Philosophy 159: Dretske and Cohen

Dretske and Cohen

Dretske is a fan of the Relevant Alternatives Theory. He thinks it does a good job of explaining his Gadwell duck/grebe case. In that case, you see a bird that has all the markings of a Gadwell duck--and in fact it is a Gadwell duck. Do you know that it is? Dretske says that depends on whether there are relevant alternatives to its being a Gadwell duck. For example, one alternative hypothesis might be that there is a species of grebe around which has the same markings. (A grebe is a kind of bird that looks like a duck but isn't.) If that is a relevant alternative, Dretske thinks, then you don't know that the bird you see is a Gadwell duck. But if it's not a relevant alternative, then you can know.

Dretske gives a similar kind of case in another paper, "Epistemic Operators." (This paper is on reserve in Robbins Library.) In that case, you're at the zoo, and in the pen in front of you is a striped horse-like animal. The sign on the pen says "Zebra." Do you know that the animal is a zebra? (Assume that in fact it is a zebra.) Dretske says: Well, what about the possibility that it's just a mule painted to look like a zebra? Do you really know that the animal is not a cleverly-disguised mule? You may have some reason to believe that it's not a cleverly-disguised mule. (Zoos don't typically try to fool people like that.) But your evidence doesn't seem to be good enough to know that it's not a cleverly-disguised mule. You haven't made any special tests, or anything like that. So Dretske thinks you don't know it. But he still wants to say that, so long as the mule hypothesis is not a relevant alternative, you can know that the animal in the pen is a zebra.

What Makes an Alternative Relevant?

Now what determines whether some possibility is a relevant alternative or not? Does the subject have to know about the grebe possibility for it to be relevant? Or is it enough if there is a suspicious ornithologist poking around in the neighborhood, who thinks there are grebes like that in the area? What if there is no ornithologist but some people think there is? What if there really are look-alike grebes in the area but no one knows that there are? In which of these cases is the grebe possibility a relevant alternative?

There are different kinds of stories one can tell about this.

On one sort of account, relevance depends solely on facts about the knower's environment. For instance, we could say that whether the grebe possibility is relevant just depends on whether there are grebes in the neighborhood.
Or we could say that the grebe possibility is relevant just in case the knower believes there to be grebes around, or has evidence that there are grebes around, or something like that. This is what Austin would say makes a possibility relevant.
Or we could say that whether some alternative counts as relevant doesn't depend so much on facts about the knower; rather it depends on facts about the conversational setting in which knowledge is being ascribed, or in which the knowledge-claim is being assessed. For instance, if we're talking about whether Julius Caesar knew he had hands, in one context we might want to count hypotheses like BIVs and evil demons as irrelevant, and in another context we might want to count them as relevant. This depends upon us, and our interests and purposes, not upon facts about Caesar. This is the kind of story the contextualist tells.

It is useful to keep these different stories distinct, though discussions of Relevant Alternatives Theories sometimes run them together. (Dretske is someone who tends to run them together. Cohen in his paper "Skepticism, Relevance and Relativity" tries to distinguish the different stories.)

Relevant Alternatives Theory and Closure

Recall the Closure Principles about knowledge that we discussed earlier. For example:

If you know that P, and you know that P logically entails not-Q, then you have to know (or at least be able to know) not-Q too.

At first glance it looks like the Relevant Alternatives Theorist will be rejecting Closure. Closure tells us that you have to know to be false everything which is (or which you know to be) an alternative to P; whereas the Relevant Alternatives Theorist says you only have to know to be false those alternatives which are relevant.

And sure enough, Dretske's version of the Relevant Alternatives Theory does reject Closure. This comes out more in Dretske's paper "Epistemic Operators."

For example, Dretske claims that you can know that the animal in the pen is a zebra on the basis of evidence which does not enable you to know that the animal isn't a cleverly-disguised mule. So here we have a counter-example to Closure:

You know that the animal is a zebra.
The claim that the animal is a zebra is incompatible with its being a cleverly-disguised mule.
Yet you don't know the claim that the animal is a cleverly-disguised mule to be false.

Dretske thinks that knowledge is closed under some known entailments. For example, he thinks that if you know that P&Q, then you know that P. (P&Q logically entails that P.) He just thinks that Closure fails when it's a matter of knowing P, and knowing that P entails not-Q, but Q is not at the moment a relevant alternative to P. In such a case, Dretske thinks, you can know P even if you're not in a position to know not-Q. You can know it's a zebra in the pen even if you're not in a position to know it's not a clerverly-disguised mule. But if the possibility of its being a cleverly-disguised mule were to become relevant, then since you can't rule it out you would no longer count as knowing that it's a zebra in the pen.

However, matters are not so straightforward. After all, it's somewhat odd to deny Closure. It's odd to say that you could know that P, and know that P entails not-Q, but not be able to put these together and draw the conclusion that not-Q.

If you hold a Contextualist version of the Relevant Alternatives Theory, then you might be able to retain some kind of Closure Principle.

According to the Contextualist, Norm can say "Caesar knew he had hands" and Skepty can say "Caesar did not know he had hands," and they need not be disagreeing. They might both be right. It's just that what they mean by "X knew he had hands" is different, since the sets of relevant alternatives are different.

Similarly, when you say "I know I have hands" and the skeptic says "You don't know you're not dreaming," the Contextualist can say you're speaking in different contexts, with different sets of relevant alternatives. It's as if Tom were to say "I'm Tom" and Jerry were to say "I'm not Tom." There's no conflict there, just a difference in context.

With this in mind, the Contextualist can go on to insist that Closure holds within any single context. In any given single context, you either know that you have hands and that you're not dreaming, or you don't know that you're not dreaming and you don't know that you have hands, either. If dreaming is a relevant alternative, then you don't know either of these things. But if dreaming is not relevant, then you can know that you have hands, and you can also know that you're not merely dreaming.

So it looks like one can be a Contextualist and still accept the Closure Principle, after all.

However, there is a small snag. If we accept Closure, then we say that (in any single context) knowing that P requires you to know to be false all the alternatives to P. However, we want to conjoin this with a Relevant Alternatives Theory, so we say that (again, in that single context) knowing that P doesn't require you to eliminate or rule out all the alternatives to P. You know the alternatives to be false but you haven't ruled them all out. This shows us that we should distinguish between merely knowing an alternative to be false and having ruled that alternative out.

Knowledge that you're not dreaming is easy to come by, on the view we're now considering. So long as we're evaluating what you know in a non-skeptical context, a context where the possibility of dreaming is not among the relevant alternatives, then you count as knowing you have a hand, that you're standing up, etc., and you also count as knowing the consequences of these things, e.g., that you're not merely dreaming. However, on nearly everybody's view it's impossible to ever "rule out" the possibility that you're dreaming. This is why Relevant Alternatives Theories seem so attractive. They say you can have perceptual knowledge without having to rule such skeptical possibilities out. So knowledge that you're not dreaming is easy to come by; but the ability to rule out the possibility that you're dreaming is hard to come by. We would need to come up with some account of "ruling out" which lets us say that.

Here's another kind of issue to think about. Suppose you're walking down the street talking to your friend on your cell phone and your friend says "Do you know whether it's raining in Harvard Square?" Since you're walking in Harvard Square you say "Yes." Now a skeptic has overheard your conversation, he butts in and says, "Hey remember the Matrix and so on. Now do you know whether it's raining in Harvard Square?" Now that the Matrix is a relevant alternative, you no longer count as knowing things about your environment. So you say "No, I don't know." The skeptic says, "So before you knew but now you don't? Isn't that a strange theory?"

The best thing for the Contextualist to say about these kinds of cases is that the skeptical context doesn't just change the meaning of the present-tense claims "I know that P"; it changes the meanings of past-tense claims, like "I knew that P," too. So in the skeptical context, "I know the weather in Harvard Square" is false and so too is "I knew the weather in Harvard Square a minute ago." But in a non-skeptical context both of those claims are true. That's because the claims say different things in the different contexts. Just as the claim "Mighty Mouse is big" says something different in a context where the comparison class is other mice than it says in a context where the comparison class is all land animals.

Does the Relevant Alternatives Theorist Beg the Question Against the Skeptic?

How satisfactory a reply to the skeptic does a Relevant Alternatives Theory really give us? The skeptic will just say, "You're begging the question against me, in claiming that my skeptical possibilities are irrelevant and don't need to be ruled out, if you're to count as having perceptual knowledge."

However, as Cohen points out in "Skepticism, Relevance and Relativity," the Relevant Alternatives Theorist can reply back to the skeptic: "Why should we accept your account of which alternatives are relevant and which are irrelevant? You're just begging the question against common sense, in claiming that those alternatives are relevant to the question whether we know."

We seem to have reached a stand-off. Both sides claim that the other side begs the question. How might this stand-off be resolved?

Well, we should distinguish two different sorts of anti-skeptical projects we might be engaged in.

The first project is to refute the skeptic using only premises that the skeptic will accept. (Call this "playing king of the hill.") Perhaps this project is doomed to fail. Perhaps any premises we'd need to show that we have knowledge would be rejected by the skeptic as question-begging.
The second project is to set our own minds at ease about the skeptic's argument. The skeptic has an argument from premises we find initially plausible to a conclusion we find unacceptable. We can't just go on believing the premises but denying the conclusion. If we want to avoid being forced to accept the skeptic's conclusion, we've got to diagnose his argument and explain where it goes wrong, or explain why although the premises seem plausible they're really false. (Call this "explaining away the skeptic in ourselves.")
If this is our project, it doesn't matter so much whether we say things the skeptic won't accept. It's only necessary that we accept them. After all, we're only trying to set our own house of beliefs in order. We just want to show to our own satisfaction that we can retain our common-sense views about knowledge without falling into the skeptic's trap.

A Problem for Contextualism

One natural way to think about the Contextualist's story is as follows. We have evidence for believing various things. And some of the things we beleive are true. But if your evidence for P is very slight, that won't be enough for you to know that P, even if P happens to be true. So we have the question: how much evidence does it take for a true belief to count as knowledge?

Now the Contextualist will answer this question by saying, "That depends upon the context." In some contexts, having evidence that makes P 99% likely might be good enough for you to know that P. In other contexts, e.g., when we're thinking about lotteries and so on, 99% might not be good enough.

What the skeptic tends to do is to get us into contexts where the bar for knowledge is very high. The Contextualist acknowledges that in contexts of that sort we don't count as having knowledge. But he says that we shouldn't lose sight of the fact that in other, ordinary contexts, where the bar for knowledge is lower, our evidence might very well be good enough for us to count as knowing.

That story does have some plausibility to it. But the problem is that it requires you to have some evidence against the skeptic's possibilities. In ordinary contexts, that evidence should be good enough to know those possibilities don't obtain; though in skeptical contexts the bar is higher and there the evidence isn't good enough for knowledge.

But do we any evidence against the skeptic's possibilities? What is our evidence that we're not in the Matrix, or not being deceived by evil demons? We would be having exactly the same experiences even if those possibilities were true. So it's not clear whether we even have any evidence at all that those possibilities don't obtain. (This is a point the Blumenfelds made, earlier.)

We need to think more about how the skeptic's arguments bear on issues about evidence and justification, not just knowledge. And we need to think about the relations between justification and knowledge. That is what we'll turn to next in the course.