The Problem of Other Minds

You have never encountered another person's pain. You have encountered faces, sounds, movements, and words, and you have encountered your own pain, which is the only one there has ever been any question of your encountering.

The previous article granted that your access to your own case is unusual even after the limits are taken into account. This one asks what follows for everyone else, and the answer is more awkward than it looks, because the obvious reply turns out to be a bad argument and its replacements are all contested.

Two problems under one name

Two versions have to be kept apart, because a solution to one is not a solution to the other.

The epistemological version asks how the belief is justified. It grants that the claim makes perfectly good sense and asks what the evidence for it is.

The conceptual version is more radical. It asks how anyone even manages to form the thought that someone else is in pain. You learned the word pain in connection with your own case, and applying it to a creature whose case you cannot examine requires that the concept extend beyond anything you have observed. Whether it can, and what makes it apply when it does, is a question that arises before any question about evidence.

Solipsism, the position that only one's own mind exists, is often introduced here and is the least interesting thing in the neighbourhood. Almost nobody holds it and it is not really the worry. The worry is that the ordinary belief may be far less well supported than its universal and unshakeable character suggests, and a belief nobody can justify is uncomfortable whether or not it is true.

What there is to go on

This is the whole of the data. The category is broader than it sounds, since it includes speech, physiological similarity, brain imaging, and the whole of a life's conduct rather than a twitch in isolation. It is also, on any view, indirect. Even a complete map of someone's brain activity is a map of their brain, and the question is what licenses the move from that to a claim about how things are for them.

The classic argument

John Stuart Mill gave the argument its standard nineteenth-century statement in his examination of Hamilton's philosophy in 1865. His reasoning was that other people have bodies like his own, which in his own case he knows to be the condition of feeling, and that they display the behaviour that in his own case he knows to be produced by feeling. So the same cause is presumably at work in them.

The argument has an obvious appeal and a well-known defect.

The defect is the sample. Every ordinary inductive generalisation rests on many observed instances, and its strength comes from the number and variety of them. Here the observed instances number exactly one, and it can never be more than one, since the only case anyone can inspect from the inside is their own. Generalising from a single case to every human being who has ever lived, on a matter about which the single case cannot be checked against any other, is not a form of reasoning anyone would accept in another context.

There is a second defect that is less often noticed and is at least as serious. The argument, if it worked, would make your belief in other minds a tentative empirical conclusion of modest strength. Nobody's belief in other minds is like that. People do not hold it with the confidence appropriate to a single-case induction, and no evidence would shake it. A proposed justification that would license far less confidence than everyone actually has is not the justification everyone is actually operating with.

Better versions

The usual repair treats the inference as an explanatory one rather than an analogical one. The hypothesis that other people have inner lives like yours explains an enormous amount: the intricacy of what they say, their responsiveness to circumstances you cannot see them tracking, the way their behaviour hangs together over decades, the fact that they produce philosophy about consciousness. No rival hypothesis explains any of it, and the alternative is not a competing theory but the absence of one.

That is a real improvement, and inference to the best explanation does not need a large sample in the way enumerative induction does. It carries its own difficulties. An explanation is best relative to the alternatives considered, and the argument gives no reason to think we have canvassed the space. It also has trouble with phenomenal consciousness in particular: what the behaviour needs explaining by is the information processing, and whether anything is felt looks explanatorily idle in a way that is exactly the difficulty the second article identified.

The conceptual attack

Ludwig Wittgenstein approached it from the other end, and his treatment in the Philosophical Investigations of 1953 changed the shape of the discussion.

Suppose everyone has a box with something in it that they call a beetle, and nobody can look into anyone else's box. The word beetle has a use in the shared language and that use cannot depend on what is in the boxes, since nobody has access to those. The contents could differ from person to person, could change constantly, or could be absent, without any difference to how the word works. Whatever gives the word its meaning, it is not the private object.

The intended application is to sensation words. If pain named a private item known only to its owner, the word could not have the public meaning it demonstrably has. So the meaning must be fixed by something available to everyone, and that is the relation between the word and the circumstances and behaviour in which it is used and taught.

The distinction between a criterion and a mere sign is what the position turns on. A rash may be a reliable sign of a disease, discovered by investigation, and the connection between them is contingent and could have been otherwise. A criterion is not like that. On this view the connection between pain behaviour and pain is not a correlation someone discovered but part of what makes the concept the concept it is, so that a creature writhing and crying out is not merely evidence for pain in the way smoke is evidence of fire.

If that is right, the epistemological problem is transformed rather than solved. There is no gap to be bridged by inference, because behaviour was never merely evidence for something else lying behind it.

The position is not comfortable and has been resisted from the start. Pretence is possible, which means the criterion can be satisfied without the state; defenders reply that a criterion is defeasible and still not merely inductive, which critics find hard to distinguish from ordinary good evidence. And the account can look as though it denies that anything is felt at all, which its defenders deny and which the wording of some of their arguments invites.

What this tier has done

Six articles have set up the field rather than settling anything in it, which is the job at this level.

A problem was stated as a conflict between claims that cannot all be held. The mental was divided into a processing side that looks amenable to physical explanation and a felt side that does not, and the vocabulary for that division was fixed. The states themselves were sorted by content and by activity. Perception was shown to resist the ordinary view of it under pressure from ordinary cases. The instrument that all of this relies on was examined and found to be reliable in some respects and not others. And the extension of mental description beyond one's own case was shown to rest on foundations nobody has been able to make secure.

The intermediate tier takes the positions themselves: the versions of dualism and what each says about causation, the identity theory and its collapse, functionalism and the objections that shaped it, and the accounts of content and perception that the arguments here were pointing toward.