Reading on Evidence?

I plan to think more about the nature of evidence and its role in epistemology. But I don’t know enough about where to start looking in the literature. Really, I only know three kinds of things.

First, there is formal work from Bayesian epistemology, and especially confirmation theory. But I’ve always found that work quite disappointing in terms of its foundations. The issue of just what the E is that goes into Pr(H | E) is never satisfactorily addressed. When you press people informally on it, they seem to fall back on an unexamined version of the phenomenological theory of evidence. This obviously isn’t part of the theory – Williamson has shown how to do Bayesian epistemology with an externalist conception of evidence – but it is what most practitioners seem to assume.

Second, there is work around the debates about evidentialism as promoted by Conee and Feldman. This is all very interesting, though the focus there was more on what evidence does than what evidence is.

Third, there are the debates started by Williamson’s defence of E=K, including in the recent volume of papers on Williamson. This has led to lots of interesting discussions, such as Clayton Littlejohn’s idea that evidence is non-inferential knowledge. I have a couple of contributions to this, an unpublished paper arguing that inductive knowledge isn’t evidence, and a small aside in Deontology and Descartes’ Demon suggesting that, for creatures like us, evidence is knowledge gained directly from a perceptual module. This seems to be an interesting, if young, field.

But there must be other stuff out there. What should I be reading?

13 Replies to “Reading on Evidence?”

  1. I have the same overall impression (and disappointment) about the current state of the literature on this general issue, Brian. But, on the “E=K”-inspired debates (and some of their connections to broadly Bayesian perspectives on evidence), I found Ram Neta’s recent BJPS paper “What Evidence Do You Have” an interesting read.

  2. Hey Brian,

    Reading through the Williamson volume, I was surprised how confident people were propositions didn’t have to be true to constitute evidence. I think you can get a nice argument for the truth requirement on evidence by thinking about the factivity of explanation and inference to best explanation. Sure, they don’t have to be true to be considered evidence, but I think W is on pretty strong ground on that one. I also think that you’re right that the conclusion of an inductive inference shouldn’t be included in your evidence. Suppose I know p on inductive grounds and deduce q because I know q is a deductive consequence of p. I don’t think that ~q is inconsistent with my evidence at any stage of the reasoning although it is inconsistent with what I know at the end of that reasoning. That’s one reason I like the restriction to something non-inferential, inference doesn’t obviously seem to me to be a process of evidence acquisition but rather a way of working from the evidence to something else. (Ordinarily I’d say that this is a minor point, but it seems to matter in defending Williamson from some of his critics who attack his truth requirement.)

    Thanks for the plug. I certainly think that non-inferential knowledge of p’s truth suffices for p’s inclusion in your evidence, but I’m worried about the necessity claim because of Gettier cases.

    Two possible alternatives to avoid Gettier cases:

    IJTE: S’s evidence includes p iff p and S is non-inferentially justified in believing p.
    IJE: S’s evidence includes p iff S is non-inferentially justified in believing p.

    Since I think that the falsity of p will give you reason not to include p in deliberation (particularly if the belief that p is playing the foundational belief role), I tend to think that the IJE and IJTE come to the same thing. BIV cases worry me a bit because my intuitions go wonky when I think about cases of veridical hallucination. On some days, that pushes me back to the evidence is non-inferential knowledge idea. On others, I want to say that p’s truth is what’s required for p to be evidence (for something or other) whereas being non-inferentially justified in believing p is what it takes to have the evidence in the sense of having the right to reason from the assumption that p. Then, I just bite the bullet and say that BIV veridical hallucination cases don’t really threaten IJTE.

    Dylan Dodd and Nico Silins got me interested in this topic as they were the first authors I saw really challenge Williamson. (See Dodd’s “Why Williamson Should Be a Sceptic” in PQ and Silins’ “Deception and Evidence” in Phil Perspectives.) Prior to reading their work, I was happy to just work from the assumption that E=K. (I have a piece coming out in Synthese that addresses one of Silins’ objections to externalism about evidence.)

    There’s also a new piece by Conee and Feldman in the Q. Smith edited collection, but I still cannot quite tell what their view is. I know that they are mentalists, but there are passages that seem sympathetic to the idea that non-inferential knowledge that p is true suffices for p’s inclusion in your evidence, passages where they object to McDowell in ways that suggest they think evidence has to consist of truths, and Feldman seems to think that the scope of non-inferential knowledge is not limited to propositions about your own mental life. These claims in combination entail that mentalism is false. So, like you, I think I know what C&F think evidence does but I can’t quite tell what they think it is.

  3. Mark Schroeder has an interesting paper called “Having reasons” that touches on some of this stuff. It’s mostly on practical rationality, but at the end he draws an interesting analogy to the perceptual case. In short, he thinks that to have a reason is to stand in a certain relation (believing) to a proposition, but a proposition cannot be a reason unless it is true. He then claims that items of evidence are reasons for belief, so: having perceptual evidence is a matter of standing in a certain relation (perception) to a proposition, but a proposition cannot be evidence unless it is true.

  4. Mark Schroeder has an interesting paper called “Having reasons” that touches on some of this stuff. It’s mostly on practical rationality, but at the end he draws an interesting analogy to the perceptual case. In short, he thinks that to have a reason is to stand in a certain relation (believing) to a proposition, but a proposition cannot be a reason unless it is true. He then claims that items of evidence are reasons for belief, so: having perceptual evidence is a matter of standing in a certain relation (perception) to a proposition, but a proposition cannot be evidence unless it is true.

    This isn’t exactly Schroeder’s view. He thinks that there are two types of reasons relations—the objective and the subjective. Objective reasons have to be true, but subjective reasons needn’t be (they just have to have the following property: if they were true, they would be objective reasons). Moreover, he thinks that one can have a subjective reason if one has a presentational attitude with that proposition as its content. He thus denies that you have to stand in some positive justificatory relationship (e.g. knowing) in order to have a (subjective) reason. He thinks that justification is a function of subjective reasons—i.e. he thinks that justification supervenes on subjective reasons. He thinks that dropping both the factivity condition on having (subjective) reasons and the high bar allows one to give a simpler, more explanatory unified account of perceptual justification. But it is very important for this story that subjective reasons needn’t be true. (Also, I should point out that he is neutral about some of the story I just told in “Having Reasons.” I take some of this story from the paper I mention below and some of his other work (Slaves of the Passions, for example).

    He has another paper forthcoming in a volume about reasons for belief that further draws the lessons out in epistemology. It’s called “What Does it take to ‘Have’ a Reason,” and it’s available on his website. (Full disclosure; I have a paper coming out in Phil Studies that is a response to his original “Having Reasons.” It’s called “Having Reasons and The Factoring Account.”)

  5. One of my grad students, Jeff Dunn, has a large chunk of his thesis—now virtually finished—devoted to this very issue. He’s done a lot of interesting work on this. You should write him to see what he’s doing: (Braden: you may remember talking with Jeff when we went out for beers at Northwestern.)

  6. For an overview of the nature of evidence you might look at the probability literature, which is not limited to a Bayesian approach. The best one volume work in this area is Achinstein’s The Book of Evidence (2001). You might also look at the debate between Maher (1996) and Achinstein (1996). To some degree the non-inferential view about evidence aligns with Maher’s view about evidence since it requires, as characterized by Achinstein, “(i) that e must be known to be true, (ii) that e must be known to be true without calculation, and (iii) that e must be known to be true without being inferred from any other proposition under consideration” (Achinstein 1996: 180). From what I can tell, however, your view is not like Maher’s in that it is an externalist view of evidence. It must be the case that e is true, but it need not be the case that e is known to be true.

    Some neglected works on evidence and epistemology are N.M.L Nathan’s Evidence and Assurance (1980) and Moser’s Knowledge and Evidence (1989). Haack’s Evidence and Inquiry (1993) is typically only mentioned in relation to structural issues in epistemology, but in formulating Founherentism Haack discussed things about evidence that it might be worth looking at. Haack talks about the distinction between evidential and non-evidential components of belief (pp. 76-77).

    Those are a few references off the top of my head.

  7. Thanks Christopher, those references are very helpful.

    I actually do think that p must be known to be evidence. But I don’t know why having a truth requirement would make the view more externalist than Maher’s. After all, any view which puts a factive requirement on evidence will have an externalist component.

  8. You’re welcome Brian. Your comment highlights some of the translation that must occur when looking at the probability literature on evidence. Perhaps that is why there has not been effective communication between the epistemology and probability literature on evidence. The distinctions at play are:

    *Probability: Subjective vs. Objective
    *Epistemology: Internalist vs. Externalist

    You’re right that a factive account will have an externalist component. I was talking across the distinctions above, and if ‘Subjective’ does not equal ‘Internal’ and ‘Objective’ does not equal ‘External’, then there is some conflation of terms in my comment. Your position now seems to be aligned with a Subjective/Externalist rendering of evidence. It is subjective in that p must be known to be true by some individual or community (note: Achinstein subscribes to an objective view of evidence in that evidence is an impersonal concept, i.e., p need not be known to be true by any individual or community). Most epistemologists endorse a subjective notion of evidence because evidence is spoken of in relation to a subject or a community of subjects. The prototypical question is whether, “S’s evidence includes e.” Your view is also externalist because of the factivity aspect. This is where Maher requires luminosity: S must know that p is true, i.e., the truth of p must be luminous to S—often as the result of empirical observation that provides non-inferential knowledge. However, this is where Achinstein recognizes the opacity of evidence. That is, p must in fact be true to count as evidence regardless of whether S knows that S knows that p is true. Things get a little tricky though. This is because Achinstein adopts an internalist constraint on evidence in addition to the externalist constraint: p must be true and S must believe that p is true (but S is not required to know that p is true, in terms of being able to access the truth of p).

    So, in looking at some of the details, it seems that your view aligns closer with Maher’s view even though there is some overlap with Achinstein’s view. Aligning with Maher exposes your view, and Littlejohn’s non-inferential view, to the objections raised by Achinstein to Maher’s view. It seems you could get some mileage out of looking at the exchange between these philosophers.

    Note: You might also pickup on the need to sync the term ‘know’ across the domains. Probability theorists and epistemologists have different levels of stringency in the application of the term (probability theorists being more lax and epistemologists being more stringent for obvious reasons).

Leave a Reply