John McDowell asks an interesting question in his "What is the Content of an Intention in Action?" If Anscombe is right that "I do what happens," then how can she also say that "my knowledge is independent of what actually happens" so that I can know that I am writing "I am a fool" on a chalkboard even if my eyes are shut? (See here for the relevant passage from Anscombe.)
Anscombe's point is that if someone asks what you are doing, you don't have to stop and look in order to answer the question. You know what you are doing non-observationally. But what if the board has been cleaned with something that prevents the chalk from making a mark? Then how could I know I was writing anything, rather than merely trying to write something, if my eyes are shut?
Anscombe says that in such cases the mistake lies in the performance, not the judgment. McDowell insists that: "it is surely wrong to suppose Anscombe's claim to be writing 'I am a fool' on the blackboard can express knowledge if those words are not getting written on the blackboard."
She might be said to have knowledge of what she takes herself to be doing, but if we say this then it sounds as though we might always have to actually check in order to be sure that we really are doing what we think we are doing (in other cases, as well as the blindfold-type case). And this is both unfashionably dualistic and absurd. We know what we are doing not only better than others know but in a different way. In fact, we know with such certainty that it almost makes no sense to say that we know it. But the chalk example shows that we can be wrong, and so knowledge claims do make sense. There is something that could be true or false, something to know or fail to know.
The problem with what Anscombe says, it seems to me, lies in her assertion that her knowledge would be the same in the case where she is writing and in the case where, unbeknown to her, something has gone wrong in the performance of her action. Her belief or certainty might be the same, but in one case she knows only what she means to do, not what she is doing. So it's odd to say that her knowledge is the same in both cases. Should we say that her mental state is the same, but that in one case it is not knowledge? That doesn't sound quite right either.
It's important to stay on Anscombe's side, though, otherwise we seem to be left saying that we don't know what we are doing most of the time (unless we neurotically look and see), just as we might in other philosophical debates say that it is not reality that horrifies some people but only putative reality, or that we don't really hear motorcycles but only sounds that we interpret as motorcycles. Sometimes it is hard to know what one is hearing or doing, but excessive focus on these exceptional cases make bad metaphysics.