Two Bootstrapping Problems

I mentioned in passing in the Kornblith post that there were two distinct puzzles about bootstrapping, but I let the point slide fairly quickly. This post is a short clarification of the two puzzles, and then a request for further info. I meant to post this before the Kornblith post, but I seem to have got muddled about when I hit ‘Draft’ and when I hit ‘Publish’.

Here’s a quick version of the bootstrapping problem. I don’t know that my colour vision is reliable, but I do know that my introspection of what colours I appear to be seeing is reliable. I look at something which happens to be red, as it turns out, a copy of Stefano Predelli’s Contexts, and reason as follows.

(1a) That appears to be red (by introspection).
(2a) That is red (by colour vision).
(3a) So appearances match reality on this occasion.

I repeat this with a bunch of other things, ending with a copy of Parts of Classes.

(1z) That appears to be purple (by introspection).
(2z) That is purple (by colour vision).
(3z) So appearances match reality on this occasion.

I then infer from (3a) through (3z), (4)

(4) My colour vision has worked perfectly the last 26 times I’ve used it.

And from that I conclude (5).

(5) My colour vision is generally reliable.

And that’s a bizarre thing to be able to conclude in this way. At both the Rutgers Epistemology Conference back in May, and the Arché Scepticism Conference back in June, there were a lot of people noting that the step from (4) to (5) is problematic in a number of ways. There may well be general inductive rules that block this particular inference. (At the Rutgers conference Jonathan Weisberg presented what is, I think, the best worked out version of such a blocking rule.) But there were also, at each conference, a number of people arguing that even if the step from (4) to (5) could be blocked, it still seems bad that various views (such as reliabilism, or some kinds of dogmatism about perception) can get to (4). (Stewart Cohen, for instance, made this point at the Arché conference.)

Now here’s a question I have for the audience. I’m a little behind in my reading, and a long way behind on my reading on what’s in print. So does anyone know where the points I mentioned in the previous paragraph are made in print? I’m really just looking for something that distinguishes arguments from the unknowability of (5) to an epistemological conclusion from arguments from the unknowability of (4) to an epistemological conclusion. (Of course, I don’t mean (4) and (5) are unknowable in general, just that they are unknowable on this occasion.) A quick scan of what’s in print didn’t reveal much, but I suspect I’m missing a lot.

8 Replies to “Two Bootstrapping Problems”

  1. Isn’t there’s some relevant discussion in Roger White’s anti-dogmatism paper? There White argues that one can infer a claim he labels NO FAULTS, which is more or less 4, if I’,m remembering correctly. This is supposed to be problematic since it’s hard to explain why NO FAULTS is true without appealing to something like 5.

    (This perhaps isn’t quite what you’re after, since it’s more a way of responding to those who think the step from 4 to 5 is resistable, rather than an argument that reaching 4 is bad independently of whether we can move from 4 to 5. But it was one of the considerations Cohen mentioned at the conference, and it seemed relevant enough to be worth mentioning.)

  2. In A Contextualist Solution to the Problem of Easy Knowledge, Ram Neta (Grazer, 2005) provides some ampliative closure schemas in an attempt to distinguish moves such as that between (4) and (5) that are acceptable from those that are not.

  3. I can’t tell if your worry is a general one about principles that get one to 5, or in particular a worry about inductive principles that get one to 5. Many people are now of the opinion that one can do bootstrapping without any inductive step at all. For instance, suppose we add to your example that you know in the background that your color vision is either reliable or anti-reliable (in which case it goes wrong every time). Then you can skip from 3a immediately to 5 via a deduction. Of course, whatever objections one has to an inductive principle that gets from 4 to 5 might also be raised about the deductive principle that would be needed to get from 3a (plus your background knowledge) to 5. But the problem doesn’t look to me like it’s about induction per se.

  4. Mike,

    That’s a nice version of the problem. I guess any case where the subject knows in advance that if they are right on some cases, they are right on all cases, will get the puzzle going without an inductive step?

    Do you know if anyone has written that up?

  5. The reliable/anti-reliable example is original to me, but as you note it’s a tidy version of a general class of cases in which a method’s providing one right answer implies that it is reliable. (Hence the immediate deductive step to that conclusion.) It’s an interesting question whether anyone has written up the fact that bootstrapping can be done without induction. I couldn’t find the point on a quick scan of the articles I have to hand, but perhaps someone with a deeper knowledge of the literature can chime in here?

  6. I clearly don’t understand something. So far as I can see, the problem starts by the time you get to 3a!

    Consider someone whose colour vision is unreliable. In fact, let him be unable to distinguish between red and green (both appear to be green to him). When looking at a copy of Contexts, his colour vision tells him it’s green. Then why wouldn’t he reason as follows:
    (1a’) That appears to be green (by introspection).
    (2a’) That is green (by colour vision).
    (3a’) So appearances match reality on this occasion.

Leave a Reply