My account of inference (see “Inferring as a Way of Knowing”) involves rejecting the view that, when you infer, you come to believe your conclusion in part because you take your premises to support it. In other words, it involves rejecting the second clause of the increasingly well-known “Taking Condition”:
The Taking Condition: “[i]nferring necessarily involves the thinker taking his premises to support his conclusion and drawing his conclusion because of that fact” (Boghossian 2014: 5, his emphases).
It’s a familiar fact about inference that you can come to know your conclusion by inferring it from your premises only if you already know your premises. But the same thing is arguably true of the sorts of “takings” mentioned in the Taking Condition. That is, arguably, you can come to know your conclusion by inferring it from your premises only if you already know that your premises support your conclusion. In both cases, the qualifier “already” is important. Your knowledge of your premises needs to be prior to your knowledge of your conclusion; otherwise, your knowledge of your premises wouldn’t ground your knowledge of your conclusion. And similarly for your knowledge that your premises support your conclusion.
Part of the idea behind my account of inference is that the second of these claims––that you can come to know your conclusion by inferring it from your premises only if you already know that your premises support your conclusion––is questionable. In fact, my suggestion is that, in the fundamental case, your knowledge that your premises support your conclusion is posterior to your knowledge of your conclusion. (The same is not true of your knowledge of your premises.)
So the difference between the standard view (which accepts the Taking Condition) and my view might be put like this: On the standard view, your knowledge that your premises support your conclusion is both causally and epistemically prior to your knowledge of your conclusion. On my view, by contrast, your knowledge of your conclusion is both causally and epistemically prior to your knowledge that your premises support your conclusion.
It was recently suggested to me that the following example poses a challenge to my view.
Suppose that, in fact, <q> and <r> together support <p>. And suppose that there are two people who are similar in the following respects:
(i) both of them know both that q and that r;
(ii) both of them infer <p> from <q> and <r>.
But suppose they differ in the following respect: the first knows that <q> and <r> together support <p>, but the second merely believes this. More carefully, both of them “take” <q> and <r> together to support <p>, but only the first’s “taking” amounts to knowledge.
Plausibly, the first person now knows that p, but the second does not––because she does not know (but merely believes) that her premises support her conclusion. The challenge, then, is to explain why the second person fails to know that p, without relying on the second clause of the Taking Condition, i.e., without relying on the idea that each subject’s “taking” is part of what grounds her conclusion-belief.
Now, I agree that I can’t meet this challenge. But I don’t think I need to. On my view, if (i) and (ii) are both true, the second person, despite merely believing that <q> and <r> together support <p>, does know the conclusion. She knows the conclusion because (i) she knew the premises, (ii) she inferred the conclusion from the premises, and (iii) if you infer something from something else you already know, you come to know it.
In other words, from my perspective, the challenge rests on the assumption that the second person’s “taking” is among the grounds of her conclusion-belief. And I reject that assumption. (To be sure, I haven’t here said why I reject it. But what I say below is part of that explanation.) So, from my perspective, there is no challenge.
This response is, I think, unlikely to persuade. Part of the reason (at least) has to do with the fact that, quite clearly, both subjects might plausibly “take” <q> and <r> together to support <p> before they perform the relevant inference. And that makes it odd to suppose, as I do, that those “takings” play no role in bringing about or grounding the subjects’ conclusion-beliefs. In particular, they surely could play such a role, even if they don’t always; and, when they do, the second subject surely won’t come to know her conclusion.
At this point, deep differences between my approach and the standard approach start to become relevant. I think it’s important to ask, at this point, how the first subject knows that <q> and <r> together support <p>, and why the second subject believes this (i.e., what grounds their respective beliefs). In particular, it’s important to ask: what’s the most basic way of coming to know such a thing? For our subjects might of course have learned that <q> and <r> together support <p> via testimony; but then we want to know how the testifiers came to know it.
Now, it seems tempting to say that the most fundamental way of coming to know that some propositions support another is by in some sense “seeing” that they do. But there’s the rub: how different is “seeing” that <q> and <r> together support <p> from inferring <p> from <q> and <r>? Well, there’s at least this difference: you can “see” that <q> and <r> together support <p> without coming to believe that p (for example, because you don’t believe both that q and that r). So “seeing” the support relation isn’t inferring. Granted. But still: does that make it different enough from inferring to do the work it needs to do here, in providing part of an account of inferring?
I don’t know how to make this thought fully precise, but it seems to me that “seeing” that <q> and <r> together support <p> involves the very same (causal-epistemic) connection between mental acts that’s involved in inferring. What’s different is just the mental acts. Let’s say that, in “seeing” that <q> and <r> together support <p>, you merely “entertain” the propositions that q, that r, and that p. Then the idea is that, in “seeing” what you see, your entertaining both that q and that r grounds your entertaining that p, in the same sense in which, in inferring <p> from <q> and <r>, your believing both that q and that r grounds your belief that p. If that’s right, then, whatever exactly this connection is, and whether it first comes into view in inferring or in merely entertaining, the connection itself is prior to your knowledge that q and r together support p. In other words: your performance of some kind of “inferential act” (as we might call it) grounds your knowledge that the premise-propositions support the conclusion-proposition––rather than the other way around. (Or, as I’ll put it in a moment: your conception of some kind of inferential act––and your conception of it as potentially knowledge-producing––grounds your knowledge that the premise-propositions support the conclusion-proposition.)
An important crutch for the standard view, then, is the idea that our knowledge of the logical relations between propositions is prior to our drawing of inferences. But I’m not convinced that this crutch can bear the weight it’s being asked to bear. For one thing, it’s not plausible that our knowledge of all of the evidential relations between propositions is prior to our drawing of inferences; and so the standard view is likely to run into problems when it comes time to explain material inferences (unless the plan is to deny that there are such things––but that, too, isn’t terribly plausible). More importantly, though, it’s not clear that our knowledge of even the purely (deductively) logical relations between propositions can be prior to our drawing of inferences. For that knowledge seems of necessity to derive from the performances of mental acts that are essentially inferential, even if we don’t want to treat them as full-blown inferences (because, for example, we take the latter always to begin from and conclude in belief).
To put it another way: to know that <q> and <r> together support <p> is to know that you could come to know that p by inferring it from <q> and <r>. This, however, is explain support relations between propositions in terms of inference. And that suggests that we should be dubious of the idea that inference can be explained in terms of support relations between propositions. And that provides some important motivation for the theory of inference I prefer.
(P.S. There’s more to be said about the example I discuss here, and I’ll return to it in another post.)
Boghossian, Paul. 2014. What is inference? Philosophical Studies 169: 1–18.
So would it be correct to say that on your view, the scenario in which p in fact follows from q and r, and in which someone knows that q and knows that r but merely believes rather than knows that p follows from q and r, is impossible?
That’s actually the issue I’m planning to write about in another post. The short answer is “No,” but there are at least a couple different kinds of thing that might be happening in such a case. One possibility is that the subject’s mere belief that p follows from q and r just isn’t “brought into play” when they infer p from q and r. That’s the sort of case I was mainly thinking of in writing this post.
If the belief that p follows from q and r is “brought into play,” though, things get complicated. I’m inclined to think that, in that case, we have an inference with the belief that p follows from q and r functioning as another premise-belief (where the subject might believe that p follows from q and r on, say, testimonial grounds, without herself being able to “see” the connection between q and r and p). But that’s not the inference represented by “q; r; therefore, p.” It’s the different inference represented by “q; r; p follows from q and r; therefore, p.”
I do think, though, that I’m committed to saying that, if your belief (or “taking”) that p follows from q and r plays the role of grounding your conclusion-belief, then we have an inference of the form “q; r; p follows from q and r; therefore, p,” and not an inference of the form “q; r; therefore, p.” And defenders of the Taking Condition are committed to denying that. But then they owe us an explanation (a non-circular one, I think) of the difference between inferences of these two forms. And I’m still not convinced that that’s possible, assuming the truth of the Taking Condition. But more about that in another post.
How delightfully Lewis Carroll-ish this is all getting. I’m surprised to hear the answer is no! Though I guess if it were yes, that would be an easy response to the objection.
So in the first branch of your two possibilities, I know that q, know that r, infer that p follows from q and r (as it in fact does), and believe that p follows from q and r, but don’t bring my belief into play when I draw the defective inference?
That’s right. And on my view, as a result, you then come to know that p follows from q and r. (More precisely, you come to know that you’ve come to know that p by inferring it from q and r.) So I end up accepting something very much like the first clause of the Taking Condition, i.e., that inferring necessarily involves taking your premises to support your conclusion.
I think the Lewis Carroll-isness of what I’m saying here is a big part of what makes it unpersuasive to defenders of the Taking Condition, though. Because of course they’ve thought about that problem, and have an answer! So I’m trying to find some different ways of explaining my doubts about the second clause of the Taking Condition, so that it doesn’t seem like I’m just repeating Carroll’s objection. So that’s what the stuff towards the end of this post is meant to be about.
By the way, Matt, it was your objection to the very first draft of this paper––in which I defended a version of the Taking Condition myself––that got me started down this road!
No way! I had an objection?
So this is going to sound a little trippy, but please bear with me. Are you saying that as a result of knowing that q, knowing that r, and believing but not knowing that p follows from q and r, because I’m not putting my belief into play in drawing the inference that p follows from q and r, I then come to know that p follows from q and r? If so, it kind of sounds like the scenario in which I know the premises but merely believe that the conclusion follows is impossible, because the moment it tries to be that scenario, as it were, it irrevocably turns into the scenario in which I know all three.
In a sense, that’s right. But I think it’s worth stressing that you continue to believe that p follows from q and r on the grounds you had earlier. So, for example, if you originally came to believe that p follows from q and r because I told you that it does, and because you thought I knew what I was talking about, you still believe it on those grounds. But you also have new grounds, and these will suffice for knowledge.
There are also going to be cases in which you know both that q and that r, believe (but don’t know) that p follows from q and r, and come to believe that p as a result––that is, there are going to be cases in which your belief that p follows from q and r does come into play. And when that happens, you won’t come to know that p follows from q and r. (I say more about this in the post I published today; this would be an inference of form (2), in the terms I use there.)
So, whether the thing you’re imagining is impossible, according to my view, depends on what exactly you’re imagining. There is something that’s going to be impossible, but it’s important to me that there are a number of very similar things that are possible. I emphasize this point because it sometimes looks to people like I’m just denying the possibility of something that obviously happens all the time. But, as I see it, that’s not what I’m doing at all. What I’m doing is suggesting that we need to be a lot more careful about how we describe what’s going on. And I think that, once that’s done, it’s no longer going to be obvious that what my theory says about the examples isn’t exactly the right thing to say.
Nice. Ok, onto the next post!