An epistemic possibility is something that might be true, given the available evidence. I might see a bird tomorrow and I might not—given my evidence now, both options are epistemically possible. Not everything is epistemically possible though, as some options are ruled out by my evidence. I’m definitely not on the moon right now (if I were, I’d be suffocating!), so that’s not an epistemic possibility for me.
Epistemic possibilities feature prominently in several arguments about what we know. Suppose, for example, that you’re in a zoo. You’re standing outside the zebra enclosure, and you see an animal that looks exactly like a zebra—horse-like shape, black and white stripes, etc. But then a skeptic comes along and argues as follows:
Maybe that animal is really a mule that has been cleverly disguised to look just like a zebra. So, maybe it’s not a zebra after all. And that means that you don’t really know that it is a zebra.
‘Maybe’ here is an epistemic possibility operator, so the skeptic is pointing out that it is epistemically possible given your evidence that the animal is a mule cleverly disguised to look like a zebra. And this seems true. All of the evidence you have about how the animal looks is just the sort evidence you’d have if it were a mule cleverly disguised to look like a zebra, so that evidence can’t rule out that possibility. From this the skeptic infers that it is also epistemically possible that the animal is not a zebra, and therefore that you do not know that it is a zebra.
Both of those inferences seem sensible, but I think the first one is a mistake. Implausible as it initially sounds, I think it is epistemically possible for you that the animal is a mule cleverly disguised to look like a zebra, but it is not epistemically possible for you that the animal is not a zebra. To see why, consider the following test for epistemic possibility:
If you have a good reason to think that, if a proposition were true, you would have different evidence, then that proposition is not epistemically possible for you. Otherwise, it is.
This test handles standard cases of epistemic possibility well. It is not epistemically possible for me that I am on the moon, says the test, since I have very good reason to believe that I’d have different evidence (e.g. evidence of looking at a grey landscape through a spacesuit helmet) if I were. It is epistemically possible for me that I will see a bird tomorrow, since I have no reason to think that I’d have different evidence than I do if I were going to see a bird tomorrow.
Using this test, we can see why the skeptic’s inference is a bad one. It is epistemically possible for you that the animal is a mule cleverly disguised to look like a zebra, says the test, since you have every reason to believe that you’d have exactly the same evidence if it were. But now think about just the proposition that the animal in the pen is not a zebra. What evidence would you have if that were true? The standard way of answering that question is to think about the most similar scenarios where the animal is not a zebra and what evidence you would have in those scenarios. Plausibly, the most similar scenarios where you’re in a zoo looking at an animal that is not a zebra are just scenarios where you’re looking at some other animal, such as a giraffe or a crocodile. And in those scenarios, you would have very different evidence. So, our test actually says that, even though it is epistemically possible that the animal you’re looking at is a mule cleverly disguised to look like a zebra, it is not epistemically possible that it is not a zebra.
We can apply this thinking to other areas of philosophy with interesting results. The problem of evil is an argument that starts with all of the bad things we observe in the world and concludes that there is no morally perfect, all-powerful God. After all, such a being would surely prevent these bad things unless there was a really good reason to allow them, and there doesn’t seem to be any such reason.
There are many ways of attacking this argument, but one of them, skeptical theism, can be understood as pointing out that we cannot rule out the possibility that there is some reason we haven’t thought of that justifies the bad things we observe. So, the skeptical theist argues:
Maybe God does exist but has some reason beyond our limited understanding of good and evil for allowing all of these bad things to happen. So, God might exist after all. Which means that you don’t know that God does not exist.
This argument has the same form as the skeptic’s argument about the zebra, and it fails for the same reason. It doesn’t follow from the fact that it is epistemically possible that God exists and has a reason for allowing all of the bad things we’ve observed to happen that it is epistemically possible that God exists. We have very good reasons to think that if God existed and had reasons beyond our understanding for allowing the bad things we’ve observed, then we would have just the evidence we do. We’d see all of the same bad things happen, and we’d be just as baffled as to how a morally perfect being could allow them. But what evidence would we expect to have if God existed? That is, what evidence would we have in the most similar scenarios in which God exists? That’s a tough question, in part because scenarios can be similar or dissimilar to each other in different ways. But the answer does not follow from the epistemic possibility that there are reasons beyond our understanding for the evils we’ve observed.
Brandon Carey
Department of Philosophy
Sacramento State
Nice post, Brandon.
ReplyDeleteI like the way your proposed condition on epistemic possibility addresses the problem, as well as its useful applications, but I can't think of an independent reason to accept it.
In defense of your imagined skeptic, I would say that a very reasonable starting point is that: If it is epistemically possible that P; & you know that P implies ~Q, then it is epistemically possible that ~Q.
(Your students will recognize this as an application of the principle of epistemic closure, which most epistemologists prefer to preserve.)
Moreover, I do not find the proposed condition independently plausible. It strikes me as a plausible condition on probability, not possibility.
And I would also suggest we now have a better vantage point from which to address the skeptical argument proposed, namely externalism.
From (the standard) externalist perspective it just does not follow from the fact that it is epistemically possible that something is not a zebra that you do not know it is a zebra. Whether I know it is a zebra does not depend on whether I can countenance the possibility that it is not a zebra, but rather whether an actual zebra is producing my belief that there is a zebra present. (Of course, stating this condition precisely, whether in terms of sensitivity or safety or some other related concept remains challenging.)
Internalists just falsely assume that if I can countenance ~P, then I do not know that P. The kernel of internalist truth is a psychological fact, viz., that if imagining ~P actually weakens our confidence in P below a minimally acceptable level, then we do not know it. But if we retain our confidence in P in spite of these generally preposterous considerations (which normally we should) then our knowledge state remains the same as before.
Thanks, Randy!
ReplyDeleteI think you're right that some closure principle for epistemic modals is what motivates the inference here. I just think that principle is false, and that the counterfactual test I propose better matches up with (my) intuitions about epistemic possibilities.
If we're on a jury, say, and all of the eyewitness testimony, camera footage, etc. points to Kira, I think I would be saying something false if I said, "Kira might be innocent." But I think I'd be saying something true if I said, "Someone might have secretly gotten extreme plastic surgery to look just like Kira and then committed this crime to frame her." If my intuitions are on track, this is a counterexample to that closure principle.
The counterfactual test I've proposed gives just the right results, though. If Kira were innocent, we'd have different evidence, testimony from witnesses that some other person committed the crime, for example. But if the crime had been committed by a perfect Kira look-alike, we'd have just the same evidence. That's basically my reason for accepting the test: that it gets cases right.
If we also accept that knowing P requires that ~P is not epistemically possible, it also provides a solution to the lottery problem. It seems that, no matter the odds, it is epistemically possible for me that my lottery ticket will win, and again the counterfactual test gives this result. If my ticket were going to win, I'd have just the same evidence that it will almost certainly lose. But the same is not true of ordinary external world beliefs, such as that I'm looking at a zebra, even if the probability of my being wrong is the same. Those seem like good results to me, since it seems to me that I know that I'm looking at a zebra, but I don't know that my lottery ticket will lose.
Regarding externalism, I'm not sure it makes much of a difference here. As I've formulated it, the counterfactual test appeals to some internalist notions of reasons and evidence, but it doesn't follow that anything that can be countenanced is epistemically possible. In fact, that would be incompatible with the counterfactual test. I can countenance or imagine that there are unicorns, but that is not epistemically possible for me. I have good reason to expect that I'd have different evidence if there were unicorns--I'd have heard about them!
And there are even externalist versions of the counterfactual test. Instead of requiring a good reason to believe that if p were true, we'd have different evidence, we could instead just require that the counterfactual is true or that I safely/sensitively/reliably believe that it is true. This would give us a purely externalist account of epistemic possibility that makes (almost) the same problem for these arguments. Depending on how we formulate it, it would be incompatible with some externalist accounts of knowledge, but all that shows is that externalists also need a theory of epistemic possibility and its relation to knowledge.
Thanks Brandon. That's all really clear, and it's almost all persuasive. But what you find intuitive, e.g., that I would be saying something false if, in spite of strong evidence of Kira's guilt and none of her innocence, I were to say Kira might be innocent, I find Super-de-dooper unintuitive. For me, that's just what evidence is, a non-demonstrative basis for rational belief. When I say the evidence points to Kira's guilt, I am thereby acknowledging that she might not be guilty, but also asserting that the probability of her guilt is very high.
ReplyDeleteOne thing that may be impeding my understanding, though, is what you actually mean by epistemically (im)possible. You propose a test for epistemic impossibility, but I don't know if you mean that to be a partial analysis or just a criterion that flows from it. Can you say any more about that?
Sorry, Randy. I missed this reply.
ReplyDeleteThe test is more or less a consequence of my proposed analysis of epistemic possibility which is:
p is epistemically possible on a body of evidence E if and only if E supports that if p were true, E might exist.
On that view, just the fact that there is some probability that p is not sufficient to make p possible, so if by 'epistemically possible' you just mean something about probability, this won't be particularly appealing.
The idea of epistemic possibility I'm after, though, is roughly what's expressed by ordinary uses of 'maybe'. If, given ordinary very good evidence that Kira is guilty, you said, 'Maybe Kira is innocent', I think you'd be saying something false, even though I agree that there is some non-zero probability of her innocence. Similarly, if I said, 'Maybe I'm on the moon', I'd just be wrong, even though there is some chance that I'm on the moon in a perfect holographic simulation or something. On the other hand, if I said, 'Maybe my lottery ticket will win', I'd be saying something true, because I have very good reason to think I might have just the same evidence if my ticket were going to win. I don't think the difference there is a matter of probability (since we can manipulate the probability in the lottery case to be whatever we want), which suggests that it results from some other kind of difference in the evidence.