I just read ‘A counterexample to the contrastive account of knowledge‘ by Jason Rourke, at the suggestion of John Danaher. I’ll paraphrase what he says before explaining why I’m not convinced. I don’t actually know much more about the topic, so maybe take my interpretation of a single paper with a grain of salt. Which is not to imply that I will tell you every time I don’t know much about a topic.

Traditionally ‘knowing’ has been thought of as a function of two things: the person who does the knowing, and the thing that they know. The ‘Contrastive Account of Knowledge’ (CAK) says that it’s really a function of three things – the knower, the thing they know, and the other possibilities that they have excluded.

For instance I know it is Monday if we take the other alternatives to be ‘that it is Tuesday and my computer is accurate on this subject’, etc. I have excluded all those possibilities just now by looking at my computer. However if alternatives such as that of it being Tuesday and my memory and computer saying it is Monday are under consideration, then I don’t know that it’s Monday. Whether I have the information to say P is true depends on what’s included in not-P.

So it seems to me CAK would be correct if there were no inherent set of alternatives to any given proposition, or if we often mean to claim that only some of these alternatives have been excluded when we say something is known. It would be wrong if knowing X didn’t rely on any consideration of the mutually exclusive alternatives, and unimportant if there were a single set of alternatives determined by the proposition whose truth is known, which is what people always mean to consider.

Rourke seems to be arguing that CAK is not like what we usually mean by knowledge. He seems to be doing this by claiming that knowing things need not involve consideration of the alternatives. He gives this example:

The Claret Case. Imagine that Holmes and Watson are investigating a crime that occurred during a meeting attended by Lestrade, Hopkins, LeVillard, and no others. The question Who drank claret? is under discussion. Watson announces ‘‘Holmes knows that Lestrade drank claret.’’ Given the question under discussion and the facts described, the alternative propositions that partially constitute the knowledge relation are Hopkins drank claret and LeVillard drank claret.

He then argues basically that Holmes can know that Lestrade drank claret without knowing that Hopkins and LeVillard didn’t drink claret, since all their claret drinking was independent. He thinks this contradicts CAK because he claims, using CAK,

 The logical form of Watson’s announcement, then, is Holmes knows that Lestrade drank claret rather than Hopkins drank claret or LeVillard drank claret.

Whereas we want to say that Holmes does know Lestrade drank claret, if for instance he sees Lestrade drinking claret, and he need not necessarily know anything about what Hopkins and LeVillard were up to.

Which prompts the question why Rourke thinks these other guys’ drinking are the alternatives to Lestrade drinking in the knowledge relation. The obvious real alternative to exclude is that Lestrade didn’t drink.

Rourke gets to something like this as a counterargument, and argues against it. He says that if ‘who drank claret?’ is interpreted as ‘work out whether or not each person drank claret’ then it can be divided up in this way into ‘Lestrade drank claret’ vs. ‘Lestrade did not drink claret’ combined with ‘Hopkins drank claret’ vs ‘Hopkins did not drink claret’ etc. However if the question is meant as something like ‘who is a single person who drank claret?’, then ‘knowing’ the answer to this question doesn’t require excluding all the alternative answers to this question, some of which may be true.

As far as I can tell, this seems troublesome because he supposes that the alternatives to the purported knowledge must be the various other possible answers to the question, if what you supposedly know is ‘the answer to the question’. The alternative answers to such a question can only be positive reports of different people drinking, or that nobody drank. The question doesn’t ask for any mentions of who didn’t drink. So what can we contrast ‘Lestrade drank’ with, if not ‘Lestrade didn’t drink’?

But why suppose that the alternatives must be  the other answers to the question? If ‘knowing who drank claret’ just means knowing that a certain answer to that question is true rather than false for instance, there seems no problem. For instance perhaps ‘I know who drank’ means that I know ‘Lestrade did’ is one answer to the question. This can happily be contrasted with ‘Lestrade did’ not being an answer for instance. Why not suppose ‘I know who drank claret’ is shorthand for something like that?

It seems that at least for any specific state of the world, it’s possible to think of knowing it in terms of excluding the alternatives. It also seems answering more difficutly worded questions such as the one above must still be based on knowledge about straightforward states of the world. So how could knowledge of at least one person who drank for instance not be understandable in terms of excluding alternatives?