgement about when and how.

Step 5: check with the second formula. Kant does this himself.

He who is thinking of making a lying promise to others will see at once that he would be using another man merely as a mean, without the latter containing at the same time the end in himself. For he whom I propose by such a promise to use for my own purposes cannot possibly assent to my mode of acting towards him ...

The lender cannot agree to being deceived, because agreeing would mean knowing, and knowing would end the deception. So the false promise uses him merely as a means. Both formulas give the same verdict.

Remember: The universal law test asks about contradiction, not consequences. The lying promise fails because a world of lying promises contains no promises to exploit.

The dispute: may you lie to the murderer at the door?

Position A, Kant's own, in its strongest form. A lie, whatever its motive, is a maxim of saying what you believe false so that another person will believe it. Universalise lying whenever you judge it beneficial, and statements lose their power to inform, so the maxim defeats itself exactly as the false promise did. The humanity formula points the same way. Deceiving someone manages their beliefs without their consent, treating their reason as an obstacle to get around rather than a mind to address. On this view the murderer is still a rational agent, and your duty not to deceive him is not something he can cancel by being wicked. You may refuse to answer, shut the door, shout a warning or fight; what you may not do is make yourself a liar. Defenders add that the rule without exceptions is what makes it trustworthy. A principle that says "tell the truth unless you judge a lie better" gives every liar a script.

Position B, in its strongest form. Kant misapplied his own test. The maxim of the person at the door is not "I will lie whenever it seems useful" but something like "I will lie to a person who is using deception to find someone he intends to murder". Universalise that, and nothing self-destructs. Ordinary communication goes on exactly as before, because the maxim applies only to would-be murderers who conceal their aims, and it works precisely because the murderer is pretending to be an ordinary inquirer. Christine Korsgaard made the best-known version of this argument in 1986. She also argued that the humanity formula is harder to satisfy, and that Kant's theory needs an account of how to act when others are doing evil, which his reply to Constant never gave. Defenders of Position B add a simpler point: the murderer has already tried to use you as a mere means, as an unwitting tool in a killing, and resisting that is not the same as doing it to him.

What would settle it. The whole dispute turns on one question the Groundwork does not answer: which description of an action is its maxim? Describe the act broadly, "lying", and it fails. Describe it narrowly, "lying to a concealed murderer", and it passes. If any detail may be written into a maxim, almost anything can pass: "I will break promises on Tuesdays to people called Morgan" universalises without trouble. If no detail may be, the test condemns things that seem plainly permitted. Kantians have offered accounts of which features belong in a maxim, usually the ones that actually move the agent to act. The dispute would be settled by an account of maxim description that both sides could accept, and none yet commands agreement.

The point: Whether Kant's ethics forbids the life-saving lie depends less on the categorical imperative itself than on how the maxim is described, and that is where the argument is still being fought.

Where the theory is attacked more generally

Conflicting duties. Perfect duties are supposed to have no exceptions. The murderer case looks like a clash between a duty not to lie and a duty to protect a friend, and Kant's scheme offers no way to rank them except to deny that one of them is a real duty in that situation.

Coldness. If moral worth belongs only to acts done from duty, then the friend who visits you in hospital because she loves you seems to earn less credit than one who visits because it is her duty. Kantians reply that Kant's claim is about what gives an action its moral worth, not about whether warmth is good, and that he thought people should cultivate sympathetic feeling.

Compare the surgeon. Set the theory beside the last lesson's transplant case. The act utilitarian has to argue about side effects to avoid approving the operation. The Kantian has a direct answer: killing the visitor for his organs uses him merely as a means, whatever the numbers. That is the strongest point in Kant's favour, and the murderer at the door is the strongest point against him.

Common misconceptions

• "The categorical imperative is just the golden rule." Kant rejects that reading in a footnote: the golden rule depends on what you happen to want, and on it "the criminal might argue against the judge who punishes him". The categorical imperative asks whether a maxim could be a law for everyone, whatever you want.
• "Universalising means asking whether everyone doing it would have bad results." The test is contradiction. The lying promise fails because it cannot exist as a universal practice, not because a world of liars would be unpleasant.
• "Kant says never treat people as means." He says never merely as means. Using a shopkeeper to buy bread is fine; deceiving him is not.
• "Kant thinks consequences never matter." They matter to prudence and to how you carry out imperfect duties such as helping others. What they cannot do is make a maxim right or give an action its moral worth.
• "Acting from duty means acting with no feeling." Kant's claim is about which motive gives an action moral worth. Doing your duty gladly does not cancel its worth, provided duty is what would move you if the gladness were absent.

Pulling it together

• Kant holds that only a good will is good without qualification, and that an action has moral worth only if it is done from duty, not merely in accordance with it.
• A maxim is the principle you act on. A categorical imperative binds whatever you want, unlike a hypothetical one.
• The formula of universal law asks whether your maxim could exist, and be willed, as a law for everyone without contradiction.
• The lying promise fails because universal false promising would leave no promises to exploit; the humanity formula condemns it because the lender cannot consent to being deceived.
• Kant told Constant that lying to the murderer at the door is wrong; Korsgaard and others argue that a properly described maxim of lying to a concealed murderer passes the test.
• The dispute turns on how maxims are to be described, which the Groundwork leaves unsettled.
• Kant's theory gives a direct answer to the transplant case and struggles with the murderer at the door.

Sources

1. Kant, I. (2004). Fundamental principles of the metaphysic of morals (T. K. Abbott, Trans.; Project Gutenberg eBook No. 5682; first published 1785). Project Gutenberg. gutenberg.org
2. Johnson, R., & Cureton, A. (2025). Kant's moral philosophy. In E. N. Zalta & U. Nodelman (Eds.), Stanford Encyclopedia of Philosophy. plato.stanford.edu
3. Wikipedia. (2026). Categorical imperative, section "Lying to a murderer". en.wikipedia.org
4. Korsgaard, C. M. (1986). The right to lie: Kant on dealing with evil. Philosophy & Public Affairs, 15(4), 325-349.

Seminary students in a hurry

In a study published in 1973, the psychologists John Darley and Daniel Batson asked students at a theological seminary in Princeton to walk to another building to give a short talk. Some were to speak on the parable of the Good Samaritan, the story of the traveller who stops to help a stranger left beaten by the road. Some were told they had time to spare; others that they were already late. On the way, each passed a man slumped in a doorway, coughing and groaning.

Of the students who were not in a hurry, 63 percent stopped to help. Of those who had been told they were late, 10 percent did. Whether a student was about to speak on the Good Samaritan made little difference; some stepped over the man on their way to talk about helping strangers.

The study raises the question at the centre of this lesson. The theories you have met so far ask what you should do: maximise happiness, or act on a maxim you could will as law. An older tradition asks what kind of person you should be, on the grounds that people act well when they have good characters, not when they consult good formulas. The seminary students seem to have had good characters and good formulas, and a few minutes of hurry beat both. Keep that in mind; it comes back at the end.

Aristotle's question: what is a human being for?

Aristotle's Nicomachean Ethics, from the fourth century BCE, begins by asking what the chief good of a human life is. Everyone calls it happiness, he says, the Greek eudaimonia, but that only names the target. Here is how he tries to say what it is, in D. P. Chase's translation.

But, it may be, to call Happiness the Chief Good is a mere truism, and what is wanted is some clearer account of its real nature. Now this object may be easily attained, when we have discovered what is the work of man; for as in the case of flute-player, statuary, or artisan of any kind, or, more generally, all who have any work or course of action, their Chief Good and Excellence is thought to reside in their work, so it would seem to be with man, if there is any work belonging to him.

A statuary is a sculptor. Aristotle goes on to strip away what humans share with other living things. Mere life and growth belong to plants; sensation belongs to horses and oxen. What remains is a life of the part of us that has reason. So the human good is that kind of life lived well, and he adds a condition in a famous image.

... for as it is not one swallow or one fine day that makes a spring, so it is not one day or a short time that makes a man blessed and happy.

In numbered lines, this is the function argument.

(1) The good of anything that has a characteristic work lies in doing that work well, as a harp-player's good lies in playing well.
(2) Human beings have a characteristic work: active life according to reason.
(3) So the human good is active life according to reason, done well, that is, in accordance with virtue.
(4) And it must be judged over a complete life, not a day.

The Greek word translated "virtue" or "excellence", arete, means the quality that makes something good at being what it is, as sharpness is the virtue of a knife. Notice that happiness, on this account, is not a feeling. It is an activity, and a whole life of it.

The argument is valid, and its weak point is premise (2). Why think human beings have a work at all, as flutes and eyes do? Evolution explains how our capacities arose without giving them a purpose. And even granting that reasoning is what is distinctive of us, distinctive is not the same as good: humans are also the only animals that run prisons. Defenders reply that Aristotle's point does not need a cosmic purpose, only the ordinary fact that we judge a human life by how well it uses the capacities that make it human, as we judge a wolf's life by how well it hunts.

The upshot: For Aristotle the good life is an activity, the exercise of reason done excellently over a whole life, not a feeling you have on a good afternoon.

How you become good, and what the mean is

Virtues, Aristotle says, are not born in us and not learned from a book. They are acquired the way skills are.

... men come to be builders, for instance, by building; harp-players, by playing on the harp: exactly so, by doing just actions we come to be just; by doing the actions of self-mastery we come to be perfected in self-mastery; and by doing brave actions brave.

This is habituation, and it has a consequence that is easy to miss. The person who has the virtue of courage does not merely manage to do the brave thing while terrified. They feel the right amount of fear, at the right things, and act well, because training has shaped their feelings as well as their behaviour. His definition of virtue follows.

Virtue then is "a state apt to exercise deliberate choice, being in the relative mean, determined by reason, and as the man of practical wisdom would determine."

This is the doctrine of the mean. Each virtue sits between two vices, one of excess and one of deficiency.

Two things stop the mean being a recipe for mediocrity. First, it is "relative", meaning relative to us and to the situation. Aristotle's own example is food: the right amount for a champion wrestler is not the right amount for a beginner. The right amount of anger at a friend who has betrayed you may be a great deal. Second, the mean is fixed by practical wisdom, phronesis: the trained ability to see what a situation calls for. There is no formula for it, which is Aristotle's point. Ethics, he says, is not an exact science, and the person of practical wisdom is the standard, not a rule.

In short: A virtue is a trained disposition to feel and act rightly, lying between two vices at the point practical wisdom finds, which may be far from the middle.

Care: a different voice

From the late 1950s the psychologist Lawrence Kohlberg scored children's moral development by how they answered dilemmas. The most famous asked whether a man called Heinz should steal a drug he cannot afford to save his dying wife. In In a Different Voice (1982), Carol Gilligan described two eleven-year-olds from those studies. Jake treated the dilemma like a maths problem about rights: the right to life outweighs the right to property, so Heinz should steal. Amy worried that Heinz would go to prison and leave his wife worse off, and thought that if he talked to the druggist they could find another way. Kohlberg's scale rated Amy's answer as less developed. Gilligan argued that it was a different moral perspective, one that sees people as a web of relationships to be maintained rather than as separate holders of rights, and that the scale was blind to it. She described the difference as one of theme, not of gender.

The philosopher Nel Noddings, in Caring (1984), made this into a theory. Morality begins not with abstract principles but with the relation of caring: a carer attends to the one cared for, receives their needs, and responds, and the one cared for in turn responds to the care. Later writers such as Virginia Held and Joan Tronto extended the ethics of care into politics, arguing that every society runs on unpaid and undervalued care for children, the sick and the old, and that a theory of justice that ignores dependency describes nobody who has ever lived.

Care ethics shares a good deal with Aristotle: both put character, perception and feeling at the centre, and both distrust formulas. But care ethics starts from dependency and from particular relationships, where Aristotle starts from the flourishing individual.

Four theories, side by side

Reading the table

Read the fourth row first. The two left-hand columns are impartial: your mother's suffering counts for no more than a stranger's. The two right-hand columns build partiality in. That is a strength if you think a person who would weigh their child's life exactly like a stranger's has something wrong with them, and a weakness if you think it licenses favouritism and neglect of people far away. Critics call care ethics parochial for this reason. Its defenders reply that care can be extended outward, through institutions, to people we will never meet.

Then the last row, for virtue ethics. The application objection says that "do what a virtuous person would do" is useless to someone who is not yet virtuous and wants to know what to do now. Virtue ethicists reply that the virtues themselves yield rules of a kind: be honest, do not be cruel, do not be cowardly. These are less precise than Kant's test or Bentham's calculus, and they think the precision of those theories is false.

Then the situationist challenge. Go back to the seminary. Philosophers such as John Doris and Gilbert Harman have argued that studies like Darley and Batson's, and another in which people who had just found a coin in a payphone were far more likely to help a stranger pick up dropped papers, 88 percent against 4, show that behaviour depends more on small features of a situation than on character. If there are no stable traits of the kind Aristotle describes, virtue ethics is about something that does not exist. Virtue ethicists answer in two ways. Aristotle never thought virtue was common, so a study showing most people lack it confirms him. And a virtue is supposed to be a complex disposition of perception, feeling and judgement, which a single choice in a corridor does not measure.

Why this matters: The theories differ not only in their verdicts but in their questions. Two ask what to do; two ask who to be and whom you are bound to, and each pair is strongest exactly where the other is weakest.

Common misconceptions

• "The mean means always being moderate." It is relative to the person and the situation, and fixed by practical wisdom. Sometimes a great deal of anger, or of generosity, is exactly right.
• "Aristotle's happiness is a good mood." Eudaimonia is an activity carried on over a complete life. One swallow does not make a spring.
• "Virtues are temperaments you are born with." Aristotle says we become just by doing just acts and brave by doing brave ones. A naturally fearless person without judgement is rash, not courageous.
• "Care ethics is an ethics for women." Gilligan described the care perspective as a theme, not a gender, and later care ethicists apply it to politics, medicine and international relations.
• "Virtue ethics gives no guidance at all." It gives rules of thumb drawn from the virtues and the example of wise people, but no algorithm, which it thinks no honest ethics can supply.

The takeaway

• Aristotle's function argument locates the human good in rational activity done well over a complete life; its weak point is the claim that humans have a characteristic work.
• Virtues are acquired by habituation and shape feelings as well as actions.
• Each virtue is a mean between two vices, relative to us and fixed by practical wisdom, not a formula for moderation.
• Gilligan argued that Kohlberg's scale missed a voice centred on care and relationships; Noddings built a theory on the caring relation.
• Utilitarianism and Kant are impartial; virtue and care ethics give special relationships a place in what we owe.
• Virtue ethics faces the application objection and the situationist challenge; care ethics faces the charge of parochialism and of entrenching unfair burdens on carers.

Sources

1. Aristotle. (1911). The Nicomachean ethics of Aristotle (D. P. Chase, Trans.; first published 1847), Books I and II. Wikisource. wikisource.org
2. Hursthouse, R., & Pettigrove, G. (2026). Virtue ethics. In E. N. Zalta & U. Nodelman (Eds.), Stanford Encyclopedia of Philosophy. plato.stanford.edu
3. Sander-Staudt, M. (n.d.). Care ethics. Internet Encyclopedia of Philosophy. iep.utm.edu
4. Doris, J., Stich, S., Phillips, J., & Walmsley, L. (2020). Moral psychology: Empirical approaches. In E. N. Zalta & U. Nodelman (Eds.), Stanford Encyclopedia of Philosophy. plato.stanford.edu
5. Darley, J. M., & Batson, C. D. (1973). "From Jerusalem to Jericho": A study of situational and dispositional variables in helping behavior. Journal of Personality and Social Psychology, 27(1), 100-108.

A lifeboat in the South Atlantic, July 1884

On 19 May 1884 the yacht Mignonette left Southampton for Sydney with a crew of four: the captain, Tom Dudley; Edwin Stephens; Edmund Brooks; and a seventeen-year-old cabin boy, Richard Parker, an inexperienced seaman. On 5 July, running before a gale some 1,600 miles northwest of the Cape of Good Hope, she sank, and the four men got away in a small open boat with almost no food and no fresh water.

By about 20 July Parker had fallen ill, probably from drinking seawater. On 23 or 24 July Dudley proposed that they draw lots to choose one man to die so that the others could eat. Brooks refused. That night Dudley raised it again with Stephens, pointing out that Parker was probably dying and that the two of them had wives and families. The next day Dudley said a prayer and killed Parker with a penknife while Stephens stood ready to hold his legs. Brooks later said he had signalled neither assent nor protest. All three fed on the body. On 29 July they sighted a sail, and a German barque took them aboard.

Back in England the survivors described what they had done openly, believing that necessity justified it. Dudley and Stephens were tried for murder. On 9 December 1884, in R v Dudley and Stephens, the High Court held that necessity is no defence to a charge of murder. They were sentenced to death with a recommendation of mercy, and the Home Secretary reduced the sentence to six months' imprisonment.

The court's answer, in the Lord Chief Justice's words

Lord Coleridge's judgment is a piece of moral argument as well as law. Here is the core of it.

To preserve one's life is generally speaking a duty, but it may be the plainest and the highest duty to sacrifice it. War is full of instances in which it is a man's duty not to live, but to die. The duty, in case of shipwreck, of a captain to his crew, of the crew to the passengers, of soldiers to women and children ... these duties impose on men the moral necessity, not of the preservation, but of the sacrifice of their lives for others ...

Who is to be the judge of this sort of necessity? By what measure is the comparative value of lives to be measured? Is it to be strength, or intellect, or what? It is plain that the principle leaves to him who is to profit by it to determine the necessity which will justify him in deliberately taking another's life to save his own. In this case the weakest, the youngest, the most unresisting, was chosen. Was it more necessary to kill him than one of the grown men? The answer must be "No".

The law's answer is not the same thing as philosophy's, and the rest of this lesson asks what the three theories you have studied say. The fastest way in is to start from an answer that sounds right and find where it goes wrong.

A plausible answer, with three bugs in it

Here is the kind of analysis a capable student writes in ten minutes. Read it and see whether you can spot the problems before they are traced below.

Utilitarianism clearly says Dudley and Stephens did the right thing: three lives were saved at the cost of one, and Parker was dying anyway. Kant says it was wrong, because killing is always wrong. Virtue ethics cannot help, because it only says "do what a virtuous person would do", and nobody knows what that is in a lifeboat. So the three theories give three different answers, and you just have to pick the one you like.

Every sentence in that paragraph contains a mistake, and the conclusion rests on all of them.

Bug one: "utilitarianism clearly says they were right"

The error is treating a head count as a calculation. Run Bentham's steps honestly and three questions appear that the student skipped.

Certainty. A utilitarian judges an act by its expected consequences given what could be known. On the day of the killing nobody knew that a ship would appear four days later. The case for killing depends on the estimate that all four would otherwise have died, and on how soon. The estimate that Parker was dying strengthens it, since the remaining life taken from him was short. But these are estimates, and the verdict moves with them.

Extent. The effects do not stop at the edge of the boat. A practice in which the strong may kill the weak whenever they judge it necessary would make everyone less safe, and Coleridge's question, "who is to be the judge?", is itself a point about consequences. Mill makes it the foundation of justice.

Justice is a name for certain classes of moral rules, which concern the essentials of human well-being more nearly, and are therefore of more absolute obligation, than any other rules for the guidance of life ...

The interest those rules protect, he says a few pages later, is "security, to every one's feelings the most vital of all interests". A rule utilitarian, and on many readings Mill, would ask which rule about killing in necessity does most good if generally accepted, and a rule letting those who profit decide is a poor candidate.

The choice of victim. Even an act utilitarian has to ask whether a fair lottery would have produced a better result than choosing the weakest, since fairness reduces fear and resentment among the survivors and among everyone who hears of it.

Debugged: an act utilitarian may approve the killing, if the estimates of certain death were sound and no better option existed. A rule utilitarian, weighing security, is likely to condemn it. "Clearly right" is the one verdict utilitarianism does not deliver.

Bug two: "Kant says killing is always wrong"

The verdict may be right, but the reason is wrong, and the wrong reason hides the most interesting thing Kant said. Kant did not hold that all killing is wrong: he allowed that you may act against a wrongful aggressor who attacks your life. Parker attacked no one. The Kantian objection is the formula of humanity from Lesson 10: Parker was killed so that others could eat him, which uses him merely as a means, in a way he could not possibly agree to. In some accounts his last words were "What me?"

Kant also wrote directly about cases like this one, in the part of the Metaphysics of Morals that William Hastie translated as The Philosophy of Law.

There can, in fact, be no Criminal Law assigning the penalty of death to a man who, when shipwrecked and struggling in extreme danger for his life, and in order to save it, may thrust another from a plank on which he had saved himself. For the punishment threatened by the Law could not possibly have greater power than the fear of the loss of life in the case in question. ... An act of violent self-preservation, then, ought not to be considered as altogether beyond condemnation (inculpabile); it is only to be adjudged as exempt from punishment (impunibile).

This is a distinction the student missed entirely: between an act being wrong and its being fit to punish. Kant's view is that the man on the plank does wrong, but that a law threatening him with a later death cannot outweigh certain drowning now, so punishment would be pointless. The English outcome, a conviction for murder followed by six months in prison, sits surprisingly close to it.

Debugged: on Kant's view killing Parker was wrong because it used him merely as a means, and the question of punishment is separate. Whether a fair lottery, freely agreed by all four, would change the verdict is disputed among Kantians: agreeing to a risk is not obviously agreeing to be used.

Bottom line: Kant's verdict turns on how Parker was treated, not on a rule against all killing, and Kant himself separated whether an act is wrong from whether the law can usefully punish it.

Bug three: "virtue ethics cannot help"

The student confused "gives no algorithm" with "gives no guidance". Aristotle discusses exactly this kind of situation in Book III of the Nicomachean Ethics, where he considers acts done under terrible pressure, like throwing cargo overboard in a storm. Such acts, he says, are mixed: voluntary when done, though nobody would choose them for their own sake. And then:

For some again no praise is given, but allowance is made; as where a man does what he should not by reason of such things as overstrain the powers of human nature, or pass the limits of human endurance. Some acts perhaps there are for which compulsion cannot be pleaded, but a man should rather suffer the worst and die; how absurd, for instance, are the pleas of compulsion with which Alcmaeon in Euripides' play excuses his matricide!

That gives the virtue ethicist real categories to work with. Was killing Parker a case for allowance, because nearly three weeks adrift in an open boat pass the limits of human endurance? Or one of the acts for which compulsion cannot be pleaded? The virtues sharpen the question. Justice asks why the weakest and youngest was chosen, the one boy who was in the captain's charge. Courage asks whether facing death was the right response to the fear of it. Practical wisdom asks what a fair procedure would have looked like, and it is worth noticing that the one man who refused, Brooks, is the one most readers find easiest to admire. A care ethicist would add that Parker was the person in the boat to whom the others owed the most care, which makes his selection worse, not better.

Debugged: virtue ethics does not calculate a verdict, but it asks which category the act falls into and what the choice of victim shows about the characters involved, and on the second question its answer is clear.

Where the three theories really part company

Fix the bugs and the student's conclusion, "three different answers, pick one", turns out to be false in an instructive way.

Read down the second column and something like agreement appears: every theory, once applied carefully, finds something seriously wrong with choosing the weakest, youngest and least able to resist. Read down the third and the agreement breaks up. The real disagreement is not about Dudley and Stephens at all. It is about whether consent to a fair procedure can make a killing permissible, and that is where you would have to argue if you wanted to settle the case.

What matters here: Rival theories often converge on a hard case once they are applied carefully, and the places where they still diverge show you which question is actually in dispute.

Common misconceptions

• "The theories always give different verdicts." Applied carefully they often agree, as they do about the choice of Parker. Their differences show up in variations, such as the lottery.
• "If the law convicts, the act was morally wrong, and if it does not, it was right." Kant separates wrongness from punishability, and Coleridge himself stressed that the court was not calling the men's deeds devilish.
• "Utilitarians just count lives." They count all the consequences they can foresee, including the effects of a practice on everyone's security.
• "Kant forbids all killing." He allowed action against a wrongful aggressor. His objection here is to using an innocent person merely as a means.
• "Virtue ethics has nothing to say about extreme cases." Aristotle distinguishes acts that deserve allowance under unbearable pressure from acts that no compulsion can excuse.

Where this leaves us

• In 1884 Dudley and Stephens killed the ailing seventeen-year-old Richard Parker to survive; the court held that necessity is no defence to murder, and the sentence was reduced to six months.
• Coleridge argued that the principle of necessity lets the person who profits decide whose life is worth less.
• A careful utilitarian analysis depends on certainty, wider effects and the choice of victim; act and rule versions can disagree.
• Kant's objection is that Parker was used merely as a means; his plank passage separates wrongdoing from what the law can usefully punish.
• Aristotle's mixed actions give categories for acts under compulsion: some deserve allowance, some should be refused even at the cost of death.
• The theories largely agree that choosing the weakest was wrong, and diverge over whether a freely agreed lottery would change the verdict.

Sources

1. Coleridge, J. D. (1884). Regina v Dudley and Stephens, 14 QBD 273 (judgment of the Lord Chief Justice). Wikisource. wikisource.org
2. Wikipedia. (2026). R v Dudley and Stephens. en.wikipedia.org
3. Kant, I. (1887). The philosophy of law (W. Hastie, Trans.; first published 1797), Introduction, "Right of Necessity". Wikisource. wikisource.org
4. Aristotle. (1911). The Nicomachean ethics of Aristotle (D. P. Chase, Trans.; first published 1847), Book III. Wikisource. wikisource.org
5. Mill, J. S. (2004). Utilitarianism, Chapter V (Project Gutenberg eBook No. 11224; first published 1863). Project Gutenberg. gutenberg.org

A giant made of people, 1651

The title page of Leviathan, published in London in April 1651, is one of the most famous images in political thought. Etched by the Parisian artist Abraham Bosse after long discussion with the author, it shows a crowned giant rising out of a landscape, a sword in one hand and a bishop's crosier in the other. Look closely at his torso and arms and you see that they are made of more than three hundred tiny human figures, all turned toward the giant's head. Above him runs a line from the Book of Job: there is no power on earth to be compared to him.

The author, Thomas Hobbes, wrote the book during the English Civil War, which ran from 1642 to 1651 and saw a king tried and executed. The picture is his argument in a single image. A state is made of its people, who have handed their power to one head, and that head's power is the only thing standing between them and something far worse. The question the image answers is one you have probably asked in a less grand form: why should I obey rules I never agreed to, made by people I did not choose?

This lesson sets three answers side by side: Hobbes's, John Locke's, and one you have already met, the argument the Laws of Athens put to Socrates in his cell.

Hobbes: the war of every man against every man

Hobbes begins by imagining people with no government at all, the state of nature. His first premise is that people are roughly equal. The spelling is the original of 1651.

Nature hath made men so equall, in the faculties of body, and mind; as that though there bee found one man sometimes manifestly stronger in body, or of quicker mind then another; yet when all is reckoned together, the difference between man, and man, is not so considerable, as that one man can thereupon claim to himselfe any benefit, to which another may not pretend, as well as he. For as to the strength of body, the weakest has strength enough to kill the strongest, either by secret machination, or by confederacy with others, that are in the same danger with himselfe.

Equal people who want the same scarce things will compete; people who cannot trust each other will strike first; and people who care about reputation will fight over insults. Hobbes draws the conclusion.

Hereby it is manifest, that during the time men live without a common Power to keep them all in awe, they are in that condition which is called Warre; and such a warre, as is of every man, against every man. ... In such condition, there is no place for Industry; because the fruit thereof is uncertain; and consequently no Culture of the Earth; no Navigation ... no Arts; no Letters; no Society; and which is worst of all, continuall feare, and danger of violent death; And the life of man, solitary, poore, nasty, brutish, and short.

Notice that "Warre" does not mean constant fighting. It means, as Hobbes says, a known disposition to fight, the way foul weather means a tendency to rain rather than a single shower. The way out is an agreement of every person with every other.

... as if every man should say to every man, "I Authorise and give up my Right of Governing my selfe, to this Man, or to this Assembly of men, on this condition, that thou give up thy Right to him, and Authorise all his Actions in like manner." This done, the Multitude so united in one Person, is called a COMMON-WEALTH ...

In numbered lines:

(1) People are roughly equal: the weakest can kill the strongest.
(2) Equal people who compete for scarce goods, distrust each other and seek reputation will come into conflict.
(3) So without a common power to keep them in awe, they are in a state of war of every one against every one.
(4) In that condition life is solitary, poor, nasty, brutish and short, which every rational person wants above all to escape.
(5) The only escape is for everyone to give up the right of governing themselves to a single sovereign with the power to keep the peace.
(6) So rational people would make that agreement, and are bound to obey the sovereign it creates.

The sovereign Hobbes arrives at is nearly unlimited, because any limit would need a judge, and a judge above the sovereign would itself be the sovereign. There is one thing, though, you cannot give away.

A Covenant not to defend my selfe from force, by force, is alwayes voyd. For (as I have shewed before) no man can transferre, or lay down his Right to save himselfe from Death, Wounds, and Imprisonment ...

Key idea: Hobbes justifies government by comparison with its absence. Almost any government, he argues, is better than the war of all against all, which is why he thinks you owe it almost unconditional obedience.

Locke: a law before any government

Locke's Second Treatise of Government, published in 1689, starts from the same place and reaches a very different conclusion, because it describes the state of nature differently.

The state of nature has a law of nature to govern it, which obliges every one: and reason, which is that law, teaches all mankind, who will but consult it, that being all equal and independent, no one ought to harm another in his life, health, liberty, or possessions ...

For Locke, people without government already have rights and already know, by reason, that they must respect each other's. The trouble with the state of nature is not constant war but inconvenience: when someone wrongs you, you have to be judge in your own case and enforce the verdict yourself. People form a government to fix that, and only by agreeing to it.

MEN being, as has been said, by nature, all free, equal, and independent, no one can be put out of this estate, and subjected to the political power of another, without his own consent.

The purpose of government, Locke says, is "the mutual preservation of their lives, liberties and estates, which I call by the general name, property". A government that turns against that purpose has broken the trust it was given.

... whenever the legislators endeavour to take away, and destroy the property of the people, or to reduce them to slavery under arbitrary power, they put themselves into a state of war with the people, who are thereupon absolved from any farther obedience ...

Hobbes's subjects can resist only to save their own lives. Locke's people can remove a government that abuses its trust. That difference is why Locke's language of consent and of life, liberty and property became the working vocabulary of the revolutions that followed.

So what?: Where you end up depends on where you start. A state of nature that is a war makes almost any sovereign worth obeying; a state of nature governed by reason makes government a trustee that can be dismissed.

Three answers, side by side

The third answer is older than both. In Plato's Crito, set in 399 BCE in the days after the trial you read in Lesson 3, Socrates' friend Crito has bribed the guards and begs him to escape. Socrates imagines the Laws of Athens questioning him, and they make an argument that sounds strikingly like a contract. This is Jowett's translation.

... we further proclaim to any Athenian by the liberty which we allow him, that if he does not like us when he has become of age and has seen the ways of the city, and made our acquaintance, he may go where he pleases and take his goods with him. ... But he who has experience of the manner in which we order justice and administer the state, and still remains, has entered into an implied contract that he will do as we command him.

Reading the table

The second row shows the family resemblance. All three ground obedience in something like an agreement. The Laws say that staying was agreeing; Hobbes says that rational people would agree; Locke says that people did agree, if only tacitly.

The bottom row shows the family weakness. Very few people have ever made an explicit agreement to obey their government. Locke saw the problem and answered it with tacit consent.

... every man, that hath any possessions, or enjoyment, of any part of the dominions of any government, doth thereby give his tacit consent, and is as far forth obliged to obedience to the laws of that government, during such enjoyment, as any one under it; whether this his possession be of land, to him and his heirs for ever, or a lodging only for a week; or whether it be barely travelling freely on the highway ...

If walking down a road is consenting, consent is so easy to give that it can no longer explain why you owe anything. David Hume pressed this in an essay called "Of the Original Contract". Saying that people consent by staying in the country of their birth, he wrote, is like saying that a man consents to the authority of a ship's captain "though he was carried on board while asleep and must leap into the ocean and perish the moment he leaves her". Hume thought the real ground of obedience was not consent at all but the plain usefulness of a system of laws that lets people live in peace.

What philosophers have tried instead. Some ground obligation in fair play: if you benefit from a cooperative scheme like a system of laws that others obey at some cost to themselves, it is unfair to take the benefits and dodge the burdens. Some ground it in gratitude, which is part of what the Laws say to Socrates when they compare themselves to his parents. Some, following John Rawls, speak of a natural duty to support just institutions whether or not you agreed to them. And some, the philosophical anarchists, conclude that there is no general obligation to obey the law as such, though there are usually excellent reasons to obey particular laws, such as the one against murder, which you would have reason to obey if it were not a law at all. Each view has serious defenders, and the question is open.

Worth holding on to: Consent is the most natural ground of an obligation to obey and the hardest to find in real life, which is why the argument has moved to fair play, gratitude and natural duty.

Common misconceptions

• "Hobbes thought people are evil." He wrote that "the Desires, and other Passions of man, are in themselves no Sin". The war of all against all comes from equality, scarcity, distrust and pride without a common power, not from wickedness.
• "The social contract is a story about a real historical meeting." For Hobbes it shows what rational people would agree to. Locke thought some governments did begin by agreement, but his argument about what makes a government legitimate does not depend on finding the date.
• "By property, Locke meant land and money." He used the word for "lives, liberties and estates" together. Critics have pointed out, though, that his theory of how land becomes property served the interests of English colonists in America against Native Americans who hunted on it.
• "Consenting to a government means agreeing with every law it makes." Consent theorists mean consent to the government's authority, which includes the authority to make laws you think are mistaken.
• "If there is no general obligation to obey, anything goes." Philosophical anarchists hold that you still have strong reasons to obey most laws; what they deny is that a law's being the law is itself a moral reason.

What to carry forward

• Hobbes argues from the equality, competition and distrust of people without government to a war of all against all, and from there to an almost unlimited sovereign.
• The one right Hobbes says cannot be given up is the right to defend your own life and body.
• Locke's state of nature has a law of reason; government exists by consent to protect life, liberty and estate, and may be resisted when it seeks arbitrary power.
• The Laws of the Crito argue that Socrates agreed to obey by staying in Athens when he was free to leave.
• Tacit consent, as Locke describes it, is so easily given that Hume compared it to consenting to a ship's captain after being carried aboard asleep.
• Alternative grounds of political obligation include fair play, gratitude and a natural duty of justice; philosophical anarchists deny any general obligation.

Sources

1. Hobbes, T. (2002). Leviathan, Chapters XIII, XIV and XVII (Project Gutenberg eBook No. 3207; first published 1651). Project Gutenberg. gutenberg.org
2. Locke, J. (2005). Second treatise of government, Chapters II, VIII and XIX (Project Gutenberg eBook No. 7370; first published 1689). Project Gutenberg. gutenberg.org
3. Plato. (1999). Crito (B. Jowett, Trans.; Project Gutenberg eBook No. 1657). Project Gutenberg. gutenberg.org
4. Dagger, R., & Lefkowitz, D. (2021). Political obligation. In E. N. Zalta & U. Nodelman (Eds.), Stanford Encyclopedia of Philosophy. plato.stanford.edu
5. Tuckness, A. (2020). Locke's political philosophy. In E. N. Zalta & U. Nodelman (Eds.), Stanford Encyclopedia of Philosophy. plato.stanford.edu

Two payslips, April 2026

The UK tax year that began on 6 April 2026 uses the same income tax bands for England as the years before it. You pay nothing on the first £12,570, the personal allowance. You pay 20 percent on income above that up to £50,270, 40 percent on income from £50,271 to £125,140, and 45 percent on anything above £125,140. Above £100,000 the personal allowance shrinks by £1 for every £2 earned, and at £125,140 it is gone. Scotland has its own bands. Now take two people in England.

A nurse earns £38,000. The first £12,570 is untaxed, and the remaining £25,430 is taxed at 20 percent: £5,086 in income tax, about 13.4 percent of her pay.

A consultant surgeon earns £140,000. He has no personal allowance, so every pound is taxable, band by band.

That is about 35.1 percent of his income. He earns under four times what the nurse earns and pays nearly ten times as much income tax. Is that just? Or is it unjust in the other direction, since he could argue that he pays for the same roads and schools many times over? Hold on to your first answer, because this lesson will test it three ways.

There is a third design, and Britain tried it. The Community Charge, known everywhere as the poll tax, replaced local property rates in Scotland from 1989 and in England and Wales from 1990. It charged a fixed amount per adult resident, with reductions for poor people, so a duke and a cleaner in the same council area paid the same basic sum. On 31 March 1990 a demonstration against it in London became a riot around Trafalgar Square, with 339 arrests and 113 people injured. The government announced its repeal in 1991, and Council Tax replaced it in 1993.

So there are three designs on the table: the same amount per head, the same rate for everyone, and rising rates. Each has been defended by serious people as the just one.

First answer: count the wellbeing

A utilitarian asks which system produces the most wellbeing, and has a strong argument for rising rates. It rests on diminishing marginal utility: each extra pound adds less wellbeing the more pounds you already have. A thousand pounds is rent to the nurse and a holiday upgrade to the surgeon. Take the same thousand from the surgeon rather than the nurse and total wellbeing is higher.

Followed all the way, that argument would take every pound above the average and give it to those below. The utilitarian stops short for a utilitarian reason: taxes change behaviour. Set the top rate too high and some people work less, move abroad or hide income, the total shrinks, and there is less for everyone. So the utilitarian's just tax is progressive, with rates set by evidence about where the losses from lower effort begin to outweigh the gains from redistribution.

The objection is one you met in Lesson 9. The utilitarian treats society's wellbeing as one big sum and does not care, in principle, how it is divided between people. John Rawls put it in a phrase: utilitarianism does not take seriously the distinction between persons.

Second answer: choose behind a veil

John Rawls's A Theory of Justice (1971) asks you to imagine choosing the basic rules of your society from what he called the original position. You are behind a veil of ignorance: you do not know whether you will be the nurse or the surgeon, rich or poor, talented or not, which sex you will be, or what you will think makes a life go well. You do know general facts about economics and psychology. Because nobody behind the veil can tailor the rules to their own advantage, Rawls argued, the principles chosen there are fair.

Rawls argued that you would choose two principles. The first guarantees everyone equal basic liberties: freedom of conscience, thought, expression and association, and equal political rights. The second governs inequalities. Positions must be open to all under fair equality of opportunity, so that people with similar talents and willingness to use them have similar chances whatever their background. And economic inequalities are allowed only if they work to the greatest benefit of the least advantaged. That last part is the difference principle.

Why choose it? Because behind the veil you might turn out to be the person at the bottom, and a rational chooser who cannot know the odds protects against the worst outcome. Notice what the difference principle allows. It does not demand equal incomes. If paying surgeons well draws able people into surgery, and that makes the worst-off patients better off than they would be under any alternative, the inequality is just. What it rules out is inequality that does nothing for those at the bottom.

Apply it to the three tax designs. The poll tax charged the cleaner the same basic sum as the duke; unless that somehow made the least advantaged better off than any alternative, it fails. A flat rate and a progressive system both have to be judged by what they do for the worst-off group, once incentives are taken into account, and that is partly an empirical question. The objection most often pressed against Rawls is that choosing to protect the worst outcome is extremely cautious: the economist John Harsanyi argued that rational choosers behind a veil would maximise the average instead, which brings the utilitarian back.

The point: The veil of ignorance turns "what is good for me?" into "what could I accept whoever I turn out to be?", and Rawls argued that the answer permits inequality only when it helps those at the bottom.

Third answer: what you are entitled to

Robert Nozick, Rawls's colleague at Harvard, replied in Anarchy, State, and Utopia (1974) that both answers ask the wrong question. They ask what pattern of distribution is best, then use the state to impose it. Nozick's entitlement theory asks instead how people came to have what they have. If you acquired something justly, or received it by a just transfer, a gift or a fair trade, from someone who held it justly, you are entitled to it, and a distribution made of such holdings is just whatever it looks like.

His most famous argument uses the basketball star Wilt Chamberlain. Start with any distribution you think just, say one chosen by the difference principle. Now suppose a million people each freely pay 25 cents to watch Chamberlain play. At the end of the season he has $250,000, far more than anyone else, and your favoured pattern is broken. Yet every step was voluntary, and each person spent what was theirs to spend. How can a just starting point plus free choices produce an injustice? Nozick concluded that preserving any pattern means constantly interfering with free exchange, and that taxing earnings to redistribute them amounts to a kind of forced labour, since part of your working time is spent, involuntarily, for others. He accepted a minimal state, funded to protect people against force, theft and fraud; what he rejected was taxation for redistribution.

The strongest objection concerns history. Nozick's theory makes present holdings just only if the chain of transfers behind them is just, and in the real world many chains run back to conquest, theft and fraud. Nozick included a principle of rectification for past injustice, but did not say how to apply it, and critics point out that correcting centuries of unjust acquisition might require redistribution on a scale no Rawlsian ever proposed. You saw one example in Lesson 13: Locke's theory of acquisition was used to treat Native American hunting grounds as available land.

Worth holding on to: Rawls and the utilitarians judge a distribution by its shape; Nozick judges it by its history. Taxes look very different depending on which you think matters.

The second problem: a school rule

The same three answers apply to things much closer to you. Suppose a sixth-form college has 20 fully funded places on a summer programme at a university, worth about £2,000 each, and 120 applicants. It has to decide how to allocate them.

Read the table by rows first and something appears: for this school good, the utilitarian and the Rawlsian often agree, because the students with the fewest chances are also the ones who would gain most. Read it by columns and the deeper difference appears: the Nozickian is not asking which pattern is best at all, but whether the procedure was fair and who has the right to decide.

What the two problems have in common

Every question of distribution, a tax, a school place, an organ for transplant, turns on some mix of five considerations: equality, need, desert or merit, entitlement through fair procedure, and priority to those worst off. The three theories weigh them differently. Utilitarians count need heavily because it predicts where resources do most good. Rawls makes priority to the worst off and fair opportunity central. Nozick makes entitlement central and denies that justice has a pattern at all. None of them treats desert, what people have earned, as the basic measure, which surprises most students; Rawls argued that even the talent and effort that earn rewards are shaped by luck of birth and upbringing.

Remember: Before arguing about whether a rule is fair, name which consideration you are relying on: equality, need, desert, entitlement or the worst off. Most disagreements about fairness are disagreements about which of these should win.

Common misconceptions

• "A higher tax band means paying the higher rate on your whole income." Rates apply to slices. The surgeon pays 20 percent on his first £37,700 like everyone else; only the pounds inside each band are taxed at that band's rate.
• "Rawls wanted everyone to have the same income." The difference principle permits inequalities that work to the greatest benefit of the least advantaged.
• "Nozick opposed all taxes." He accepted taxation to fund a minimal state that protects people against force, theft and fraud; he rejected taxation to redistribute.
• "A flat tax means everyone pays the same amount." A flat tax charges everyone the same rate, so the richer pay more pounds. A head tax like the poll tax charges the same amount.
• "The veil of ignorance asks what a selfish person would choose." The parties choose rationally for themselves, but because the veil hides who they are, their choice has to be acceptable from every position in society.

The short version

• In 2026-27, a nurse on £38,000 in England pays about 13.4 percent of her income in income tax and a surgeon on £140,000 about 35.1 percent, because rates rise band by band.
• The poll tax charged a fixed amount per adult and provoked the Trafalgar Square riot of 31 March 1990; it was replaced in 1993.
• Utilitarians favour progressive taxes because of diminishing marginal utility, limited by the effects of taxes on effort.
• Rawls's parties behind the veil choose equal basic liberties, fair equality of opportunity, and inequalities only where they benefit the least advantaged.
• Nozick judges holdings by their history of just acquisition and transfer; his Wilt Chamberlain case argues that keeping any pattern requires interfering with liberty.
• The same frameworks apply to a school rule, where the utilitarian and the Rawlsian often converge and the Nozickian asks a different question.

Sources

1. GOV.UK. (2026). Income Tax rates and Personal Allowances, current rates and allowances for the tax year 6 April 2026 to 5 April 2027. HM Revenue & Customs. gov.uk
2. Wikipedia. (2026). Community Charge. en.wikipedia.org
3. Freeman, S. (2023). Original position. In E. N. Zalta & U. Nodelman (Eds.), Stanford Encyclopedia of Philosophy. plato.stanford.edu
4. Lamont, J., & Favor, C. (2017). Distributive justice. In E. N. Zalta & U. Nodelman (Eds.), Stanford Encyclopedia of Philosophy. plato.stanford.edu
5. Feser, E. (n.d.). Nozick, Robert: Political philosophy. Internet Encyclopedia of Philosophy. iep.utm.edu

A watch on a heath, 1802

The most famous argument for God's existence in English begins on a walk. This is the opening of William Paley's Natural Theology, first published in 1802, in the twelfth edition of 1809.

In crossing a heath, suppose I pitched my foot against a stone, and were asked how the stone came to be there; I might possibly answer, that, for any thing I knew to the contrary, it had lain there for ever: nor would it perhaps be very easy to show the absurdity of this answer. But suppose I had found a watch upon the ground, and it should be inquired how the watch happened to be in that place; I should hardly think of the answer which I had before given, that, for any thing I knew, the watch might have always been there. Yet why should not this answer serve for the watch as well as for the stone? why is it not as admissible in the second case, as in the first? For this reason, and for no other, viz. that, when we come to inspect the watch, we perceive (what we could not discover in the stone) that its several parts are framed and put together for a purpose ...

A few pages later he states the conclusion he draws about the watch: "the inference, we think, is inevitable, that the watch must have had a maker". Then he turns to the eye, the hand, the joints of animals, and argues that they show the same marks of purpose, only more so. Charles Darwin, as a student at Cambridge, admired the book greatly.

This lesson takes three classic arguments for God's existence and runs the same procedure on each. It does not tell you whether God exists. It shows you how each argument works at its strongest, where each one is attacked, and what its defenders say back, so that whatever you conclude, you conclude it for reasons you can state.

The procedure

Six steps, the same for all three arguments.

1. Read the text, in the author's words.
2. Number the premises, adding any the author left unstated.
3. Check validity. If the premises were true, would the conclusion have to be?
4. Find the contested premise, the one the argument's critics actually deny.
5. State the strongest objection to that premise, and the strongest reply.
6. Change one input and see whether the result changes, which tests whether the form or a premise is at fault.

Argument one: design

Steps 1 and 2. Paley's argument, in numbered lines:

(1) A watch has parts framed and put together for a purpose.
(2) Whatever has parts framed and put together for a purpose was made by a designer.
(3) Living things, such as the eye, have parts framed and put together for a purpose, to an even greater degree than a watch.
(4) So living things were made by a designer.

Step 3. Valid, if premise (2) is read as a universal claim.

Step 4. The contested premise is (2): that purposive structure can only come from a designer.

Step 5. The strongest objections came from two directions. In Dialogues Concerning Natural Religion, published in 1779 after his death, David Hume gave the sceptical character Philo an argument that even granting design, nothing follows about what the designer is like.

This world, for aught he knows, is very faulty and imperfect, compared to a superior standard; and was only the first rude essay of some infant deity, who afterwards abandoned it, ashamed of his lame performance: it is the work only of some dependent, inferior deity; and is the object of derision to his superiors ...

The second objection came in 1859, when Darwin's On the Origin of Species described a process, natural selection acting on variation over immense time, that produces structures adapted to a purpose with no one intending them. That is a direct challenge to premise (2).

Step 6: change one input. Replace "a designer" in premise (2) with "a designer or a process of natural selection". The argument no longer delivers its conclusion for living things, because an alternative explanation is now available. That is why defenders of the design argument moved to ground where natural selection cannot operate: the laws and constants of physics themselves. Physicists have found that several fundamental quantities, such as the strength of gravity relative to other forces, seem to lie within narrow ranges outside which stars, stable atoms or chemistry as we know it could not exist. This is the fine-tuning argument, and it is the design argument's strongest modern form: there was nothing for natural selection to select from before there were laws of physics.

Critics reply in three ways. If there are many universes with different constants, a multiverse, some are bound to permit life, and we could only ever find ourselves in one that does. We have no measure of how probable different constants were, so "improbable" may mean nothing. And a designer complex enough to set the constants needs explaining too. Defenders answer that a multiverse is itself unobserved and far from simple, and that the argument claims only that design is the better explanation, not that it is proved.

Key idea: Darwin removed the design argument's force for biology by supplying an alternative explanation. The argument survives, contested, in physics, where the alternatives are a multiverse or brute chance.

Argument two: a first cause

Step 1. In the thirteenth century Thomas Aquinas gave five short arguments for God's existence, the Five Ways. Here are the second and third, in the translation of the Fathers of the English Dominican Province.

The second way is from the nature of the efficient cause. In the world of sense we find there is an order of efficient causes. There is no case known (neither is it, indeed, possible) in which a thing is found to be the efficient cause of itself; for so it would be prior to itself, which is impossible. Now in efficient causes it is not possible to go on to infinity ... Therefore it is necessary to admit a first efficient cause, to which everyone gives the name of God.

The third way is taken from possibility and necessity, and runs thus. We find in nature things that are possible to be and not to be, since they are found to be generated, and to corrupt, and consequently, they are possible to be and not to be. But it is impossible for these always to exist, for that which is possible not to be at some time is not. Therefore, if everything is possible not to be, then at one time there could have been nothing in existence. Now if this were true, even now there would be nothing in existence, because that which does not exist only begins to exist by something already existing. ... Therefore we cannot but postulate the existence of some being having of itself its own necessity, and not receiving it from another, but rather causing in others their necessity. This all men speak of as God.

Step 2. The Third Way, numbered:

(1) Some things are contingent: they come into being and pass away, so they might not have existed.
(2) Each contingent thing, at some time, does not exist.
(3) So if everything were contingent, at some time nothing would have existed.
(4) If at some time nothing existed, nothing would exist now, since things begin to exist only through something already existing.
(5) Something exists now.
(6) So not everything is contingent: there is a necessary being, which we call God.

Step 3. The move from (2) to (3) is invalid as it stands. That each thing fails to exist at some time does not mean there is one time at which everything fails to exist together, just as each person at a party leaving at some point does not mean there was a moment when the room was empty, if new guests kept arriving. Defenders repair this by adding premises, for example that past time is finite, or by moving to a stronger form.

Step 4, the strongest modern forms. There are two. The contingency argument, developed from Leibniz, rests on the principle of sufficient reason: every contingent fact has an explanation. The whole collection of contingent things is itself a contingent fact, so it too needs an explanation, which cannot be another contingent thing, since that would be part of what needed explaining. So there is a necessary being. The kalam argument, named after the tradition of medieval Islamic theology in which it was developed, is simpler: whatever begins to exist has a cause; the universe began to exist; so the universe has a cause. Its defenders support the second premise both philosophically, arguing that an actually infinite past is impossible, and with Big Bang cosmology.

Step 5. The oldest objection to the contingency argument is in Hume's Dialogues, and surprisingly it is spoken by Cleanthes, the character who defends design.

... the uniting of these parts into a whole, like the uniting of several distinct countries into one kingdom, or several distinct members into one body, is performed merely by an arbitrary act of the mind, and has no influence on the nature of things. Did I show you the particular causes of each individual in a collection of twenty particles of matter, I should think it very unreasonable, should you afterwards ask me, what was the cause of the whole twenty.

Bertrand Russell pressed the same point in the twentieth century as the fallacy of composition: because each part of the universe is contingent, it does not follow that the universe as a whole is. Defenders reply that some wholes clearly do need explaining beyond their parts, and that the principle of sufficient reason, which we rely on constantly, gives no exemption to the largest fact of all. Against the kalam, critics ask whether "cause" and "begin" apply to the beginning of time itself, where there is no earlier moment for a cause to occupy.

Step 6: change one input. Replace "God" with "the universe itself" as the necessary being. Hume's Cleanthes asks exactly this: why may not the material universe be the necessarily existent being? Nothing in the argument as stated rules it out. That shows the argument's real gap. Even if it proves a first cause or a necessary being, a further argument is needed to show that this being is personal, intelligent or good, which is what most people mean by God. Defenders accept this and offer further arguments for those attributes.

The upshot: The cosmological argument's strongest forms rest on the principle of sufficient reason or on the universe having a beginning, and even if sound they reach a first cause, not yet the God of any religion.

Argument three: the greatest conceivable being

Step 1. Around 1078 Anselm, then a monk at the abbey of Bec in Normandy and later Archbishop of Canterbury, wrote a prayer called the Proslogion that contains the most debated argument in the philosophy of religion, the ontological argument. This is Sidney Norton Deane's translation. "The fool" is the one in the psalm who says in his heart that there is no God.

... this very fool, when he hears of this being of which I speak, a being than which nothing greater can be conceived, understands what he hears, and what he understands is in his understanding; although he does not understand it to exist. ... And assuredly that, than which nothing greater can be conceived, cannot exist in the understanding alone. For, suppose it exists in the understanding alone: then it can be conceived to exist in reality; which is greater. Therefore, if that, than which nothing greater can be conceived, exists in the understanding alone, the very being, than which nothing greater can be conceived, is one, than which a greater can be conceived. But obviously this is impossible. Hence, there is no doubt that there exists a being, than which nothing greater can be conceived, and it exists both in the understanding and in reality.

Deane's dashes around the phrase in the first sentence appear here as commas. Step 2, numbered:

(1) God is a being than which nothing greater can be conceived.
(2) Even someone who denies that God exists understands this idea, so such a being exists at least in the understanding.
(3) Existing in reality as well as in the understanding is greater than existing in the understanding alone.
(4) Suppose that being existed in the understanding alone. Then a greater being could be conceived: the same being, existing in reality.
(5) But then the being than which nothing greater can be conceived would be one than which something greater can be conceived, which is a contradiction.
(6) So that being exists in reality.

Step 3. The argument is a reduction to absurdity, and it is harder than it looks to show that it is invalid. That is exactly what made the first objection so ingenious.

Step 6 first: change one input. Soon after Anselm wrote, a monk called Gaunilo of Marmoutiers wrote a reply "on behalf of the fool". He kept the form and changed the subject. Imagine an island, lost somewhere in the ocean, more excellent in its riches and delights than any other land. You understand the idea, so it exists in your understanding. An island that existed in reality would be more excellent than one that existed only in the mind. So, by the same reasoning, the most excellent island must exist. Since it plainly does not, something in the form of the argument must be wrong. This is a parody objection, the counterexample method of Lesson 1 applied to a whole argument.

Step 5: objections and replies. Anselm replied that his reasoning applies only to that than which nothing greater can be conceived, not to the greatest thing of some limited kind. Modern defenders sharpen the point: an island has no maximum, since you can always conceive one with more palm trees or better beaches, while a being with maximal power, knowledge and goodness does have one. The second great objection came from Kant in the eighteenth century, and it attacks premise (3). Existence, he argued, is not a property that adds to what a thing is. A hundred real coins contain not one coin more than a hundred imagined coins; the difference is whether the concept is instantiated, not what it contains. If existence is not a way of being great, premise (3) fails. Defenders reply with a modal version, most influentially Alvin Plantinga's in the 1970s: if it is even possible that a maximally great being exists, where maximal greatness includes existing necessarily, then such a being exists. Critics answer that the key premise, that such a being is possible, is exactly as doubtful as the conclusion, since the same logic run from "it is possible that no such being exists" proves that none does.

In short: The ontological argument tries to prove God's existence from the idea of God alone. Gaunilo's island tests its form, Kant's coins test its premise about existence, and the modern modal version moves the whole dispute to whether a maximally great being is possible.

Where each argument stands

Notice what the table shows. Every one of the arguments is valid in its strongest form, and every one of them turns on a single premise that intelligent people accept and intelligent people deny. That is the state of the question, and it is why the question is still asked.

Common misconceptions

• "If everything has a cause, what caused God?" The strong versions do not say everything has a cause. They say whatever begins to exist, or whatever is contingent, has one, and they claim God is neither. The real question is whether that exemption is justified or special pleading.
• "Refuting an argument for God shows that God does not exist." As Lesson 1 showed, an unsound argument leaves its conclusion open. The same holds for arguments against God.
• "Darwin refuted the design argument." He supplied an alternative to Paley's biological version. The fine-tuning version concerns physics, where natural selection does not apply.
• "The ontological argument is just a word trick." Philosophers have struggled for nine centuries to say exactly what is wrong with it, and the two classic diagnoses, Gaunilo's and Kant's, are themselves disputed.
• "If these arguments worked, they would prove a particular religion true." At most they reach a designer, a first cause or a greatest possible being. Getting from there to the God of any particular faith needs further argument.

Putting it together

• Paley argued from purposive structure, as in a watch or an eye, to a designer; Hume questioned what could be inferred about the designer, and Darwin supplied an alternative explanation for living things.
• The strongest modern design argument appeals to the fine-tuning of physical constants; critics reply with a multiverse or doubts about the probabilities.
• Aquinas's Third Way moves invalidly from each thing's not existing at some time to nothing existing at once; the contingency and kalam arguments are its stronger successors.
• The composition objection, from Hume's Cleanthes and Russell, and the gap between a first cause and God are the main problems for the cosmological argument.
• Anselm's argument tries to derive God's existence from the idea of a greatest conceivable being; Gaunilo's island parodies its form, and Kant denies that existence adds greatness.
• Each argument, in its strongest form, is valid and turns on one contested premise, which is where any honest disagreement has to be conducted.

Sources

1. Paley, W. (1809). Natural theology: Or, evidences of the existence and attributes of the Deity (12th ed.; first published 1802), Chapter I. Darwin Online. darwin-online.org.uk
2. Hume, D. (2003). Dialogues concerning natural religion, Parts V and IX (Project Gutenberg eBook No. 4583; first published 1779). Project Gutenberg. gutenberg.org
3. Aquinas, T. (2006). Summa theologica, Part I, Question 2, Article 3 (Fathers of the English Dominican Province, Trans.; Project Gutenberg eBook No. 17611). Project Gutenberg. gutenberg.org
4. Anselm. (1903). Proslogium (S. N. Deane, Trans.), Chapter II. Wikisource. wikisource.org
5. Reichenbach, B. (2026). Cosmological argument. In E. N. Zalta & U. Nodelman (Eds.), Stanford Encyclopedia of Philosophy. plato.stanford.edu

All Saints' Day, Lisbon, 1755

At about twenty to ten on the morning of Saturday 1 November 1755, the Feast of All Saints, an earthquake with its epicentre in the Atlantic, around 290 kilometres southwest of Lisbon, struck the city. Seismologists now estimate its magnitude at 7.7 or greater. Fires broke out across the city and a tsunami followed, and between them they almost completely destroyed Lisbon. Estimates of the dead run to tens of thousands.

The disaster shook more than buildings. Some decades earlier Leibniz had coined the word theodicy for the attempt to justify God in the face of the world's imperfections, and had argued that this is the best of all possible worlds. Voltaire answered with a poem on the Lisbon disaster and then with Candide (1759), whose hero is told, through one catastrophe after another, that all is for the best.

A city destroyed on a holy day by an earthquake that no human being caused is the sharpest form of the oldest objection to belief in God. This lesson sets out that objection at its strongest, then the strongest replies, and then an argument of a completely different kind, Pascal's wager, which asks what you should do if argument cannot settle the question at all. It reaches no verdict.

Epicurus's old questions

In Part X of Hume's Dialogues Concerning Natural Religion, the sceptic Philo puts the problem in four sentences, which he attributes to the Greek philosopher Epicurus.

EPICURUS's old questions are yet unanswered. Is he willing to prevent evil, but not able? then is he impotent. Is he able, but not willing? then is he malevolent. Is he both able and willing? whence then is evil?

Philosophers distinguish two kinds of evil here. Moral evil is the suffering caused by free choices: cruelty, war, murder. Natural evil is the suffering that comes from nature: earthquakes, disease, drought. Lisbon was natural evil on a vast scale. In 1955 J. L. Mackie turned Epicurus's questions into what is now called the logical problem of evil.

(1) If God exists, God is all-powerful, all-knowing and wholly good.
(2) A wholly good being eliminates evil as far as it can.
(3) An all-powerful, all-knowing being can eliminate all evil.
(4) Evil exists.
(5) So God does not exist.

The argument is valid, and the claim is strong: that God and evil cannot both exist, as a matter of logic. Everything depends on premise (2).

Position A: evil counts against God

The logical version, and what happened to it. In 1974 Alvin Plantinga answered Mackie with the free will defense. A world containing creatures who are free, and who freely do good, may be better than a world of creatures who are not free at all. But if creatures are genuinely free, whether they do wrong is up to them, not to God, so it is possible that God could not create a world containing free creatures who never go wrong. If that is even possible, premise (2) is false as stated: a wholly good being might permit evil for the sake of a greater good that could not exist without the risk of it. Plantinga claimed only that this is a possible reason, not God's actual one, and a possible reason is all it takes to show that there is no contradiction. Mackie himself later conceded that the defense shows the central doctrines of theism to be consistent after all, while doubting that it solves the problem. As the Internet Encyclopedia of Philosophy's survey puts it, many philosophers concluded that there must be more to the problem than the logical version captures.

The evidential version. That "more" was supplied in 1979 by William Rowe. Rowe asked you to picture a fawn, trapped in a forest fire started by lightning, badly burned, lying in agony for several days before it dies, with no human being ever knowing. His argument does not claim a contradiction. It claims that evil like this is evidence.

(1) There exist instances of intense suffering which an all-powerful, all-knowing being could have prevented without losing some greater good or permitting some evil equally bad or worse.
(2) A wholly good, all-knowing being would prevent any intense suffering it could, unless it could not do so without losing some greater good or permitting some evil equally bad or worse.
(3) So there does not exist an all-powerful, all-knowing, wholly good being.

Premise (2) is one most theists accept. The weight is on premise (1), and Rowe's support for it is inductive: we can think of no greater good that the fawn's days of agony serve, and after long and careful looking, the best explanation of our finding none is that there is none. Lisbon adds numbers to the fawn's intensity: tens of thousands dead, children among them, in a disaster no free choice caused. The free will defense does not obviously reach it.

Why this matters: The debate moved from "God and evil are inconsistent", which is widely thought to have been answered, to "the evil we actually see makes God improbable", which is where it is fought today.

Position B: evil does not count decisively against God

Theists have three main replies to the evidential argument, and they work in different ways.

Soul-making. Drawing on the second-century bishop Irenaeus, John Hick argued in Evil and the God of Love (1966) that the world is not meant to be a comfortable home but a "vale of soul-making". Courage needs danger; compassion needs suffering to answer; responsibility needs a world of stable natural laws in which actions have predictable effects, including the law that fire burns and the laws that move the plates of the earth. A world rearranged by constant miracles to spare everyone harm would be a world where nobody could become good by choosing to be. On this view Lisbon is the price of a world with reliable laws, and the response to it, rescue, rebuilding and care, is the kind of moral growth that such a world makes possible.

The best objection is the fawn. No soul is made by suffering that no one sees, and the amount and distribution of suffering, falling on infants as well as adults, seems far more than soul-making could require.

Skeptical theism. The second reply attacks Rowe's inference. From "we can see no good this suffering serves" Rowe concludes "there probably is none". But that inference is only as strong as our ability to see such goods, and the gap between human understanding and an all-knowing mind is, by hypothesis, enormous. A toddler held down for a vaccination can see no reason for the pain either. Skeptical theists hold that our failure to find a reason is therefore weak evidence that none exists. The objection is that this proves too much: if we cannot judge whether suffering serves a hidden good, we seem unable to judge whether we should prevent suffering ourselves, since for all we know it serves a good we cannot see.

Turning the argument around. Rowe himself described a third move, which he called the G. E. Moore shift. His argument runs from premise (1) to the conclusion that God does not exist. A theist who has strong independent reasons to believe God exists, perhaps from the arguments of Lesson 15, can run it backwards: God exists; premise (2) is true; so premise (1) is false, and the fawn's suffering serves some good after all, even if we cannot see it. Which direction is more reasonable depends on how strong each side's other evidence is.

What would settle it. For the logical version, the dispute is largely settled at the level of consistency, since the free will defense shows how God and evil could coexist. For the evidential version, three things would move it: an account of goods that plainly require suffering like the fawn's, which would support the theist; a principled reason to trust or distrust our judgements about hidden goods, which would decide the skeptical theist's reply; and the balance of all the other evidence about God, since, as the Moore shift shows, the argument from evil is one consideration to be weighed against others, not a verdict on its own.

The core of it: Both sides agree that a good God would not permit pointless suffering. They disagree about whether any suffering is shown to be pointless, and that turns on how far human judgement can see.

Pascal's bet: when the evidence runs out

Blaise Pascal, the French mathematician who helped found probability theory, left notes for a defence of Christianity when he died in 1662, published as the Pensées. In one of them he grants something the arguments above never grant: that reason cannot settle whether God exists. He then asks what you should do anyway. This is W. F. Trotter's translation.

Yes; but you must wager. It is not optional. You are embarked. Which will you choose then? Let us see. Since you must choose, let us see which interests you least. You have two things to lose, the true and the good; and two things to stake, your reason and your will, your knowledge and your happiness; and your nature has two things to shun, error and misery. ... Let us weigh the gain and the loss in wagering that God is. Let us estimate these two chances. If you gain, you gain all; if you lose, you lose nothing. Wager, then, without hesitation that He is.

This is Pascal's wager, and it is not an argument that God exists. It is an argument about what it is rational to do when you do not know, and it is one of the first uses of what is now called decision theory. Set it out as a table.

In numbered lines: (1) You cannot avoid choosing, since not wagering for God is in effect wagering against. (2) Reason cannot settle which way to bet. (3) As long as the chance that God exists is greater than zero, an infinite reward multiplied by that chance outweighs any finite cost. (4) So it is rational to wager for God.

The many gods objection. Pascal's table has two columns, but the same reasoning could back a bet on any god who offers an infinite reward, including gods who punish belief in the others, or a god who rewards honest unbelievers. Once the columns multiply, the wager no longer tells you which way to bet. The Stanford Encyclopedia's survey calls this the objection generally regarded as the most important.

The objection from belief. You cannot believe something just by deciding to, any more than you can decide to believe it is raining. Pascal anticipated this, and his answer is practical.

Follow the way by which they began; by acting as if they believed, taking the holy water, having masses said, etc. Even this will naturally make you believe, and deaden your acuteness.

Critics add that a God worth betting on might not reward belief adopted as a bet, and that believing by training yourself not to question is a strange thing for a mathematician to recommend. Defenders reply that Pascal's point is about the passions that block belief, not about switching off thought, and that a great many things people come to believe, they come to believe by living as if they were true. Philosophers of decision theory have also pointed out that infinite rewards make the arithmetic misbehave: even a strategy like tossing a coin to decide whether to wager has infinite expected value, since there is some chance it leads to belief.

Bottom line: The wager changes the question from "is it true?" to "what is it rational to do if I cannot tell?", and its strongest critics accept the question and dispute the arithmetic and the two-column table.

Common misconceptions

• "The problem of evil proves that God does not exist." The logical version is widely thought to have been answered by the free will defense. The evidential version claims only that evil makes God improbable, and how improbable depends on the rest of the evidence.
• "A theodicy says that suffering is good." It says that God may permit suffering for the sake of goods, such as freedom or moral growth, that could not exist without the possibility of it.
• "Free will explains all the evil in the world." It addresses moral evil. Natural evil, such as the Lisbon earthquake, needs another reply, such as the soul-making theodicy's appeal to stable natural laws.
• "Pascal's wager is an argument that God exists." It is an argument that believing is the prudent bet if reason cannot decide. Pascal grants that reason cannot decide.
• "Pascal thought you could simply choose to believe." He agreed you cannot, and proposed a route: act as believers do, and belief will follow.

What you now know

• The Lisbon earthquake of 1 November 1755 destroyed a city on a holy day and gave Voltaire his counterexample to the claim that this is the best of all possible worlds.
• Hume's Philo states Epicurus's questions; Mackie's logical problem claims that God and evil are inconsistent.
• Plantinga's free will defense offers a possible reason for evil, which is enough to show consistency; Mackie later conceded the point.
• Rowe's evidential argument uses cases like the fawn to argue that pointless suffering makes God improbable.
• Hick's soul-making theodicy, skeptical theism and the G. E. Moore shift are the main replies, and each faces a serious objection.
• Pascal's wager argues that betting on God is rational under uncertainty; the many gods objection and the problem of believing at will are its best-known difficulties.

Sources

1. Hume, D. (2003). Dialogues concerning natural religion, Part X (Project Gutenberg eBook No. 4583; first published 1779). Project Gutenberg. gutenberg.org
2. Pascal, B. (2006). Pascal's Pensées (W. F. Trotter, Trans.; T. S. Eliot, Intro.), section 233 (Project Gutenberg eBook No. 18269). Project Gutenberg. gutenberg.org
3. Beebe, J. R. (n.d.). Logical problem of evil. Internet Encyclopedia of Philosophy. iep.utm.edu
4. Trakakis, N. (n.d.). The evidential problem of evil. Internet Encyclopedia of Philosophy. iep.utm.edu
5. Hájek, A. (2022). Pascal's wager. In E. N. Zalta & U. Nodelman (Eds.), Stanford Encyclopedia of Philosophy. plato.stanford.edu

A conflict about money, resolved

In 2010 the psychologist Daniel Kahneman and the economist Angus Deaton published an analysis of more than 450,000 responses to a daily Gallup survey of Americans. They measured two different things. Life evaluation was how people rated their lives when asked to think about them as a whole. Emotional wellbeing was the quality of their everyday experience: how much joy, stress, sadness and anger they had felt the day before. Life evaluation rose steadily as income rose. Emotional wellbeing rose too, but showed no further progress beyond an annual income of about $75,000. Health, caring for others, loneliness and smoking predicted daily emotions more strongly than income did.

In 2021 the psychologist Matthew Killingsworth, sampling people's feelings at moments throughout their day on a continuous scale, found no such plateau. So in 2023 the two sides did something unusual: Killingsworth, Kahneman and Barbara Mellers worked together, as an adversarial collaboration, to find out who was right. Their reanalysis found that the flattening was real, but only for the least happy group of people. Among happier people, happiness kept rising with income, and in the happiest group it accelerated.

Notice what the research had to decide before it could measure anything: what happiness is. A life can feel pleasant from moment to moment and be judged poor, or be judged well spent and feel stressful day to day. Which of those is a good life, or is it something else again? Philosophers call this the question of wellbeing, what makes a life go well for the person living it, and the ancient world gave three answers that are still the main options.

Epicurus: pleasure, rightly understood

Epicurus, who taught in a garden outside Athens around 300 BCE, held that pleasure is the good. His letter to his student Menoeceus survives because the historian Diogenes Laertius copied it into his Lives of the Eminent Philosophers. This is C. D. Yonge's translation of 1853, and it corrects at once the idea you probably have of what a philosopher of pleasure recommends.

... we think, contentment a great good, not in order that we may never have but a little, but in order that, if we have not much, we may make use of a little, being genuinely persuaded that those men enjoy luxury most completely who are the best able to do without it; and that everything which is natural is easily provided, and what is useless is not easily procured. And simple flavours give as much pleasure as costly fare, when everything that can give pain, and every feeling of want, is removed; and corn and water give the most extreme pleasure when any one in need eats them.

When, therefore, we say that pleasure is a chief good, we are not speaking of the pleasures of the debauched man, or those which lie in sensual enjoyment, as some think who are ignorant, and who do not entertain our opinions, or else interpret them perversely; but we mean the freedom of the body from pain, and of the soul from confusion.

The key terms are the absence of bodily pain and ataraxia, tranquillity of mind. To reach them, Epicurus sorts desires into kinds.

Of the desires, some are natural and necessary, some natural, but not necessary, and some are neither natural nor necessary, but owe their existence to vain opinions.

Drink when you are thirsty is natural and necessary, and easy to satisfy. Expensive food is natural but unnecessary. Fame and honours are vain: they are hard to get, harder to keep, and the anxiety of chasing them destroys the tranquillity they were supposed to bring. So Epicurean hedonism ends up recommending simple food, close friends, freedom from fear, and not wanting much. Epicurus added that prudence is the greatest good of all, since it is impossible to live pleasantly without living prudently, honourably and justly.

Key idea: For Epicurus the good is pleasure, but the pleasure that matters most is the absence of pain and anxiety, which is best reached by wanting little.

The Stoics: only what is up to you

Epictetus was born into slavery around 50 CE at Hierapolis, in what is now Turkey, and spent his youth in Rome as the slave of a wealthy freedman who served the emperor Nero. At some point he became lame. Freed, he taught Stoic philosophy, and when the emperor Domitian banished philosophers from Rome he opened a school at Nicopolis in Greece. He wrote nothing; his pupil Arrian recorded his teaching, including a short handbook, the Enchiridion, which begins like this in Thomas Wentworth Higginson's translation.

There are things which are within our power, and there are things which are beyond our power. Within our power are opinion, aim, desire, aversion, and, in one word, whatever affairs are our own. Beyond our power are body, property, reputation, office, and, in one word, whatever are not properly our own affairs.

From that division the Stoic view of the good life follows. Only virtue, the right use of what is within our power, is truly good; health, money and reputation lie outside it and cannot make a life good or bad. What disturbs us is our judgement about them.

Men are disturbed not by things, but by the views which they take of things. Thus death is nothing terrible, else it would have appeared so to Socrates.

Demand not that events should happen as you wish; but wish them to happen as they do happen, and you will go on well.

This is what modern readers find useful in Stoicism: exam results, other people's opinions and the weather are not in your power, and wasting your peace on them is a mistake of judgement you can correct. Now read the same principle at its hardest.

Never say of anything, "I have lost it," but, "I have restored it." Has your child died? It is restored. Has your wife died? She is restored. Has your estate been taken away? That likewise is restored.

Here the view strikes most readers as inhuman. The Stoic reply is that grief of the right kind is not forbidden; what is corrected is the judgement that something outside your will was yours to keep, and a man who had himself been owned by another had reason to know what can and cannot be taken from you.

The point: The Stoic good life lies entirely in what is up to you, your judgements and your choices, which is what makes it invulnerable and what makes it, to its critics, too hard.

Aristotle: virtue, and enough of the rest

You met Aristotle's answer in Lesson 11: the human good is rational activity in accordance with virtue over a complete life. What distinguishes him from the Stoics is his treatment of luck. This is Chase's translation.

Still it is quite plain that it does require the addition of external goods, as we have said: because without appliances it is impossible, or at all events not easy, to do noble actions: for friends, money, and political influence are in a manner instruments whereby many things are done ...

... for many changes and chances of all kinds arise during a life, and he who is most prosperous may become involved in great misfortunes in his old age, as in the heroic poems the tale is told of Priam: but the man who has experienced such fortune and died in wretchedness, no man calls happy.

Priam was the king of Troy who lived to see his sons killed and his city burned. Aristotle's point is that virtue is the heart of a good life but not the whole of it. You need friends to be generous to, enough money to act, health to do anything at all, and a fortune that does not collapse around you. Virtue can keep a good person from becoming miserable, he thinks, but it cannot guarantee happiness against everything.

Three answers, side by side

Reading the table

Read the fourth and fifth rows against the research. Kahneman and Deaton also reported that low income made the emotional pain of divorce, ill health and loneliness worse. Aristotle would not be surprised: lacking external goods mars a life. The Stoic would say the research measures feelings produced by judgements about things outside our power, which a trained judgement could change, and that it does not measure the good at all. Epicurus would point out that the strongest predictors of daily emotion, health and loneliness, are exactly the things he told his followers to care about: freedom from pain and the company of friends.

Then read the first row against the two measures. Emotional wellbeing is roughly what a hedonist cares about. Life evaluation is a person's own judgement of their life, which is still a feeling-involving report, not Aristotle's objective standard. None of the three ancient views says that wellbeing is simply whatever you judge it to be.

Worth holding on to: Before you can say whether money, fame or anything else makes a life better, you have to say what a good life is. Epicurus, the Stoics and Aristotle give three different answers, and each can explain some of the research in its own terms.

A test for hedonism: the experience machine

In Anarchy, State, and Utopia (1974), Nozick asked you to imagine a machine that could give you any experiences you wanted for the rest of your life: writing a great novel, making a friend, winning a final. While plugged in, you would not know you were in a tank with electrodes attached to your brain; it would all feel real. Would you plug in for life?

Most people say no. Nozick drew the lesson that we want to actually do things and actually be a certain kind of person, not merely to have the experience of doing and being. If wellbeing were nothing but pleasant experience, the machine would be the best life available, so the widespread refusal counts against hedonism.

(1) If hedonism is true, a life on the experience machine is at least as good for the person as any life outside it.
(2) A life on the experience machine is not as good for the person as a life of real achievement and real relationships.
(3) So hedonism is false.

The argument is valid, so hedonists must deny premise (2), and some do: they argue that our refusal reflects fear of the unfamiliar or a mistaken attachment to "reality", not a judgement about wellbeing. Notice, too, what the argument cannot do. It counts against the view that pleasure is the only thing that makes a life go well. It does not show that pleasure does not matter, and neither the Stoics nor Aristotle would plug in, for their own reasons: the Stoic because virtue requires real choices, Aristotle because a good life is an activity, not the experience of one.

Common misconceptions

• "Epicureans were party-goers." Epicurus explicitly rejects "the pleasures of the debauched man" and recommends corn, water, friendship and freedom from anxiety.
• "Stoicism means feeling nothing." Epictetus's target is judgement: the belief that what is beyond your power is yours to control. The aim is a correct response, not an absence of response.
• "Aristotle thought money makes you happy." External goods are instruments and conditions of a good life. The good itself is virtuous activity.
• "Research has shown that money does not buy happiness." The 2023 adversarial collaboration found that happiness rises with income for most people, levelling off only for the least happy. What the research cannot settle is which kind of happiness a good life consists in.
• "Refusing the experience machine proves pleasure is worthless." It shows at most that pleasure is not the only thing that matters.

Looking back

• Kahneman and Deaton separated emotional wellbeing from life evaluation; the 2023 adversarial collaboration found happiness levelling off with income only for the least happy.
• Epicurus identifies the good with pleasure, understood as freedom from bodily pain and mental disturbance, reached by wanting little and keeping friends.
• Epictetus divides things into what is within our power and what is not; for the Stoics only virtue is good, and misfortune cannot touch it.
• Aristotle holds that a good life needs external goods as well as virtue, so great misfortune, like Priam's, can spoil it.
• The experience machine counts against the view that wellbeing is only pleasant experience; it does not show that pleasure is unimportant.

Sources

1. Diogenes Laertius. (2018). The lives and opinions of eminent philosophers (C. D. Yonge, Trans.; first published 1853), Book X, Epicurus (Project Gutenberg eBook No. 57342). Project Gutenberg. gutenberg.org
2. Epictetus. (2014). The enchiridion (T. W. Higginson, Trans.; Project Gutenberg eBook No. 45109). Project Gutenberg. gutenberg.org
3. Aristotle. (1911). The Nicomachean ethics of Aristotle (D. P. Chase, Trans.; first published 1847), Book I. Wikisource. wikisource.org
4. Crisp, R. (2026). Well-being. In E. N. Zalta & U. Nodelman (Eds.), Stanford Encyclopedia of Philosophy. plato.stanford.edu
5. Killingsworth, M. A., Kahneman, D., & Mellers, B. (2023). Income and emotional well-being: A conflict resolved. Proceedings of the National Academy of Sciences, 120(10), e2208661120. pubmed.ncbi.nlm.nih.gov

A personality quiz, and 87 million people

In 2013 a data scientist called Aleksandr Kogan and his company built a Facebook app called "This Is Your Digital Life". It asked users a series of questions to build a psychological profile. About 270,000 people installed it. Through the way Facebook's platform then worked, the app could also collect data about each user's Facebook friends, who had never installed anything. By the time the story broke in 2018, through a former employee of the British consulting firm Cambridge Analytica, the app was estimated to have harvested data from up to 87 million Facebook profiles, which the firm had used for political advertising, including in the 2016 United States presidential campaigns.

The consequences for the company were real. On 24 July 2019 the US Federal Trade Commission announced that Facebook would pay a record-breaking $5 billion penalty and accept new restrictions, to settle charges that it had violated a 2012 order by deceiving users about their ability to control the privacy of their information. In the UK, Facebook agreed to pay the Information Commissioner's Office a fine of £500,000 for exposing users' data to a serious risk of harm. Cambridge Analytica filed for bankruptcy.

Now make the problem personal. Suppose you were one of the 87 million. You never installed the app. Nothing bad happened to you that you know of: no money lost, no messages leaked to your family. Were you wronged? And if you were, what exactly was the wrong? That question sounds easy, and the three standard answers to it give different results.

First answer: privacy protects you from harm

The most influential test for when society may interfere with a person comes from John Stuart Mill's On Liberty (1859).

That the only purpose for which power can be rightfully exercised over any member of a civilised community, against his will, is to prevent harm to others. His own good, either physical or moral, is not a sufficient warrant.

Mill was writing about the power of governments and public opinion over individuals, but his principle suggests an account of privacy: collecting and using information about someone is wrong when, and because, it harms them or exposes them to a risk of harm. On this account the harvested friends were wronged to the extent that the data exposed them to real risks: of being targeted, deceived, discriminated against, or having their votes influenced by messages tuned to their weaknesses. The UK regulator's wording, a "serious risk of harm", fits this view exactly.

The problem is the case with no harm at all. Imagine someone who reads your diary every night for a year, tells no one, changes nothing, and is never discovered. Most people think you have been wronged, and a purely harm-based account struggles to say how.

Second answer: privacy is control

A second account says privacy is your control over information about yourself: who gets it, and what they do with it. On this view the friends were wronged whether or not harm followed, because information about them was taken without their consent. Their friends who installed the app could consent for themselves; they could not consent on behalf of 87 million others.

The problem runs the other way. You do not control most information about you, and nobody thinks you should: your neighbour sees when you leave the house, your teacher knows your marks, the shop knows what you bought. If privacy were control, ordinary life would be one long violation of it.

Third answer: information flows have norms

The philosopher Helen Nissenbaum proposed in 2004 an account that handles both problems, called contextual integrity. Every context in which we share information, medicine, friendship, school, shopping, comes with norms about where that information may flow next. Your doctor may tell a specialist about your condition; she may not tell your employer. A privacy violation is a flow of information that breaks the norms of the context in which it was shared.

(1) Information shared in a context carries that context's norms about where it may go next.
(2) Friends shared their likes, birthdays and posts with each other under the norms of friendship.
(3) Those norms do not permit that information to be passed to a firm building psychological profiles for political advertising.
(4) So the flow broke the context's norms, and the friends were wronged, whether or not any further harm followed.

Contextual integrity explains the diary reader, since reading a diary breaks the norms under which it was written, without calling every glance from a neighbour a violation. The Stanford Encyclopedia's survey of privacy and information technology lists it among the main moral reasons for protecting privacy, alongside the prevention of harm, the unfairness of large inequalities of information, and the threat to autonomy when people know they are watched. Its critics reply that the norms of a context are often unclear or disputed, especially online, where the contexts are new and the companies write the rules.

So what?: "Were you wronged if nothing bad happened?" gets three different answers. The harm view says only if you were put at risk; the control view says yes, because you did not consent; contextual integrity says yes, because the information went somewhere the context never allowed.

The second problem: an experiment on your feed

In June 2014 researchers from Facebook and Cornell University published a study in the Proceedings of the National Academy of Sciences. For about 689,000 Facebook users, they had reduced the amount of emotional content in the News Feed. When positive posts from friends were reduced, users went on to write fewer positive posts and more negative ones; when negative posts were reduced, the opposite happened. The authors described it as experimental evidence of emotional contagion through social networks, while noting that the effects were small. The users were not told they were in an experiment. After a public outcry, the journal published an editorial expression of concern about whether the study had met the principle of obtaining informed consent and allowing participants to opt out, and a privacy group filed a complaint with the Federal Trade Commission.

This is the attention economy made visible. Services like Facebook are free to use because they are paid for by advertisers, and what advertisers buy is attention. A company paid for attention has every reason to design its feed to capture and hold it, and the experiment showed that the same feed can also shift what people feel.

The case against the experiment. Kant's formula of humanity from Lesson 10 is the natural tool. Deliberately altering people's emotional environment to see what happens to them, without their knowledge, treats them as material for someone else's purposes, in a way they could not agree to without the experiment losing its point. That is the mark of manipulation: influence that works around a person's capacity to reason rather than through it. Persuasion gives you reasons you can weigh; manipulation changes you by a route you cannot inspect.

The case for it. Defenders pointed out that the News Feed is always filtered by an algorithm choosing what users see; the study changed one filter among thousands that are changed routinely to test what works, and the only unusual thing was that the results were published. Newspapers choose frightening headlines and supermarkets arrange shelves to make you buy more. Users had agreed to a data policy. If this was manipulation, so is much of ordinary commercial life.

The reply to the defence goes to consent. Agreeing to terms of service that almost nobody reads, and that cannot be refused without giving up a service your friends all use, is not what researchers mean by informed consent, which requires understanding what you are agreeing to and a real option to say no. The disagreement that remains is whether routine commercial testing should meet the standard that research on human beings is held to.

In short: The emotional contagion study turned a hidden feature of the attention economy into a published result, and the argument about it is an argument about consent and manipulation, not about whether Facebook meant well.

Four questions for any data or attention practice

Work the two problems and a general method appears. Before judging a practice, ask four questions.

Notice that the four questions do not deliver one verdict automatically; they locate the disagreement. Someone who defends targeted political advertising has to say which answer in the first column they dispute. Someone who condemns every experiment on a platform has to explain why it is worse than the tests that shape every feed anyway.

Remember: Harm, consent, context and manipulation are four different grounds for objecting to a data practice. Say which one you are relying on, because each has different counterexamples.

Common misconceptions

• "If you have nothing to hide, you have nothing to fear." Privacy also protects autonomy, relationships and freedom from misuse. People change what they say and do when they know they are watched, which is called the chilling effect, even if nothing is done to them.
• "Information that is already public cannot be a privacy violation." Contextual integrity explains why moving public scraps of information into a new context, and combining them into a profile, can still break the norms under which they were shared.
• "Clicking I agree is consent." Informed consent requires understanding what is agreed and a real option to refuse. Unread terms attached to a service you cannot realistically avoid meet neither condition well.
• "Manipulation needs lies." An influence can manipulate by selecting true information, or by exploiting mood and habit, so long as it works around your ability to weigh reasons rather than through it.
• "The 2014 study proved that Facebook controls our moods." The measured effects were small, and critics questioned the method, which inferred emotion by counting positive and negative words in posts.

Summing up

• The Cambridge Analytica app, installed by about 270,000 people, harvested data from up to 87 million profiles; the FTC imposed a $5 billion penalty on Facebook in 2019.
• A harm-based account, drawing on Mill's harm principle, struggles with wrongs that cause no harm; a control account makes ordinary life a violation.
• Contextual integrity locates the wrong in information flowing in breach of the norms of the context where it was shared.
• The 2014 emotional contagion study altered about 689,000 users' feeds without their knowledge and drew an editorial expression of concern about consent.
• Manipulation works around a person's reasons; persuasion works through them. Terms of service rarely amount to informed consent.
• Four questions, about flow, consent, harm and manipulation, locate what is really disputed in any data or attention practice.

Sources

1. Wikipedia. (2026). Facebook-Cambridge Analytica data scandal. en.wikipedia.org
2. Federal Trade Commission. (2019, July 24). FTC imposes $5 billion penalty and sweeping new privacy restrictions on Facebook [Press release]. ftc.gov
3. Kramer, A. D. I., Guillory, J. E., & Hancock, J. T. (2014). Experimental evidence of massive-scale emotional contagion through social networks. Proceedings of the National Academy of Sciences, 111(24), 8788-8790. pubmed.ncbi.nlm.nih.gov
4. Mill, J. S. (2011). On liberty, Chapter I (Project Gutenberg eBook No. 34901; first published 1859). Project Gutenberg. gutenberg.org
5. van den Hoven, J., Blaauw, M., Pieters, W., & Warnier, M. (2024). Privacy and information technology. In E. N. Zalta & U. Nodelman (Eds.), Stanford Encyclopedia of Philosophy. plato.stanford.edu

9.58 p.m., North Mill Avenue, Tempe

On the night of 18 March 2018 a modified 2017 Volvo XC90 was completing its second loop of a test route in Tempe, Arizona. It belonged to the Advanced Technologies Group of Uber, which had fitted it with a developmental automated driving system, and it had been driving itself for about nineteen minutes. In the driver's seat sat a human operator whose job was to watch the road and take over if anything went wrong. The road was dry and lit by street lamps. The car was travelling at about 45 miles an hour.

At about that time Elaine Herzberg, who was forty-nine, began walking across the road pushing a bicycle, at a place where there was no crosswalk. According to the National Transportation Safety Board, the driving system detected her 5.6 seconds before impact. It kept tracking her until the crash, but it never accurately classified her as a pedestrian or predicted her path; the recorded data show it treating her first as an unknown object, then as a vehicle, then as a bicycle. By the time it determined that a collision was imminent, it was too late for its braking specifications, and the system had been designed not to brake hard on its own, relying instead on the operator to intervene. The operator was looking down. Police later reported that her phone was streaming a television programme. Herzberg died of her injuries in hospital, the first recorded pedestrian killed by a self-driving car.

Who was responsible? The operator, the engineers, the company, the state that allowed the testing, the victim, the car? This lesson follows the case through the institutions that tried to answer that question, and the philosophical tools arrive as the case needs them.

The oldest test for blame

The first tool is two thousand three hundred years old. Aristotle opens Book III of the Nicomachean Ethics by saying that praise and blame belong only to voluntary actions, and then says what makes an action involuntary. This is Chase's translation.

Now since Virtue is concerned with the regulation of feelings and actions, and praise and blame arise upon such as are voluntary, while for the involuntary allowance is made, and sometimes compassion is excited, it is perhaps a necessary task for those who are investigating the nature of Virtue to draw out the distinction between what is voluntary and what involuntary; and it is certainly useful for legislators, with respect to the assigning of honours and punishments. Involuntary actions then are thought to be of two kinds, being done either on compulsion, or by reason of ignorance.

Philosophers still use exactly these two conditions. The control condition: you are responsible only for what you had some control over. The epistemic condition: you are responsible only for what you knew, or should have known. Run them over the parties in the car.

The car had "control" in a mechanical sense, but it did not know anything in the sense Aristotle means, and it had no idea what a person is. Nobody proposes to blame it. The operator had control, since she could brake, and she should have known the risk, since watching the road was her job. The engineers had no control on the night, but they had made choices months earlier, about the braking design and about the operator's role, whose risks they knew or should have known. So far the conditions point at humans. The question is how many of them.

What matters here: Responsibility needs control and knowledge. A machine that has neither in the moral sense is not the kind of thing that can be blamed, which pushes the question back onto the people who built it, deployed it and watched it.

March 2019: the prosecutors decide

Criminal law answered first. Because the county prosecutor's office normally responsible had worked with Uber on a road safety campaign, the case went to a neighbouring county. On 4 March 2019 the Yavapai County Attorney wrote that there was no basis for criminal liability for Uber. The operator was later charged with negligent homicide, pleaded guilty to endangerment, and was sentenced to three years' probation. In the eyes of the criminal law, one person answered for the death.

That outcome makes sense given what criminal law is for: it asks whether a particular person's conduct meets the definition of a particular crime, and it is built around individuals who act. But notice what it leaves out. The operator was one of Uber's staff, placed in a job that Uber had designed, in a car whose emergency braking had been designed out, after Uber had reduced the crew in each car from two employees to one.

November 2019: the safety board decides

The National Transportation Safety Board investigates crashes to prevent the next one, not to punish, and it reached a different kind of answer. Its finding, as published on its investigation page, reads:

The probable cause of the crash in Tempe, Arizona, was the failure of the vehicle operator to monitor the driving environment and the operation of the automated driving system because she was visually distracted throughout the trip by her personal cell phone. Contributing to the crash were the Uber Advanced Technologies Group's (1) inadequate safety risk assessment procedures, (2) ineffective oversight of vehicle operators, and (3) lack of adequate mechanisms for addressing operators' automation complacency, all a consequence of its inadequate safety culture. Further factors contributing to the crash were (1) the impaired pedestrian's crossing of N. Mill Avenue outside a crosswalk, and (2) the Arizona Department of Transportation's insufficient oversight of automated vehicle testing.

The board's original text sets off its last clause with a dash, printed here as a comma. Count the parties: the operator, the company, the pedestrian, the state. Where the criminal law found one person to charge, the safety board found a web of causes. This is what philosophers call the problem of many hands. When many people and organisations each contribute to an outcome, it becomes hard to say who is responsible for what, and easy for each to point at the others. The Stanford Encyclopedia's entry on computing and moral responsibility traces the problem through earlier disasters, such as the Therac-25 radiation machine, which massively overdosed patients in the 1980s through a mix of software faults and institutional failures.

The phrase "automation complacency" matters most. People asked to watch an automated system that almost never fails are known to stop paying attention. Uber, the board found, had no adequate mechanism for dealing with that. So the operator's distraction, which the criminal law treated as her fault alone, was also, on the board's analysis, a predictable result of how the job was designed.

The upshot: Causal responsibility and blame are different things. Several parties can be among the causes of a death, and more than one of them can deserve blame for a separate failure of their own.

The responsibility gap

The Tempe case is, in one way, the easy kind. The system was not very sophisticated, the humans around it made identifiable choices, and Aristotle's conditions could be applied to each of them. Philosophers worry about a harder kind. Systems that learn from data behave in ways their designers did not program line by line and cannot fully predict. If such a system causes harm, the designers may lack the knowledge Aristotle's second condition requires, the users may lack the control his first condition requires, and the system itself is not a moral agent. Then, the argument goes, nobody is responsible. This is the responsibility gap, a phrase associated especially with Robert Sparrow's work on autonomous weapons.

(1) A person is blameworthy for a harm only if they had control over it and knew, or should have known, it might occur.
(2) For some harms caused by learning systems, no designer, operator or user had that control and knowledge.
(3) The system itself is not a moral agent and cannot be blamed.
(4) So for some such harms, no one is blameworthy.

The argument is valid, and there are three main replies, as the Stanford Encyclopedia's entry on the ethics of artificial intelligence surveys them. The first attacks premise (2): design systems so that some human always keeps meaningful human control, meaning they can understand what the system is doing and intervene in time, so that the gap never opens. The Tempe car failed that standard, since the system was not designed to alert the operator or brake on its own. The second accepts the conclusion and says that deciding to deploy a system you cannot fully predict is itself something you can be blamed for, which moves the responsibility back to the decision to deploy. The third says the right question is not who is to blame but who should bear the risk and the cost, a question of fair distribution rather than of guilt.

February 2024: a chatbot and a funeral

The law has begun to answer the gap in its own way. In November 2022 Jake Moffatt's grandmother died, and he went to Air Canada's website to book flights to her funeral. The website's chatbot told him he could buy tickets at full price and claim the cheaper bereavement fare afterwards, within ninety days. That was wrong: the airline's actual policy, on another page of the same website, did not allow claims after travel. When Air Canada refused the refund, Moffatt took it to British Columbia's Civil Resolution Tribunal.

Air Canada argued that the chatbot was "a separate legal entity that is responsible for its own actions". The tribunal member called the submission remarkable and rejected it. The chatbot was part of Air Canada's website, and the airline was responsible for all the information on its website, whether it came from a static page or a chatbot. In Moffatt v. Air Canada (2024) the airline was held liable for negligent misrepresentation and ordered to pay the fare difference with interest and fees, well under a thousand Canadian dollars in all.

The sum was small; the principle was not. A company that deploys an automated system to speak or act for it cannot step out from behind it when it goes wrong. That is the second reply to the responsibility gap, put into law: the decision to deploy carries the responsibility.

Why this matters: "The AI did it" has not so far been accepted as an answer, either by safety investigators or by courts. They keep finding a human decision behind the machine.

Common misconceptions

• "If a machine made the decision, nobody is responsible." The NTSB found an operator, a company and a state responsible for failures of their own, and the Air Canada tribunal held the company responsible for its chatbot.
• "Responsibility is a fixed amount to be shared out." The operator can be fully responsible for her distraction and Uber fully responsible for designing a job that invited it. These are separate failures, not slices of one pie.
• "Being a cause means being to blame." A gust of wind can cause a crash and deserves no blame. The pedestrian's crossing was a contributing factor; whether she was blameworthy is a separate question about her control and knowledge.
• "The guilty plea shows the operator was the only one at fault." A plea settles a legal charge against one person. The safety board's analysis found organisational failures that criminal law was not asked to judge.
• "A system that makes decisions is a moral agent." Choosing an output is not the same as understanding what is at stake. Whether future systems could meet Aristotle's conditions is an open question; current ones are not treated as meeting them.

Recap

• On 18 March 2018 an Uber test vehicle in self-driving mode struck and killed Elaine Herzberg in Tempe; the system detected her 5.6 seconds before impact but never classified her as a pedestrian, and relied on the operator to brake.
• Aristotle's two conditions, control and knowledge, still frame responsibility: involuntary actions are done under compulsion or through ignorance.
• Prosecutors found no basis for criminal liability for Uber; the operator pleaded guilty to endangerment.
• The NTSB named the operator's distraction as the probable cause and Uber's safety culture, the pedestrian's crossing and Arizona's oversight as contributing factors, an instance of the problem of many hands.
• The responsibility gap argument says learning systems may cause harms for which no one meets the conditions of blame; replies include meaningful human control, blaming the decision to deploy, and fair distribution of risk.
• In Moffatt v. Air Canada (2024) a tribunal rejected the claim that a chatbot was responsible for its own actions and held the airline liable.

Sources

1. National Transportation Safety Board. (2019). Collision between vehicle controlled by developmental automated driving system and pedestrian, Tempe, Arizona, March 18, 2018 (Investigation HWY18MH010). ntsb.gov
2. Aristotle. (1911). The Nicomachean ethics of Aristotle (D. P. Chase, Trans.; first published 1847), Book III. Wikisource. wikisource.org
3. Noorman, M. (2023). Computing and moral responsibility. In E. N. Zalta & U. Nodelman (Eds.), Stanford Encyclopedia of Philosophy. plato.stanford.edu
4. Müller, V. C. (2026). Ethics of artificial intelligence and robotics. In E. N. Zalta & U. Nodelman (Eds.), Stanford Encyclopedia of Philosophy. plato.stanford.edu
5. Wikipedia. (2026). Moffatt v. Air Canada (2024 BCCRT 149). en.wikipedia.org

A question in a footnote, 1789

In a footnote to Chapter XVII of his Introduction to the Principles of Morals and Legislation, Bentham turned from the law of human beings to the treatment of animals. Here is the passage as the Stanford Encyclopedia's entry on the moral status of animals quotes it.

The day may come, when the rest of the animal creation may acquire those rights which never could have been withholden from them but by the hand of tyranny. The French have already discovered that the blackness of skin is no reason why a human being should be abandoned without redress to the caprice of a tormentor. It may come one day to be recognized, that the number of legs, the villosity of the skin, or the termination of the os sacrum, are reasons equally insufficient for abandoning a sensitive being to the same fate. What else is it that should trace the insuperable line? Is it the faculty of reason, or perhaps, the faculty for discourse? ... the question is not, Can they reason? nor, Can they talk? but, Can they suffer?

"Villosity" means hairiness, and the end of the os sacrum is where a tail begins. Bentham's comparison is deliberate and uncomfortable: the difference between a human and a dog, he suggests, may one day look as irrelevant to how they should be treated as the colour of a person's skin.

The question he asked is live law. On 28 April 2022 the UK's Animal Welfare (Sentience) Act received Royal Assent. It set up an Animal Sentience Committee to report on how government policy affects the welfare of animals as sentient beings, and it defined "animal", for its purposes, as any vertebrate other than a human being, any cephalopod mollusc, such as an octopus or squid, and any decapod crustacean, such as a crab or lobster. Parliament had drawn Bentham's line at the capacity to suffer, and drawn it well beyond mammals. The dispute this lesson sets out is whether that line is the right one.

The other answer: animals as machines

Descartes, whose dualism you met in Lesson 7, gave the opposite answer a century and a half earlier. In Part V of his Discourse on the Method (1637), in Veitch's translation, he argues that animals' skill at some tasks shows the absence of mind, not its presence.

It is also very worthy of remark, that, though there are many animals which manifest more industry than we in certain of their actions, the same animals are yet observed to show none at all in many others: so that the circumstance that they do better than we does not prove that they are endowed with mind, for it would thence follow that they possessed greater reason than any of us, and could surpass us in all things; on the contrary, it rather proves that they are destitute of reason, and that it is nature which acts in them according to the disposition of their organs ...

He goes on to compare them to a clock, which keeps time better than we can while understanding nothing. If animals are mechanisms without minds, there is nothing it is like to be one, in the phrase you met from Nagel, and so nothing that can be wronged.

Kant took a more careful middle position. Only rational beings are ends in themselves, and beings that are not rational have, in his words as the Stanford Encyclopedia quotes them, "only a relative value as means". Yet Kant condemned cruelty to animals, on the ground that a person who is cruel to animals becomes hard in dealing with people. On this indirect duty view we have duties concerning animals but not duties to them: the wrong of kicking a dog lies in what it does to the kicker and to other humans.

Position A: animals count for their own sake

From suffering. In Animal Liberation (1975), Peter Singer built Bentham's footnote into an argument. His principle of equal consideration of interests says that like interests count equally, whoever has them. A pig's interest in not suffering is like a child's interest in not suffering, so it counts the same. Giving it less weight merely because the pig belongs to another species is what Singer calls speciesism, a prejudice of the same form as racism or sexism. Notice that this is not a claim that animals and humans must be treated identically. A pig has no interest in voting, so equal consideration gives it no vote. What it gives the pig is an equal claim not to suffer.

From being someone. Tom Regan, in The Case for Animal Rights (1983), rejected Singer's utilitarian framework. What matters, he argued, is that many animals, like humans, are subjects of a life: they have beliefs, desires, memory, a sense of their own future, and a welfare that matters to them whatever anyone else thinks. Such beings have rights, and rights, unlike interests in Singer's calculation, may not be traded away for a greater total good. On Regan's view, using animals as mere resources is wrong even if it could be made painless.

The argument from marginal cases. Both positions rest on an argument whose name, as the Stanford Encyclopedia notes, is unfortunate: the argument from marginal cases.

(1) If only beings with reason and self-awareness have moral status, then newborn infants and some severely cognitively disabled humans have no moral status.
(2) Newborn infants and severely cognitively disabled humans do have moral status.
(3) So it is not only beings with reason and self-awareness that have moral status.
(4) The most plausible property that infants share with the beings we count is the capacity to suffer and to have a welfare of their own, which many animals also have.
(5) So many animals have moral status too.

Premises (1) to (3) are modus tollens and valid. The weight is on premise (4).

The point: If moral status rests on the capacity to suffer, species membership by itself cannot decide who counts, and the argument from marginal cases is designed to show that no other line is consistent with how we treat infants.

Position B: moral status is tied to persons and relationships

Defenders of a sharper line between humans and animals have their own strong arguments, and they are not the same as Descartes'.

Morality as an agreement among agents. On contractualist views, morality is a system of rules that rational agents could agree to for their mutual benefit. Animals cannot make or keep agreements, so they are not parties to it. They may still be protected, because many humans care about them and because cruelty corrupts character, which brings back Kant's indirect duty.

Replies to marginal cases. Such views answer premise (4) in several ways. Infants belong to a kind of being, the rational kind, whose normal members are persons, and they will usually grow into persons themselves. Human beings with severe cognitive disabilities stand in relationships of family and community that give others special duties toward them, a point care ethicists from Lesson 11 would press. The critic's reply is that these answers either make moral status depend on what others feel about a being, which seems the wrong kind of thing, or give an animal and a human with the same capacities different status on grounds the animal cannot help.

The practical point. Some defenders of Position B add that equal consideration for animals, taken seriously, would require vast changes: to diet, to farming, to medical research and to the control of pests. They treat that as a reason to doubt the premise. Their opponents treat it as the conclusion.

What would settle it. Part of the dispute is empirical: which animals can suffer, and how much. That is why the UK bill was amended to include octopuses and lobsters only after a scientific review found strong evidence of sentience, and why the Act lets the list be extended as the evidence changes. The rest is philosophical: whether the capacity to suffer is enough for moral status, or whether rational agency or membership in a moral community is also needed. Evidence can extend the circle of sentient beings; only argument can say whether sentience is what the circle is drawn around.

Beyond animals: can a river have standing?

Environmental ethics pushes the question further. Anthropocentrism holds that only humans have moral standing, so nature matters as it affects human beings, now and in the future. Biocentrism extends standing to every living thing. Ecocentrism extends it to wholes: species, ecosystems, the land. The forester Aldo Leopold put the ecocentric principle in A Sand County Almanac (1949): "A thing is right when it tends to preserve the integrity, stability, and beauty of the biotic community."

In 1973 the philosopher Richard Routley devised a thought experiment to test anthropocentrism, the last man argument. Imagine that after a world catastrophe, the last surviving human being, knowing he will soon die, sets about destroying every remaining living thing. No human interest is harmed, since there will be no humans left to be harmed. If anthropocentrism is true, he does nothing wrong. Routley pointed out that most people nonetheless judge that he does something wrong, which suggests that nature has value that does not depend on human interests.

Now the real case. The Whanganui River, the third-longest in New Zealand at about 290 kilometres, is of central importance to the Māori iwi of the region, who had pursued claims over it for well over a century. On 15 March 2017 the New Zealand Parliament passed the Te Awa Tupua (Whanganui River Claims Settlement) Act, which gave the river a legal identity with the rights, duties and liabilities of a legal person, to be represented by two officials, one appointed by the iwi and one by the government. The minister responsible, Chris Finlayson, said that some people would find it strange, but that it was "no stranger than family trusts, or companies, or incorporated societies".

The case can be read two ways, and the disagreement between the readings is the environmental dispute in miniature. An ecocentric reading says the law recognises that the river has standing of its own, as the Māori understanding of the river as an ancestor holds. An anthropocentric reading, closer to the minister's comparison with companies, says legal personhood is a device for protecting what the river means to people, and that nothing in the law requires thinking the river itself can be wronged.

So what?: The question of moral status keeps widening, from humans to sentient animals to living things to whole rivers, and at each step the dispute is the same: whether value requires a being who can experience it, or whether some things matter even with no one to notice.

Common misconceptions

• "Equal consideration means treating animals exactly like humans." It means giving like interests equal weight. Animals lack many human interests, such as in voting or education, so equal consideration does not give them those things.
• "Singer and Regan hold the same view." Singer is a utilitarian who weighs interests; Regan argues for rights that may not be traded for greater total good. They agree that animals count, and disagree about how.
• "The indirect duty view permits cruelty to animals." Kant condemned cruelty to animals. The view locates the wrong in its effects on human character and relationships rather than in a wrong done to the animal.
• "A river that is a legal person must have feelings." Legal persons include companies and trusts, which have no feelings. Whether the Whanganui has moral standing is a separate question from its legal status.
• "Science can settle which beings have moral status." Science can show which beings are likely to be sentient. Whether sentience is what matters morally is a philosophical question.

What to remember

• Bentham argued that the question is not whether animals can reason or talk but whether they can suffer; the UK's 2022 sentience law now covers vertebrates, cephalopods and decapods.
• Descartes treated animals as mechanisms without minds; Kant held that we have only indirect duties concerning animals.
• Singer's equal consideration of interests counts like interests equally across species; Regan argues that subjects-of-a-life have rights.
• The argument from marginal cases says that any line excluding animals by rationality also excludes some humans; the replies appeal to kinds, potential and relationships.
• Anthropocentrism, biocentrism and ecocentrism differ over whether humans, all living things or whole ecosystems have standing; Routley's last man tests anthropocentrism.
• New Zealand made the Whanganui River a legal person in 2017, a case that can be read in ecocentric or anthropocentric terms.

Sources

1. Gruen, L., & Monsó, S. (2024). The moral status of animals. In E. N. Zalta & U. Nodelman (Eds.), Stanford Encyclopedia of Philosophy. Includes the quoted passage from Bentham (1789), Chapter XVII. plato.stanford.edu
2. Descartes, R. (1993). Discourse on the method of rightly conducting one's reason and of seeking truth in the sciences (J. Veitch, Trans.; Project Gutenberg eBook No. 59; first published 1637), Part V. Project Gutenberg. gutenberg.org
3. Animal Welfare (Sentience) Act 2022, c. 22, section 5. legislation.gov.uk. legislation.gov.uk
4. Brennan, A., & Lo, N. Y. S. (2021). Environmental ethics. In E. N. Zalta & U. Nodelman (Eds.), Stanford Encyclopedia of Philosophy. plato.stanford.edu
5. Wikipedia. (2026). Whanganui River, section on legal personhood. en.wikipedia.org

A famous man at a standstill

By the late 1870s Leo Tolstoy had written War and Peace and Anna Karenina and was one of the most celebrated writers alive. In a short book he called A Confession, which church censors suppressed when it first went to press in 1882 and which was published in Geneva in 1884, he described what happened to him next. This is the translation by Louise and Aylmer Maude.

And all this befell me at a time when all around me I had what is considered complete good fortune. I was not yet fifty; I had a good wife who loved me and whom I loved, good children, and a large estate which without much effort on my part improved and increased.

My life came to a standstill. I could breathe, eat, drink, and sleep, and I could not help doing these things; but there was no life, for there were no wishes the fulfilment of which I could consider reasonable. If I desired anything, I knew in advance that whether I satisfied my desire or not, nothing would come of it. Had a fairy come and offered to fulfil my desires I should not have known what to ask.

He put his state into an old fable.

... a traveller overtaken on a plain by an enraged beast. Escaping from the beast he gets into a dry well, but sees at the bottom of the well a dragon that has opened its jaws to swallow him. ... Then he sees that two mice, a black one and a white one, go regularly round and round the stem of the twig to which he is clinging and gnaw at it. And soon the twig itself will snap and he will fall into the dragon's jaws. The traveller sees this and knows that he will inevitably perish; but while still hanging he looks around, sees some drops of honey on the leaves of the twig, reaches them with his tongue and licks them.

The black and white mice, he explains, are night and day, the dragon is death, and the honey is the ordinary pleasures of family and work, which no longer tasted sweet to him. Tolstoy lived another thirty years and did some of his most admired work in them. His crisis is worth following because he turned it into an argument, and because the questions it raised are ones most thoughtful people meet at some point. If they ever stop feeling like a philosophical puzzle and start to feel like despair, that is worth saying out loud to someone you trust, a friend, a parent, a teacher or a doctor; there are also free helplines whose whole job is to listen.

The question he could not answer

In chapter V Tolstoy states the question in three forms.

It was: 'What will come of what I am doing to-day or shall do to-morrow? What will come of my whole life?' Differently expressed, the question is: 'Why should I live, why wish for anything, or do anything?' It can also be expressed thus: 'Is there any meaning in my life that the inevitable death awaiting me does not destroy?'

Two questions are tangled together there. One is about meaning: what, if anything, makes a life worth living in a way that goes beyond whether it is pleasant? The other is about death: does the fact that it all ends take the meaning away? This lesson takes them one at a time, as Tolstoy did.

Philosophers distinguish meaning from happiness. A life plugged into the experience machine from Lesson 17 could be happy; very few people think it would be meaningful. And a meaningful life, such as that of someone who gave up comfort to care for a sick parent, need not be the happiest available. The Stanford Encyclopedia's entry on the meaning of life sorts the answers into families.

Supernaturalism. Meaning requires a relation to God, or a soul that outlasts death. Without one, everything we do is eventually undone, which is exactly Tolstoy's fear.

Subjectivism. A life is meaningful to the extent that the person cares about what they do and finds fulfilment in it. Nothing outside the person has to certify it.

Objectivism. Meaning requires engaging with things of real worth, often summed up as the good, the true and the beautiful: love and friendship, understanding, creating or appreciating something excellent. Caring is not enough if what you care about is worthless.

A hybrid. The American philosopher Susan Wolf has argued that meaning arises when subjective attraction meets objective attractiveness: when you are gripped by something that is actually worth being gripped by.

Key idea: Meaning and happiness come apart. The live dispute is whether meaning depends on something beyond this life, on what you care about, on what is really worth caring about, or on a combination.

A test case: Sisyphus

In Greek myth the gods condemned Sisyphus to roll a stone up a hill, watch it roll down, and roll it up again for ever. In 1942 Albert Camus made him the hero of The Myth of Sisyphus, arguing that the universe offers no meaning and that the right response is not despair but defiance; his essay ends with the line "One must imagine Sisyphus happy."

In 1970 the American philosopher Richard Taylor sharpened the test. Suppose the gods, in a moment of mercy, implanted in Sisyphus a burning desire to roll stones, so that he wanted nothing more. Is his life now meaningful?

Taylor himself drew the subjectivist conclusion. Most of his critics find it hard to believe that implanting a desire could make a pointless task meaningful, which is the main intuition behind objectivism. The subjectivist's reply is that the objectivist has to say who decides what is worthwhile, and that every candidate list sounds suspiciously like the tastes of philosophers.

How Tolstoy came through

Tolstoy's own answer was supernaturalist, reached by an unexpected route. He looked first to science and philosophy and found nothing, and then noticed that millions of people who knew nothing of either lived without despair. In chapter XII he writes:

... I was saved only by the fact that I was able to tear myself from my exclusiveness and to see the real life of the plain working people, and to understand that it alone is real life.

He concluded that their lives had meaning because of their faith, and he spent the rest of his life working out a religion of his own, stripped of most of the Church's doctrine. A naturalist reads the same scene differently: what sustained the peasants, on this reading, was the work itself, their families and their communities, which are available without any supernatural belief. Tolstoy's experience does not settle which reading is right. It does show that the question can be lived through and answered, not only argued about.

The core of it: Tolstoy found meaning by turning from abstract questions to how ordinary people actually lived. Whether what he found there was faith or the life itself is the dispute between supernaturalist and naturalist views.

Is death bad for the one who dies?

Now the second question. Epicurus, whose account of pleasure you read in Lesson 17, argued in the same letter to Menoeceus that death should not be feared. This is Yonge's translation.

Accustom yourself also to think death a matter with which we are not at all concerned, since all good and all evil is in sensation, and since death is only the privation of sensation. ... Therefore, the most formidable of all evils, death, is nothing to us, since, when we exist, death is not present to us; and when death is present, then we have no existence.

(1) Something is bad for you only if you can experience it.
(2) When death is present, you do not exist, and so experience nothing.
(3) So death is not bad for you.
(4) So fearing death is irrational.

The Roman poet Lucretius, an Epicurean, added a second argument in his poem On the Nature of Things. This is William Ellery Leonard's verse translation of 1916.

Look back: Nothing to us was all fore-passed eld
Of time the eternal, ere we had a birth.
And Nature holds this like a mirror up
Of time-to-be when we are dead and gone.
And what is there so horrible appears?
Now what is there so sad about it all?
Is't not serener far than any sleep?

"Eld" means past ages. This is the symmetry argument: you were not troubled by the eternity before you were born, and the eternity after you die is its mirror image, so it should not trouble you either.

The deprivation reply. The strongest answer to Epicurus, developed by Thomas Nagel in a 1970 essay called "Death", denies premise (1). You can be harmed without experiencing the harm: a person whose friends mock him behind his back, or whose life's work is secretly destroyed, is worse off even if he never finds out. Death is bad, on this account, because it deprives you of the goods you would have had if you had lived on: the next forty years of friendship, work and discovery. It is not an experience; it is a loss. The Stanford Encyclopedia's entry on death sets out this deprivationist defence and the puzzle it raises about when exactly death harms you.

The reply to Lucretius. Deprivationists answer the symmetry argument by pointing to an asymmetry. Death takes away a future you would otherwise have had. The time before your birth took away nothing, since, Nagel argued, you could not have been born much earlier and still been you, with your origins and history. Critics reply that the asymmetry is in our attitudes, not in the facts: we care more about future goods than past ones, and Lucretius's point is that this bias is irrational.

Would it be better never to die?

If death is bad because it deprives us of goods, would living for ever be better? In 1973 Bernard Williams argued no, in an essay called "The Makropulos Case", named after a play whose heroine has lived for over three hundred years and grown utterly bored. Williams distinguished conditional desires, which we have only on condition that we go on living, such as the desire for food, from categorical desires, which give us a reason to go on living at all: to finish the novel, to see the children grown, to understand something. An endless life, he argued, would sooner or later exhaust the categorical desires of a person who remained recognisably the same person, and then there would be nothing left to live for but tedium. His critics reply that some goods, such as friendship, music or walking by the sea, do not wear out with repetition, and that a person might keep changing in ways that renew what they want.

Williams's argument connects the two halves of this lesson. If he is right, death is not what destroys the meaning of a life, as Tolstoy feared; it may be part of what gives a life its shape.

Common misconceptions

• "Epicurus thought that since death is nothing, life is nothing." The opposite: he argued that understanding death correctly "makes the mortality of life pleasant to us", by removing the longing for immortality.
• "If there is no cosmic purpose, life has no meaning." That is the supernaturalist view, and it is one view among several. Subjectivists and objectivists both hold that meaning can arise within a life.
• "A meaningful life is just a happy one." The experience machine and the contented Sisyphus both separate the two. Most philosophers who write on the subject treat meaning as a distinct kind of value.
• "The deprivation account says the dead suffer." It says nothing of the kind. Death harms by leaving you worse off than you would otherwise have been, not by being felt.
• "Tolstoy's recovery proves that meaning requires religion." It shows how one person answered the question. Whether his answer was the only one available is exactly what the dispute is about.

Pulling it together

• Tolstoy, famous and fortunate, found his life at a standstill and asked whether it had any meaning that inevitable death does not destroy.
• Supernaturalists locate meaning in God or an immortal soul; subjectivists in caring about what one does; objectivists in engaging with what is truly worthwhile; Wolf in the meeting of the two.
• Taylor's Sisyphus, given a desire to roll stones, separates the views: happy on any account, meaningful only for the subjectivist.
• Epicurus argued that death is nothing to us because when it is present we are not; Lucretius added that the time after death mirrors the time before birth.
• The deprivation account replies that death harms by taking away goods we would have had, and finds an asymmetry between prenatal and posthumous time.
• Williams argued that living for ever would be intolerable once categorical desires ran out, which suggests mortality may give a life its shape.

Sources

1. Tolstoy, L. (1921). A confession (L. Maude & A. Maude, Trans.), chapters IV, V and XII. Wikisource. wikisource.org
2. Diogenes Laertius. (2018). The lives and opinions of eminent philosophers (C. D. Yonge, Trans.; first published 1853), Book X, Epicurus' letter to Menoeceus (Project Gutenberg eBook No. 57342). Project Gutenberg. gutenberg.org
3. Lucretius. (1997). On the nature of things (W. E. Leonard, Trans.; first published 1916), Book III (Project Gutenberg eBook No. 785). Project Gutenberg. gutenberg.org
4. Metz, T. (2026). The meaning of life. In E. N. Zalta & U. Nodelman (Eds.), Stanford Encyclopedia of Philosophy. plato.stanford.edu
5. Luper, S. (2026). Death. In E. N. Zalta & U. Nodelman (Eds.), Stanford Encyclopedia of Philosophy. plato.stanford.edu

An assignment, 800 words

Here is the kind of question a first philosophy paper is set: "Nozick's experience machine shows that hedonism about wellbeing is false." Discuss, in about 800 words. You met the machine in Lesson 17. This lesson takes that question from the blank page to a finished paper, prints the paper in full, marks it against a rubric, and then shows what the same paper would have to become if its central reply failed.

First, what the task is not. It is not a report on what philosophers have said about the machine, and it is not a statement of your feelings about pleasure. It is an argument. You take one position, give reasons for it that a reader who does not already agree could accept, face the strongest objection to it, and answer. James Pryor's widely used guidelines for philosophy students make the same point: treat your reader as someone who does not yet accept your position and whom you are trying to persuade, so that nothing can be left as obvious. Everything in the procedure follows from that.

The procedure, in seven steps

1. Pin down the question. Say in one sentence what would count as answering it. Here: does the machine give good reason to think hedonism false?
2. Reconstruct the argument fairly, in numbered premises, as its best defender would put it. You cannot criticise what you have not first stated well.
3. Choose a thesis: one clear claim that a reasonable reader could deny. "The experience machine raises interesting questions" is not a thesis, because nobody denies it.
4. Plan the shape: introduction with thesis, reconstruction, your argument, the strongest objection, your reply, conclusion. For 800 words that is six paragraphs.
5. Write your argument with every premise stated. The premise you leave unsaid is the one a marker will attack, as Lesson 1 warned.
6. Find the strongest objection, not the easiest. Steelman it: state it as its best defender would. A paper that defeats a weak objection proves nothing.
7. Reply, and concede what you must. A reply can show the objection fails, show it misses your particular argument, or accept part of it and narrow your thesis. Then conclude by saying exactly what you have shown and what remains open.

Remember: A philosophy paper is one argument, done carefully: a thesis, reasons for it, the best case against it, and an answer.

The steps worked on the question

Steps 1 and 2. Hedonism about wellbeing says that a life goes well for the person living it to the extent that it contains pleasure and lacks pain. Nozick's argument, reconstructed: if hedonism is true, a life on the machine is at least as good for you as any life off it; a life on the machine is not as good for you as a real life of achievement and relationships; so hedonism is false. Valid. Everything turns on the second premise.

Step 3. Two theses are available: that the machine does refute hedonism, or that it does not. Either can earn full marks. The model paper below argues the first, and the last section of this lesson shows how the second would be argued.

Step 6. The obvious objection, that some people would plug in, is weak: the argument never claimed everyone would refuse. The strongest objection is empirical and specific. In a 2010 paper, the philosopher Felipe De Brigard reported a study in which people were asked to imagine they were already on a machine and could leave. Their choices shifted with what the "real" life was said to be, which he attributed to status quo bias. That objection attacks the evidence for the key premise, so it is the one the model paper takes on.

The model paper

Does the Experience Machine Refute Hedonism?

Hedonism about wellbeing is the view that a person's life goes well for them to the extent that it contains pleasure and lacks pain. In Anarchy, State, and Utopia, Robert Nozick asked us to imagine an experience machine that could give us any experiences we wanted for the rest of our lives, while we floated in a tank believing it all to be real. Most people say they would not plug in. In this paper I argue that the experience machine gives us good reason to reject hedonism. I first set out Nozick's argument, then defend its key premise against the strongest objection to it, which comes from experimental work on status quo bias.

The argument has three steps. (1) If hedonism is true, a life on the machine would be at least as good for the person living it as any life off it, since the machine can supply at least as much pleasure. (2) A life on the machine would not be as good for the person as a comparable life of real achievement and real relationships. (3) So hedonism is false. The argument is valid, so everything depends on premise (2). Nozick supports it by appeal to our reactions: we want to actually do things and to be a certain kind of person, not merely to have the experience of doing and being. If careful people judge that plugging in would be a mistake for their own sake, that is evidence that premise (2) is true.

I think premise (2) is true, and the best way to see why is to consider someone else's life rather than your own. Imagine two people, Asha and Ben, whose experiences are identical from the inside. Each feels that they have written a novel that moved thousands of readers, and that they have a friend who would do anything for them. Asha really did write the novel and really has the friend. Ben is on the machine, and his novel and his friend are illusions. If hedonism is true, Asha's and Ben's lives went equally well for them, since their pleasures were the same. But almost everyone who considers the case judges that Asha's life went better for Asha than Ben's went for Ben. What Ben lacked was not a feeling. He lacked the things his feelings were about. That is exactly what premise (2) claims, and the comparison supports it without asking anyone to imagine leaving their own life.

The strongest objection comes from Felipe De Brigard, who argued that our reluctance to plug in reflects status quo bias, the tendency to prefer whatever situation we are already in, rather than a judgement about wellbeing. In a 2010 study he asked 72 undergraduates to imagine that they were already on an experience machine and asked whether they would disconnect and return to real life. When told nothing about that real life, 54 percent chose to disconnect; when told that their real life was as a prisoner in a maximum security prison, only 13 percent did. If people's choices change when the starting point changes, the objection runs, then the original reactions are evidence of our attachment to the familiar, not of what is good for us, and premise (2) loses its support.

This objection is serious against the version of the argument that relies on each reader's choice about their own life, and I concede that such choices are distorted by where we start. But my defence of premise (2) did not rely on that choice. The comparison between Asha and Ben asks no one to leave or enter anything, so there is no status quo for a bias to favour: we are judging two lives from outside. Nor does De Brigard's result show that people prefer machine lives as such. The prisoners' choice is easily explained by how bad the real life on offer was, which even someone who rejects hedonism expects to matter. So the evidence of bias weakens one route to premise (2) but leaves the comparative route standing.

I have argued that the experience machine gives good reason to reject hedonism about wellbeing. The argument turns on the claim that a life of illusory achievement is worse for the person than a real one with the same experiences. De Brigard's work shows that our choices about our own lives are unreliable evidence for that claim, but judgements about the lives of others avoid the bias he identified. If those judgements were also shown to be distorted, the argument would need another defence. As it stands, hedonism has not answered it.

That paper is about 800 words. Notice its moves. The thesis is in the first paragraph. The argument under discussion is set out in numbered premises and its weak point named. The writer's own contribution, the Asha and Ben comparison, is in the third paragraph. The objection is stated with its evidence, in a form its author would accept. The reply concedes something real and then shows the objection misses the particular argument given. The conclusion says what would change the verdict.

The rubric

Marking the model. Thesis 4: one arguable claim in the first paragraph. Reconstruction 4: numbered, fair, key premise named. Argument 4: the Asha and Ben comparison is the writer's own and supports the contested premise directly. Objection 4: De Brigard's is the strongest objection, and it is stated with its evidence. Reply 4: it concedes that first-person choices are biased and then shows why the comparison escapes the bias. Clarity 4. Accuracy 3: the study is reported accurately, but a full paper would give its reference in a bibliography rather than in passing. Total 27 out of 28.

Worth holding on to: The marks go to the objection and the reply as much as to the argument. A paper that faces the best case against it and survives is worth more than one that never looks.

One input changed: when the objection wins

Now change one fact. Suppose a later study found that when people judged the lives of Asha and Ben, their answers also flipped depending on which life was described as the normal one in their society. Then the reply in paragraph five would fail, because the comparative route to premise (2) would be distorted in the same way as the first-person route.

A dishonest paper would keep its thesis and hope the marker had not read the new study. An honest paper changes its thesis to what can still be defended. The revised thesis might read: The experience machine does not by itself refute hedonism, because every reaction to it that we can test seems to track where we start rather than what is good for us; it shows, at most, that hedonism conflicts with our untutored judgements. The objection section would now come from the other side: a defender of Nozick arguing that some judgements are not biased, perhaps judgements made about a stranger's life with no society's normal named at all. And the conclusion would say what evidence would restore the original argument.

That revised paper argues the opposite thesis, and on the rubric it could score exactly as highly as the first. This course takes no side on whether hedonism is true. What earns the marks is the quality of the argument, not the side it lands on.

Bottom line: When the strongest objection wins, narrow or reverse your thesis. Changing your mind for a good reason is the whole point of the exercise.

Common misconceptions

• "A good paper covers every view on the topic." At 800 words, one argument done properly beats a survey done thinly. Mention other views only where they bear on your argument.
• "Raising an objection weakens my paper." It is the part markers look for first. Answering the strongest objection is what shows your thesis is worth believing.
• "Lots of quotations show I understand." Put arguments in your own words and quote only where the exact wording matters to your point.
• "In philosophy any opinion is as good as any other." Positions are judged by the reasons given for them. An opinion with no reasons scores at the bottom of the argument row, however interesting.
• "I have to prove my thesis beyond all doubt." You have to show it is the best-supported position given the strongest objection, and say what would change your mind.

The takeaway

• A philosophy paper argues for one thesis a reasonable reader could deny, to a reader who does not already agree.
• Reconstruct the argument you discuss fairly, in numbered premises, and name its key premise.
• Your own contribution is an argument, like the Asha and Ben comparison, that bears directly on the contested premise.
• Choose the strongest objection, state it as its defenders would, and reply by meeting it or conceding part of it.
• Conclude with what you have shown and what would change the verdict.
• If the objection wins, revise the thesis. A well-argued paper on either side of a contested question can earn full marks.

Sources

1. Pryor, J. (n.d.). Guidelines on writing a philosophy paper. jimpryor.net
2. Crisp, R. (2026). Well-being. In E. N. Zalta & U. Nodelman (Eds.), Stanford Encyclopedia of Philosophy. plato.stanford.edu
3. Wikipedia. (2026). Experience machine, sections on status quo bias and De Brigard's study. en.wikipedia.org
4. De Brigard, F. (2010). If you like it, does it matter if it's real? Philosophical Psychology, 23(1), 43-57.
5. Nozick, R. (1974). Anarchy, state, and utopia (pp. 42-45). Basic Books.