I like evidentialism.
(EJ) S is justified in believing P iff P is best supported by S’s total evidence.
I also like phenomenal conservatism.
(PC) If it seems to S that P and S has no defeaters for P, then S is justified believing P.
I take it that seemings (absent defeaters) are a kind of supporting evidence so I think that PC fits well with EJ. Lots of fun puzzles and problems for both of these, but that’s not my present concern. Let’s move on to defeaters.
I won’t get into all the nitty, gritty details concerning defeaters – but I am sympathetic to the idea that if one has strong seemings that one’s input source is unreliable, then one has a defeater.
For example, suppose it seems to me that the table is red. Now suppose it also strongly seems true to me that the source of the seeming are my eyes, and it strongly seems true to me that my eyes are generally unreliable with respect to forming color beliefs. If this is true, then I think I have a defeater, and I’m inclined to think that I should suspend judgement.
Now, I’m not a reliabilist. That is, I think you can be justified in believing some proposition even if your faculties are not reliable. Furthermore, I don’t think PC is true because seemings are reliable indicators of truth.
But here is something that is, I think, prima facie odd about my position. One might ask (and I have been asked this) how could evidence of unreliability undermine justification if reliability had nothing to do with justification? That’s the question I want to tackle here.
If this really is to be a problem, then there must be some principle concerning the relationship between instantiations of properties and facts about whether one of those properties factors into the analysis of the other. I think something like that is being assumed when the above question is asked.
It could be that the following principle is being assumed. (Let P and Q range over properties).
(IP) If the instantiation of not-P entails Q is not instantiated, then P factors into the analysis of Q.
Someone might employ (IP) to say that my view commits me to the thesis that reliably formed belief factors into the the analysis of justification. Because I’m saying that the unreliably formed beliefs entail the lack of justification.
But notice that’s not what I’m saying. I’m not saying that unreliability entails lack of justification. I’m saying that evidence of unreliability counts as a defeater. So at most, evidence of unreliability would entail lack of justification. I it’s easy to confuse something in the neighborhood of (IP) with something like (IP*)
(IP*) If my having evidence of the instantiation of not-P entails not-Q, then P factors into the analysis of Q.
If this principle were true, then I might have problems. But this principle is surely false. Suppose I have evidence that someone is not a smoker. This entails some other things about my mental states, like I’m not unconscious. But the property of being a smoker doesn’t factor into the analysis of the property of being-unconscious.
That’s all for now. This was just a little warm-up exercise to see if there was a problem here that would be worth fleshing out more. I think there may be.
Andrew,
It seems like the kind of defeater in question here is an undercutting defeater, one that undermines the source of one’s justification. That is to say, if I have a justified token j which supports a belief token p, if a defeater comes along that undermines the justification-producing type J which produced j, this wouldn’t necessarily imply that the justification for my belief p has been negated (contra reliabilism), but it would force me to acquire some additional justification for p–j would no longer be sufficient evidence. I think you can require defeasibility of this type without the stronger thesis of reliabilism.
Ezra
Hey Ezra,
You’re absolutely right. What I’m discussing her would qualify as an undercutting defeater (as opposed to a rebutting defeater).
And I’m also in agreement that you can allow for these without being committed to tacking on a reliability requirement.
Ultimately, I’m just trying to come up with the most charitable interpretation of what someone might have in mind when they raise the above question.
Andrew,
I think I disagree with you but I am not sure why. Let me see if I understand your position. Here are two theses, which I will call “Weak Reliabilism” (WR) and “The Defeater Principle” (DP):
(WR) S is justified in believing P only if S’s belief that P was formed by a reliable belief forming process.
(DP) If it seems to S that P was formed by an unreliable belief forming process, then S is not justified in believing that P.
I take it that you want to endorse (DP) and reject (WR). It is clear that this is a consistent position. But, it seems weird to me. You seem to suggest that one way of stating your position is as follows: even though reliability has nothing to do with justification, if it seems to you that your belief was formed unreliably, you have a defeater for your belief. But, now I have the question you wanted to tackle: why is it that such a seeming will act as a defeater for one’s justification when the thing that is seemed has nothing to do with justification?
The following No Difference Principle seems plausible to me:
(ND) If the reliability of a belief forming process X makes no difference to whether a belief is justified and S knows this to be the case, then whether it seems to S that X is unreliable makes no difference to whether a belief is justified.
Now it seems that ~(WR), (DP) and (ND) are incompatible. That is, there are incompatible if ~(WR) is a way of saying that reliability makes no difference to justification and (DP) is a way of saying that seeming unreliability does make a difference to justification. If all this is correct and (ND) really is plausible, then we have a reason to either accept (WR) or reject (DP).
Oops, I should say
“they are incompatible if ~(WR) is a way of saying that reliability makes no difference to justification and (DP) is a way of saying that seeming unreliability does make a difference to justification and at least one person knows that ~(WR)”
That seems like a fair characterization of the position. I give up on (ND). Here’s one way to motivate the rejection.
Having evidence that a belief that is formed by an unreliable process is but one way to have evidence that the belief is not likely to be true. When you have evidence that a belief is not likely to be true, then you have good reason to give up the belief. It’s your evidence that the belief is not likely to be true that undermines the justification. Doesn’t it seem plausible that evidence concerning the likelihood of truth is what’s relevant?
It’s the evidence concerning likelihood the makes the justificatory difference – the actual likelihood of the belief’s truth is irrelevant.
What do you think?
I have a couple of questions. You want to say that a evidence that a belief is not likely true gives you good reason to give up the belief even though the likelihood of the belief makes no difference to justification. Do you also endorse that a seeming that supports that a belief is not likely true is a defeater for the justification for that belief even though the likelihood of the belief makes no difference to justification? If you do endorse that claim, then can’t we ask a question here that is similar to the one asked above? Why is it that such a seeming will act as a defeater for one’s justification when the thing that is seemed has nothing to do with justification?
Also, I am not sure I understand what kind of likelihood you are talking about. It is clearly not objective likelihood since we all agree that you can be justified in believing something that is objectively unlikely. So, it must be some kind of conditional likelihood. But what kind of conditional likelihood?
I was thinking objective likelihood. I agree with you that you can be justified in believing something that is objectively unlikely. I’m actually starting to wonder if we’re not on the same page here.
If you had evidence that a belief was objectively unlikely to be true, doesn’t that seem like good reason to suspend judgement? So…if you think that the mere fact that a belief is objectively unlikely doesn’t matter for the justificatory status of the belief and you think that evidence of objective unlikelihood does – aren’t we in the same boat?
Regarding the modified question. You say yourself that we can be justified in believing something that is objectively unlikely. So, it’s objective unlikelihood is not relevant to the justification. I’m willing to maintain that a seeming that supports thinking that the proposition is objectively unlikely would count as a defeater.
Regarding the modified question. I’m not sure how to go about answering the question. But perhaps considering the following case will help yield an answer.
Consider a simple case of it seeming that the table is red. I take it that the seemings that tables are red are relevant to whether or not I’m justified in believing that the table is red. However, the mere fact that the table is red is not relevant (except for perhaps the role it plays in causing me to have the seeming, but that’s not the sort of relevance we’re talking about – by “relevant” we mean something like it’s a feature that would factor into a condition on justification).
So, in general, it’s not weird to say that a seeming that P is relevant to the justification that P, but the fact that P is not.
That’s not an answer to the modified question, but perhaps a reason to think that there must be some good answer. I’ll have to think more about answering the question.
(Thanks for taking time to discuss this…this is really, really helpful…at first I thought there wasn’t a real puzzle here, but now I’m seeing that there is one)
Hey,
I haven’t thought about all your responses, but I do have something to say about likelihood. I think we can be justified in believing something that is objectively unlikely and, moreover, I think we can be justified in believing something and believing that it is objectively unlikely. Consider a scientist who is observing a particle. He might know that it is objectively unlikely that the particle will decay at 12:00. But, nevertheless, he observes the particle decay at 12:00 and is thereby justified in believing that it decays at 12:00. In this situation, not only is it objectively unlikely that the particle decays at 12:00 but the scientist knows this fact and nevertheless knows (by observation) that the particle decays at 12:00.
Some people might object to the above story by claiming that something cannot both be true and objectively unlikely. But, I disagree. I think that objective likelihood and unlikelihood has to do not only with actual truth but also with truth in some relevant set of nearby possible worlds. So, something might be true but also unlikely (if for example it is false in most nearby worlds).
I guess I don’t think that objective likelihood has much to do with justification (if anything at all). But, there is a kind of epistemic likelihood which is related to justification. Epistemic likelihood is the likelihood of a proposition for S given S’s epistemic position. Now, I take it that this kind of likelihood does have something to do with justification. If P is epistemically unlikely for S, then S is not justified in believing that P. But, whether S’s epistemic position includes facts about reliability is something that is at issue. If reliability is included as part of S’s epistemic position, then whether S’s belief was formed reliably or not will have an affect on whether that belief is epistemically likely and hence have an affect on whether S is justified in his/her belief.
This discussion has provided me with an excellent opportunity to procrastinate on grading! I’ve been checking your blog in between grading individual tests.
Good point. I agree with you with respect to what you say about objective likelihood. I guess I did mean something slightly different from objective likelihood.
A reliabilist defines the reliability of the faculty in terms of truth at nearby worlds where that faculty was used to form the belief. So I guess the likelihood at play here is conditional (in some sense) in that the class of nearby worlds is restricted by the faculty being used.
I was hesitating to call it conditional, because conditional is sometimes used to mean something like subjective. That’s not what I meant.
I also share your idea that there is some sense of epistemic likelihood that is relevant to justification…I just thought that the reliabilist notion of likelihood would be outside of that because I thought the feature of objective likelihood that made it irrelevant to justification would be shared by the kind of likelihood that the reliabilist is talking about – namely the likelihood is a fact that the agent doesn’t necessarily have evidence for.