McDowell’s argument goes deep and, in the end, concerns the nature of experience. But it is helpful to work up to that from his treatment of other minds. That’s what he does, too.
The problem of other minds starts from the assumption that whilst one can experience another’s behaviour, one cannot experience their mental states. (There is a trivial truth in this claim but it can also be taken to imply a significant epistemological point.) So the challenge is then to answer this question:
How can what we can experience (behaviour) yield knowledge of what we cannot experience (other minds)?
(In the philosophy of mind, this set up is usually accepted and the problem answered in its own terms. But McDowell aims to challenge the terms of the debate, to dissolve the problem before it starts. So I will ignore answers such as ‘the argument from analogy’ or ‘theory theory’.)
Behaviourism
One answer is to say that mental concepts are merely short-hands for behaviour. If so, a judgement about behaviour can support a logical deduction to judgement about mental states: behaviourism. BUT behaviourism is implausible (because of pretence / shamming).
Traditional view of criteria
So the erring Wittgensteinians put forward a weaker version of behaviourism. Mental concepts are not quite short-hands for behaviour but when those concepts are defined they include links to forms of behaviour. That is how little Ludwig learns the concept of pain, for example, by looking at others’ behaviour. So such behaviour is – a priori and by definition – good evidence for mental states. But – learning from the objection to behaviourism – such behavioural ‘criteria’ can be defeated / undermined. The behavioural criteria can be satisfied (that is, the right kind of behaviour can be present) but, for some reason or other, the appropriate mental state is not present (it is merely acting, eg.).
McDowell’s objection to the traditional view of criteria
It cannot give us knowledge of other minds. There is NEVER knowledge, even when there is no acting. Why not? Because if the criteria can be satisfied in the absence of the mental state, then even if all goes well and they are not defeated, still that is merely a matter of luck. All I am aware of is that the criteria are satisfied. But whether the other person really is in the mental state for which her behaviour is a criterion is beyond my ken: beyond anything I can experience.
McDowell’s alternative
The above objection suggests that the only way experience of others’ behaviour can yield knowledge of their mental states is if criteria are NOT defeasible. If they are satisfied, the other person must be in the right mental state.
There’s another consequence of that. Criteria are not types of behaviour at least when specified in non-mind-presupposing terms. If they were, and if they were not defeasible, then that would be behaviourism and we’ve already rejected that.
Two responses to that consequence:
1) a criterion is a criterion only in a context; it is a one-off arrangement of the elements. (This is like Williams’ contextualism but applied to the specific case of justification of knowledge of others’ mental states from behaviour.)
2) But it can be seen as a type as long as it isn’t given in non-mind-presupposing terms. All pain behaviour has the common feature that it expresses pain.
So what is going on? There’s a positive story and a concession.
The positive story: the original assumption that judgements of behaviour are more secure than judgements of other minds is wrong. Once one has been inducted into a language (a language of other mental states), one can see in others’ behaviour, the expression of their mental states. One categorises the behaviour in terms of the mental states.
This positive story is helped by a historical diagnosis. Philosophers have only ignored the expressive possibility of behaviour because of an objectifying attitude to the body which is itself the result of Cartesianism because Descartes splits the mind off from a merely objective res extensa.
The concession: we do not always get things right. But when we do not that is not because the criteria are satisfied but defeated. But rather, we think the criteria are satisfied though they are not.
But this means that, as long things do not go wrong, what we experience is enough to reach out to (the expression of) other minds. The mistake is to think that there is something common between the cases of things going well or badly: the same behaviour. But it isn’t. In one case the behaviour expresses the mental state and in the other it is not. Thinking that there is means that we never have knowledge of other minds. If there isn’t then we sometimes have knowledge of other minds (when all goes well).
The link to the argument from illusion
The problem of other minds starts by taking behaviour which stops short of mental states to be all that can be experienced. Such behaviour is what is common between cases of things going well or badly.
Scepticism about the external world draws a similar conclusion from cases of illusions. Since we can confuse the experience of an illusion of a dagger and the experience of a real dagger, all that there can be to experience is what is common to both: the highest common factor. But if so, experience can never yield knowledge of the world since, if there is a dagger there, that’s merely a matter of luck.
Just as behaviour can be mind-expressing, so experience can be dagger-involving.