What I believe about easy knowledge
Mon 28 Apr 2008 06:10 AM
I've been thinking about this since the conference a couple of weeks ago.
The problem of easy knowledge is alleged to put the kibosh on reliabilism.* Consider, for example, a situation in which I make a series of perceptual judgments. There are many piles of cardboard tokens on the table and I count the number in each. For the first pile, I form a belief B1 about the number of tokens in it.I can reason
1. B1. (That is, B1 is true.)
2. I believe B1.
3. Therefore, I formed a true belief here.
4. Repeating this process for each pile: I formed a true belief in every case.
5. So the process at work in these instances of chit-counting is reliable.
6. Having a belief formed by a reliable process is sufficient for knowledge. (Reliabilism.)
7. Therefore, I know each of the beliefs.
There is no problem with saying that line 7 is true, because we are setting aside sceptical worries here. Nevertheless, there is something wrong with thinking that I can learn line 7 in this way. I extrapolate knowledge about my knowing from mere belief by a kind of legerdemain.
An epistemic rabbit from a doxastic hat!
Second-order knowledge is not really so easy, so something has gone wrong. If premise 6 is where the absurdum comes in, then this is a reductio of reliabilism.
There are several ways to respond:
First, one might problematize the step from the collection of instances to the claim that they are formed by reliable process (4-5). In his conference paper, Colin Caret argues that this inductive move doesn't come for free. It requires a further premise that all of these instances of pile counting are the result of the same process. That premise, he argues, requires some heavy hitting cognitive science. The easy knowledge is not so easy after all.
Second, one might insist that similar problems arise for non-reliabilist epistemologies. As such, it is everybody's problem and not a reductio of reliabilism. Hilary Kornblith suggested this approach after Colin's talk. It has the same bitter taste that accompanies any tu quoque reply.
Third, one might deny that I am entitled to intermediate conclusion 3. This seems awkward, because 3 is a deductive consequence of 1 and 2. Certainly the inference is unobjectionable in the third-person:
A1. P is true.
A2. Colin believes P.
A3. Therefore, Colin believes something true.
Consider, however, what I need to do in order to assess the soundness of this alternate argument. In order to know A2, I need to learn Colin's doxastic state. In order to know A1, I need to learn about whatever part of the world P is about. This is not trivial, at least not as trivial the easy argument is supposed to be. It won't be a reductio if we change the first-person 'I' to the third-person 'Colin'.
When I ask myself whether Colin believes P, I observe him and infer his doxastic state. (Perhaps I just ask him and accept his answer.) When I ask myself whether I believe P, however, I do not simply introspect and determine my preexisting doxastic state. As Gareth Evans and Richard Moran have noted, my determining whether 'I believe that P' is true typically just becomes my determining whether 'P' is true. Determining whether I believe P requires determining whether P, and I mobilize whatever cognitive resources are relevant to it.
In the argument, I am supposed to learn about my own doxastic state by introspection. My introspection just notices that I've formed a belief in B1 by whatever cognitive process is at work. This is the very same state of affairs recorded in premise 1, when I work through the argument for myself. I am tempted to say that 1 and 2 are not even distinct premises here.**
In any case, I don't see how I can accept premise 1 and premise 2 in the way the easy knowledge argument requires. They only make sense as distinct claims if we treat 2 as a third-person belief ascription, but then it no longer comes along for free. So the problem is not with reliabilism at all.
* Specifically, this is the bootstrapping version of the problem.
** Moran argues that 2 as distinct from 1 is coherent, but that rational agents almost never pose 2 in that way. (He calls it the 'theoretical question.') Ron (in conversation) suggested some cases where one might actually pose it. Perhaps I just need for this context to make it impermissible to pose 2 in its distinct, theoretical sense.***
*** I am treating it here as if it's a semantic point. A parallel point might be made if it's a pragmatic difference. (I just footnoted a footnote. I think that is a blogging first for me.)