Talking to a friend in the history department recently, she asked me what my research process is. Do I use Obsidian? Or Scrivener? Or some elaborate system of index cards like Mr Casaubon in Middlemarch? Which made me realise that I have no system—or at least the haphazard series of disjointed activities that I do have doesn’t deserve to be called a system. I download papers, perhaps read them, leave them on my desktop until it looks too cluttered to be tolerable, at which point I delete them to quiet the discomfort their messiness causes me. Along the way, I think of what I want to say about the topic, sometimes jot down something cryptic on a pad of paper I have a high chance of losing, and finally write it up longhand on paper first then type it up before going through a hundred redrafts and submitting. But I thought perhaps I need to do something better than this; in particular, perhaps I need to properly record my reactions to the papers I read. I’m not going to blog about every single paper I read, nor even every book. But for the more substantial ones, I thought part of my process could be writing up thoughts for a blog, and that’s what you find here: my reactions upon reading Jennifer Lackey’s rich and important ‘Credibility and the Distribution of Epistemic Goods’.
I discern three main theses in this paper:
(1) contra Fricker, the form of epistemic injustice known as testimonial injustice can arise just as much from credibility excesses as from credibility deficits;
(2) contra Fricker, what is required of us is not that we match our credibility judgment of a person’s testimony to the evidence we have that they are offering the truth, but that we should match it to the evidence we have and should have that they are offering the truth;
(3) contra Fricker, the epistemic good of credibility is sometimes a finite good, so that giving much of it to one person requires you to give less of it to others.
Together, Lackey takes these to show that, contra Fricker, much epistemic justice is a question of distributive justice, where the good to be distributed is credibility.
Lackey’s examples in favour of (1) are extremely compelling. First, she considers a case in which you match your credibility judgment about woman’s answer to a scientific question exactly to the evidence she’s offering the truth, in line with Fricker’s central norm, but you set your credibility judgment about a man’s answer to the same question far above the evidence he’s offering the truth. In such a case, no-one suffers a credibility deficit, and yet an epistemic injustice has occurred, and that must therefore be attributed to the credibility excess given to the man. Second, she considers a case in which you match your credibility judgment about a young black man’s answer to a question about drugs far above the evidence he’s offering the truth, and you do so because you think nearly all young black man are extremely knowledgeable about drugs. Again, there is only a credibility excess, and yet there is an epistemic injustice.
The argument for (2) is equally strong. If prejudice or bias or bigotry of any form can lead you to respond poorly to the evidence you have, either by judging one person less credible than your evidence suggests or judging another more credible, then it can also lead you to ignore or fail to gather evidence you should gather. If I believe no woman can be a serious scientist, then I will not gather the available evidence about a particular woman’s scientific credentials and track record of expertise, simply assuming that would be a waste of time. Because of that, I might match my credibility judgment to the evidence I actually have that she’s offering the truth, while nonetheless treating her unjustly from an epistemic point of view, because I should have gathered the evidence about her expertise and, if I had, I would have judged her credibility much higher. There’s an interesting question about when someone should gather evidence, and I discuss it at length here. For instance, is it only evidence they didn’t gather because of prejudice that we must add, or all evidence they should have gathered but didn’t, such as evidence they didn’t gather because they were distracted, or because they incorrectly calculated it wouldn’t be worth the effort to gather it, and so on. But, however we answer that, Lackey is surely right that there are clear cases of this.
I’m less convinced by the argument for (3). Lackey’s argument is based on cases in which testimony clashes: perhaps a victim accuses a perpetrator and the perpetrator denies it; or a person confesses to a crime and then retracts their confession. Of the latter case, Lackey says: “To give credibility to the confessing self is ipso facto to deny it to the recanting self. Credibility becomes scarce.” So, she concludes, credibility is a finite resource and so its fair distribution is a matter of distributive justice. But I don’t think it’s right to describe these cases in these terms, and I think that viewing them as cases of distributive justice could potentially be risky. In the end, I think Fricker is right to say that credibility is a good that wears its proper distribution “on its sleeve”.
At the risk of sounding like a broken record, I think this is a case where it’s helpful to frame the discussion in terms of credences or subjective probabilities, rather than beliefs. Take the case of recanting confession. Start with the original confession. What is our credibility judgment about it? Fricker’s proposed norm, which tells you to match your credibility judgment to the evidence that the person is offering the truth, sounds very much like the Bayesian norm that your credence that they are guilty should be the evidential probability that they are guilty conditional on (i) the fact of their confession and (ii) your background evidence (and, as Lackey says, perhaps also the evidence you should have collected but didn’t). That is, it should be P(Guilty | Confession of Guilt & Background Evidence). And similarly with the retraction. Suppose you only hear them say they are innocent–you were out the room for the initial confession. Then your credence that they are guilty should be the evidential probability they are guilty conditional on the fact of their plea of innocence and your background evidence, i.e., P(Guilty | Plea of Innocence & Background Evidence). Now, Lackey says that, when the person confesses and then retracts, credibility becomes finite and whatever credibility you give to their earlier self and their confession of guilt you must withhold from their later self and their plea of innocence. But I’m not sure that’s quite right. It might be that P(Guilty | Confession of Guilt & Evidence) and P(Guilt | Plea of Innocence & Evidence) are both very high, because this person is generally trustworthy. Now you learn that they’ve actually both confessed to guilt and plead innocent, retracting their earlier confession. Those probabilities, and so your credibility judgments of the earlier and later self don’t change, or don’t change much. But your new credence in their guilt should be its evidential probability conditional on both confession and plea, i.e., P(Guilty | Confession of Guilt & Plea of Innocence & Evidence). Now, this can be anything from very low to very high depending on the background evidence: perhaps you have extremely strong evidence that very trustworthy individuals like this one are most likely to tell the truth the second time, i.e. in the plea, or perhaps you have evidence they’re more likely to tell it the first time, i.e. in the confession, or perhaps you know it’s pretty 50-50, and so your conditional probability should be around 50%. But the important point is that, whatever you should do is simply set by the evidence you have; and it’s set in exactly the way Fricker suggests. There are no considerations of distributive justice beyond this. And it’s not that you come to think of either earlier or later self as less credible. You might continue to consider both extremely reliable testifiers. But when two extremely reliable testifiers say contradictory things, you must take that into account when you set your credences, and to do that you just do whatever the evidence says you should.
Lackey says that, in the end, you must believe either the earlier confessing self or the later pleading self. But that isn’t true. You might end up 50-50 between them, or you might end up thinking guilt very slightly more likely given the confession-plus-retraction, but certainly not strong enough to believe in guilt. So, very often, if such cases of clashing testimony, we simply believe neither, but rather have some credences in the proposition that don’t rise to the level of belief. Now, of course, in the end, given the nature of many criminal justice systems, you must come down on one side or the other, either declaring ‘guilty’ or ‘not guilty’ (unless you’re in Scotland, where you can declare ‘not proven’!). But a verdict of ‘not guilty’ does not mean that you believe the person innocent. It just means that your evidence is insufficient to declare them guilty.
In some ways, I think this is just a nitpicky point, but I also think that considering these clashes of testimony in terms of distributive justice does run a risk: you risk thinking that, while Fricker’s norm governs cases of individual testimony and credibility judgments about them, it simply falls silent in cases of clashing testimony, and so where it says nothing you should try to distribute credibility equally, or in some other way that counts as a fair distribution. But that’s wrong. You should distribute it exactly as your evidence suggests conditional on the clash itself. So, when we hear both confession and retraction, we shouldn’t try to give equal credibility to both earlier and later selves on the grounds of distributive fairness. We should give credibility to their guilt exactly in line with how likely it is to be true conditional on the fact that they confessed and retracted and our background evidence.
Richard Pettigrew's homepage 