It’s not the thought that counts

Let’s say that the mentalist about evidence believes the following supervenience thesis:

M: Necessarily, if A and B are in the same non-factive mental states from the cradle to the grave, A and B will share the same evidence from the cradle to the grave.

Here’s an argument against mentalism, so understood:

(1) We have non-inferential knowledge of the external world.
(2) If we know p non-inferentially, p is part of our evidence.
(3) If ~p, p is not part of our evidence.
(4) It is possible for someone to be in just the same non-factive mental states as any one of us and believe mistakenly that p.
(5) We know p non-inferentially.
(C) It is possible for someone to be in just the same non-factive mental states as any one of us and while p will not be part of their evidence, p will be part of ours.

Let’s bracket (1) for now. I think Feldman and Pryor do a nice job defending the claim that we could have such knowledge. No point in conceding (1) but denying (5). Let’s bracket (4) for now. I don’t think anyone was going to take issue with it. Now our question is this: should the mentalist deny (2), (3), or both? I think these should the yadda yadda yadds deny blah blah blah claims are funny, so let me rephrase. What plausible responses could the mentalist offer in response to the argument apart from the table hammering response that says that it is just intuitively obvious that (C) is false. If you don’t like the argument, that’s fine, but a diagnosis is what I’m looking for.

In his paper in the new volume on Williamson, Goldman suggests that our evidence will consist of the contents of our non-inferentially justified beliefs. I’ve seen similar things suggested by a number of people. Since no one seems to like the view that says there are no false, justified beliefs (apart from Sutton and myself), the view that Goldman is proposing as an alternative to Williamson’s appears to be a view that denies (3). Williamson remarks that on this sort of view:

it is not even clear that one’s evidence must be consistent, let alone true … there are grave difficulties in making sense of the evidential probabilities on inconsistent evidence, since conditional probabilities are usually taken to be undefined when conditioned on something inconsistent. In particular, any proposition has a probability 1 conditional on itself and any contradiction has probability 0 on anything … but these constraints cannot both be met for probabilities conditional on a contradiction” (2009: 310).

For obvious reasons, those of us who think that justified beliefs are factive will not think that the propositions you justifiably believe form an inconsistent set of propositions but those who deny this will have a harder time denying that it’s possible for someone to justifiably believe inconsistent sets of propositions. The questions then are these. If you deny that evidence is factive, what view do you adopt to rule out the sort of case that worries Williamson? Is it a big deal if you adopt a view of evidence on which evidential probabilities are undefined?

It seems that we have the beginning of an argument for the factivity of evidence. Let me add to the evidence for factivity. Consider:
(6) There is a reason for him to think he has a flu; namely, that his doctor told him he has come down with a flu.

The following just sounds contradictory:
(7) There is a reason for him to think he has a flu; namely, that his doctor told him he has come down with a flu. Having said that, his doctor didn’t tell him that.

Obvious explanation: claims of the form ‘There is a reason for him to believe p; namely, that q’ entail, inter alia, q. Maybe there’s some subtle difference between reasons to believe and bits of evidence, but it’s not as if things get better if we use evidence-talk instead of reasons-talk:
(8) What’s his evidence for believing that he has a flu? That his doctor told him that he has come down with a flu. Having said that, his doctor never told him that he had a flu.

You can say, ‘Well, but his evidence really will be a mental state of some sort he could be in even if his doctor hadn’t spoken to him or some fact about such a mental state’. You can say that (6) is really just elliptical for some longer description of mental facts that would be true whether the subject was hallucinating, dreaming, or veridically perceiving. Okay, but that’s just using Factivity and mentalism to redescribe the evidence in terms of truths related to but distinct from those I say are entailed by (6). Things don’t get better when you say:
(9) What’s his evidence for believing that he has a flu? That he seemed to hear his doctor tell him that he has come down with a flu. Having said that, I should note that it never seemed to him that his doctor said this.

Finally, think about the role that evidence plays in inference, such as inference to the best explanation. Suppose you know that p is part of your evidence. (Note: I didn’t say, ‘Suppose you know that p is true’). Question: is there something further you need to know in order to know that it is either a brute fact that p or there is some explanation for p? Intuitively, it seems to me that anyone who knows that p is part of her evidence is thereby in a position to know that either there is an explanation for p or p is a brute fact. The first disjunct entails p. If you know there’s some explanation for p, you know there’s some true statement of the form ‘p because q’, and we know that ‘p because q’ entails that both p and q are true. The second disjunct entails p. You know that ‘It’s a brute fact that p’ entails p. So, you know that if p is part of your evidence, a disjunction that entails p is true. So, you know that if p is part of your evidence, p is true. I can’t see how that could be unless p’s being evidence requires, inter alia, that p be true.

Those who deny factivity I think have to say that when you know that p is part of your evidence, you could still fail to be in a position to know whether the question ‘Why is it that p?’ rests on a mistake. (The question ‘Why do we lose weight when we die?’ rests on a mistake because the answer is ‘We don’t’ and I’m saying that once you know that p is part of your evidence, you no longer have to worry whether the question ‘Why is it that p?’ rests on a mistake.)

Knowledge and Possession
According to (2), you don’t need more than non-inferential knowledge of p’s truth to acquire p as part of your evidence. We can call this ‘IKSE’ for short (it’s short for immediate knowledge suffices for the possession of evidence and probably contains lots of stuff that you won’t need but couldn’t hurt). Suppose S knows p non-inferentially. One possibility is that p isn’t the sort of thing that could be evidence. Is that because evidence can’t be a proposition or fact? Maybe, but let’s assume that evidence is propositional just for fun. I take it that most of Williamson’s critics aren’t criticizing him simply on the grounds that they don’t think evidence is propositional. If evidence are the thoughts (true or false) that you have in mind, p will be in mind when you know p non-inferentially. If evidence consists of facts or true thoughts/propositions, we don’t have to worry about whether p since it is known to be true. So, maybe the worry about IKSE is this: there’s more to possessing a piece of evidence than knowing non-inferentially that the proposition that constitutes evidence is true.

Okay, so suppose we talk about reasons for belief. If you know p non-inferentially and know q to be a deductive consequence of p, can’t we say that you have a reason to believe q because you know p non-inferentially? I think so. Aren’t you being awfully picky about evidence if you say that I’m ignoring the all important distinction between reasons for believing that explain why a belief is justified and evidence proper?

When I see the sorts of things that get floated as platitudes about evidence, it seems that _things we know non-inferentially_ fit these platitudes. If our evidence is that which we have to go on in trying to settle a question, you clearly _can_ go on the things you know non-inferentially. Maybe the platitude is really that evidence is that which we can properly rely on in theoretical deliberation. Okay, is it _improper_ to rely on p when you know p without needing to know anything else? If it is improper to rely on p for the purposes of theoretical reasoning where the impropriety is epistemic, that should affect the justificatory status of the belief in p. That would seem to prevent the belief from constituting knowledge. Wait, have we just hit upon an argument for IKSE? If you are non-inferentially justified in believing p, it cannot be epistemically improper or impermissible to rely on p in theoretical reasoning. So, there are no counterexamples to IJSE, the claim that if p is a piece of evidence and you are non-inferentially justified in believing p, p is part of your evidence. No counterexamples to IJSE means no counterexamples to IKSE because IJSE entails IKSE! Fantastic.

I think there’s a decent case to be made against mentalism, but I could be horribly mistaken. Yes, I realize that the data is complicated. I’ve never claimed that there’s no evidence against Factivity and I’ve never claimed that there’s no evidence against IKSE. But, there are arguments that suggest that that evidence is misleading. Are they any good?

Late addition. I was eating a taco (yes, it was delicious!) and I realized I forgot to post something that I’ve been thinking about lately that I think is interesting. If I try hard enough, I can make myself think it is related to Greg’s question below.

In the post, I mention an argument for factivity that focuses on the role that evidence plays in explanatory inferences. The example I had in mind was that of non-normative explanations of observational evidence. I gather some evidence by means of observation. If know that p is part of my evidence, I know that p is either a brute fact or an explanandum for which I should find an explanans, explanandum and explanans propositions are true, the same is true for brute facts, logic, and factivity. What about normative explanations? Here, I have in mind claims of the form ‘You shouldn’t believe p because q’. Now, I think that when I know that q strongly disconfirms p and know that q is part of my evidence, I can often say things like this that are true. This seems very much in the spirit of evidentialism. But, it seems that ‘You shouldn’t believe p because q’ entails both ‘You shouldn’t believe p’ and ‘q’. So, while I think that evidentialists will want say that it is evidence that explains the epistemic oughts and shoulds, there is some linguistic evidence that suggests that they are on the hook for accepting the factivity of evidence that derives from the factivity of explanations.


It’s not the thought that counts — 10 Comments

  1. Perhaps it reduces burden shifting, but why would a mentalist start with (M) but grant (3)? Isn’t that the nub of it?

  2. Hey Greg,

    Good question. That’s what I had always assumed. If they deny (3), I’d say that that’s great. I reject mentalism, so I needn’t worry about all that squirming. M –> ~3, but 3, so ~M. I’m happy to do some explanatory burden shouldering. Maybe I can’t convince mentalists, but as I don’t see much evidence that supports M I feel obliged on broadly evidentialist grounds to reject it or at the very least suspend judgment on it in light of the evidence for 3.

    That was partially sort of tongue in cheek. I think some mentalists accept 3 (Conee and Feldman in the Q. Smith edited collection seem to assume it in criticizing McDowell but it’s buried in a footnote and their commitment to it is not very explicit). The mentalist could insist that the propositions that get into our evidence are propositions about matters that we would not be mistaken about if our experiences were not veridical, our memories not genuine, our moral intuitions supported Republican positions, etc… So, either they can retreat to a more classical version of foundationalism that denies we can have knowledge of the external world that isn’t based on knowledge of the internal world or allow that we have such knowledge but insist that the scope of our evidence is not as extensive as the scope of our non-inferential knowledge. For some reason, our evidence will consist of truths we know via introspection but not any truth that can be known via observation unless it can also be known via introspection.

    What I’ve found in trying to criticize the view is a lot of squirming. It’s like trying to catch a squid. I think I have the thing in my grasp and the waters get murky with defensive sprays of ink. I run this argument and I can’t get two people to agree on which premise the mentalist will deny.

  3. Maybe I should say more: Denying (3) does not entail Goldman’s alternative; Even so, if one is worried about inconsistent evidence, then simply add a consistency condition for non-factive evidence; Or, in the end, if you like Goldman’s view, you’re likely to find living with some undefined probabilities preferable to an unrealizable model of evidence.

  4. Hey Greg,

    I think that’s right. You can deny (3) and still avoid Goldman’s alternative view, but at some point someone like Williamson is going to say that we have evidence, it fixes our evidential probabilities, and without (3) the restriction to consistent propositions is ad hoc. If someone happens to believe inconsistent propositions (e.g., when you take a series of experiences at face value where those experiences in total represent an impossible state of affairs but you don’t appreciate that this is so) or happens to justifiably believe them, what explains how these propositions are excluded from the evidence if (3) is false?

    Maybe the thought is that living with undefined probabilities isn’t the worst thing in the world.

  5. Ah, I see. Interesting footnote! I can understand your frustration, then.

    I like the idea of non-factive evidence because I think we deal in incomplete, noisy, uncertain information. It might be that a Williamsonian model is helpful as a last step, from highly refined evidence to a decision model, say, and so we are interested in different questions. That’s some part of it.

    Another part might be the focus on propositional formulae and a numerically determinate probability function. If one insists on looking at the problem with these tools, then, yes, consistency is going to be important, the fuss over conjunction in lottery and preface cases will look like a mistake, and Lockean detachment will look pointless.

    This last part is to say that maybe the mentalists have their heart in the right place (from my point of view), but they are wedded to the wrong tools…or to tools that will make for squisky, squirmy arguments about factivity.

  6. Clayton,
    Why does IJSE entail IKSE? It seems to me it doesn’t.

    Most people (sounds like you’re an exception) think there are immediately justified false beliefs. Say we go with the majority here. I have an immediately justified false belief that p. According to IJSE, p is part of my evidence. But according to IKSE it’s not. So if IJSE, then ~IKSE. So IJSE doesn’t entail IKSE.

    When you said IJSE entails IKSE, were you assuming that there’s no immediately justified false beliefs? To assume that in an argument against mentalism seems like begging the question to me.

  7. Hey Dylan,

    Good point. I didn’t formulate the relevant paragraph as carefully as I should have. I wrote:
    “If you are non-inferentially justified in believing p, it cannot be epistemically improper or impermissible to rely on p in theoretical reasoning. So, there are no counterexamples to IJSE, the claim that if p is a piece of evidence and you are non-inferentially justified in believing p, p is part of your evidence.”

    The following seems like a possibility: it possibly proper to treat p as a reason even if p is not a reason (e.g., when you have good reason to think it is a reason). I took IJSE to be something like this:

    (IJSE) If (p is a piece of evidence & S belief in p is non-inferentially justified), p is part of S’s evidence.

    I don’t think that ‘p is a piece of evidence’ entails ‘p is part of S’s evidence’, so the first conjunct in the antecedent of the conditional is not redundant if you think that evidence consists of truths but don’t think that all non-inferentially justified beliefs are true. If you think evidence can consist of propositions true or false, that proviso might be vacuous. We don’t need a similar proviso in IKSE because knowledge is factive.

    So, assuming that S knows p entails S’s belief that p is justified, I think IKSE does follow from IJSE if we mind the proviso in the antecedent of the conditional.

  8. So you meant to combine IJSE with the factivity of evidence. Put those together and you get IKSE. Is that it?

    OK, what do you think about this? I have an immediately justified belief that there’s a sheep in the meadow. It’s immediately justified by my experience, which represents there as being a sheep in the meadow. I’m not sure what it takes to be a piece of evidence. But maybe is a piece of evidence here (?) Then by IJSE, it’s part of my evidence. But it’s one of those non-inferential Gettier cases — there is a sheep in the meadow, but I’m looking at a white rock that looks just like a sheep from my perspective. So I don’t know there’s a sheep in the meadow. So according to IKSE, isn’t part of my evidence.

    I have a feeling this isn’t really a counterexample, and it all has to do with my not really grasping what it means for p to be a piece of evidence.

  9. Hey Dylan,

    It’s a tough question. My intuitions about these kinds of cases just aren’t clear at all. Suppose we start with the constitution question, what is a piece of evidence. I say it’s a fact. Sure, it’s a fact that stands in relations to other things just like uncles stand in relations to other things, but an uncle is still a human male. We turn to the possession question. I take this to be partially normative. Facts about what you should/shouldn’t believe seem to follow from facts about the evidence you have, so maybe facts about evidence possession should be sensitive to normative stuff. No normativity out without normativity in. Then the question, I guess, is this: assuming p is a reason, could p fail to play the evidence/reason role if justifiably believed without further supporting evidence for p?

    I’m sort of inclined to say that it cannot. But, I’m also sort of inclined to say that my intuitions about j-ascription are constrained by/polluted by intuitions about the proper treatment of considerations in deliberation. To the extent that I don’t think it’s improper to reason from the premise that there’s a sheep in the field, I’d say J’d belief and thing believed is evidence. However, given the kind of disconnect here between fact that makes the attitude true and the surrounding mental states, I’m also sort of inclined to say that the proposition believed is not evidence. It is, at the very least, something that makes it very likely that the agent will be misled.

    Further case that worries me: veridical perception cases with BIVs. I think Jonathan Ichikawa deserves credit on that one, but my memory could be failing me.

    As for what it means for p to be a piece of evidence, I think one platitude is that p is your evidence when p is evidence and it is the sort of thing you can (properly) reason from when trying to settle an issue. That doesn’t settle the constitution question, but I take it that one thing that evidence is supposed to do is explain why it is that certain attitudes are rational. If facts are what explain things and evidence is thought to sometimes play the role of explanans in normative explanations, more evidence that evidence consists of facts. If evidence is thought to be the things that play the explanandum role, the thing that plays that role is a fact.

    I don’t know if that helps.

Leave a Reply

Your email address will not be published. Required fields are marked *