Peer disagreement: Why it matters and what the proper response to it is.

By request from Stillwater and Rose, here is a post on the epistemology of peer disagreement. Being the kind of lazy sod that I am in my deepest of hearts, I realised that since I had already had done a good discussion of the various views previously, I’ll just reproduce that discussion more or less intact. The context of the discussion below is me arguing that under widespread disagreement any reasonable view about the proper response to peer disagreement will require us to severely moderate our views. I argue that this would be the case even if the proper response to individual peer disagreement is not necessarily to moderate judgement.

Before we proceed with the discussion. I would like to elaborate what I mean by peer disagreement. Peer disagreement basically is what happens when two or more epistemic peers disagree. Widespread peer disagreement is wha happens when many epistemic peers disagree with many people on both sides of the issue. Defining what counts as an epistemic peer is slightly tougher as many different philosophers have used different notions of peerhood. The most stringent notion of peerhood is one in which two people count as epistemic peers only if they share the same evidence and have the same attitude towards the evidence. i.e. they draw the same kinds of connections and relations etc. A looser notion of peerhood is about two people who are equally competent when it comes to a particular question or set of questions. We can obviously see that the latter notion is wider than the former and includes the former as well. Two people could still be equally competent in a field of enquiry even if they don’t share the same evidence or don’t have the same attitude towards the evidence that they do have. Of course two people who have the same evidence and the same attitude towards it have to be equally competent. Below in the excerpt, I have examined three different views: the equal weight view, Kelly’s total evidence view and Enoch’s first-person inescapability view.

One of the key assumptions that motivates the problem of peer disagreement is the uniqueness thesis. The uniqueness thesis basically says that for any proposition P and set of evidence E, there is only one credence level C about P which is most epistemically appropriate. Obviously, if more than one credence was appropriate, then people could unproblematically reasonably disagree as long as their disagreement was limited to those appropriate credences. Of course, given the uniqueness thesis, what this still allows that people with different evidence (and priors) may permissibly have different levels of confidence in P. The uniqueness thesis also allows that evidence may be misleading. i.e. it may really be the case that P, but the body of evidence E is such that E indicates not-P. Credence levels range from 0 which indicates complete disbelief to 0.5 which is perfect agnosticism to 1 which is complete belief. So, it might be that the state of the evidence is such that we cannot say one way or another about P. In that case we should be agnostic about P. None of what I’ve said here indicates that we will know what that best credence level is, only that it exists. It may still be the case that a matter may be very difficult to assess and that some connections may be difficult to see. However, even if difficult to see, those connections are still there. The basic idea working here is that there is no fundamentally indeterminate piece of evidence. Any particular piece of evidence in its appropriate context evidences a particular conclusion by some determinate amount whether or not we can know what that amount is. Let us move on to a discussion of the various views.

In order to say that we should moderate judgement [under widespread peer disagreement], we don’t need to say that we should moderate judgement under individual peer disagreement. While that would certainly help, it is not necessary. To see why, I will try to reconstruct what the best response to the equal weight view is and try to show that it still recommends moderating our views under widespread disagreement.

Here, I will use Thomas Kelly’s total evidence view[1] as the most reasonable alternative to the equal weight view. The equal weight view is the view that when I find that someone who has exactly the same evidence that I do, and is equally as good as I am at assessing the evidence, but comes to a different conclusion about a particular proposition than I do, we should both moderate our views until we agree. Equal weight theorists suppose that this result is stable even when we weaken our notion of peerhood to one in which someone is a peer if he is as likely as I am to get the answer correct. This looser notion of peerhood says nothing about whether we both possess the same evidence or have the same attitude with regards to the evidence that we do have. If I have evidence that you don’t have and symmetrically you have evidence that I don’t, then, the fact of our disagreement is a fact about the relevance of our “personal evidence” to our relative conclusions. i.e. the fact that you arrive at a conclusion H reflects some part of the evidence that you do in fact possess and the fact that I arrived at not-H reflects some aspects of my evidence. The equal weight theorist could then argue that we should therefore suspend judgement until we have acquired the other person’s evidence and then made a re-evaluation. If, on the other hand, I have all the evidence that you have, as well as more, then I am in a better epistemic position than you and need not and indeed should not defer to you. But, in that case, I may not necessarily count you as a peer. Similarly, we could be peers and still evaluate the evidence in a different way. In that case, the flaws that I have in my reasoning need not be the flaws that are present in your reasoning. However, given that there are flaws in my reasoning (which I may at present be unable to identify), I should moderate my own judgement.

Kelly notes, however, that here is a problem with the equal weight view. The equal weight view presumes that the fact of disagreement swamps other kinds of facts. Kelly notes correctly that the fact of disagreement cannot constitute first order evidence against a proposition. Rather, the fact of disagreement constitutes a higher order evidence and instead bears on the strength and nature of our first order evidence. So, when I find that a peer disagrees with me, all that disagreement does is call into question the strength of the first order evidence with respect to the conclusion. However, this is not conclusive. It is possible that my peer could be wrong. And the mere fact that he is wrong on this one instance does not demote him from peerhood. There is nothing that prevents someone who usually very reliable from drawing unreasonable conclusions from the given evidence every once in a while. However, Kelly notes that disagreement even in such a case is still evidence even though it is misleading evidence. However, this higher order evidence need not undermine totally the confidence I have on the strength of my evidence. The reason for this is very simple. Kelly says that the views we can reasonably draw from our evidence is constrained  by what we can reasonably believe about our evidence. However, if we have anything to say about our evidence at all, surely the actual state of the evidence is to play a part in it. The degree to which disagreement ought to affect our judgement about a particular case really depends on the strength of our own first order evidence. If my evidence for the proposition I hold is indeed strong, then it follows that there is unlikely to be reasonable evidence against the conclusion and the less that peer disagreement should move me. The weaker my evidence is, the greater the possibility that someone can disagree and have reasonable evidence to back him up. There are of course some cases where our primary evidence may be sufficiently weak that we are able to conceive of sufficient counter arguments that would give us reason to completely suspend judgement. But this need not happen in every case of peer disagreement. Of course, if you find that many of your peers disagree with you, then, the likelihood of your initial evidential assessment being correct decreases, or at least it does so in most cases. The question of whether there are any exceptions will be explored later. Nevertheless, in most cases, whenever you find that you are in a distinct minority among your peers, it is often the case that you should moderate your views significantly.

The result above may not work for other alternatives to the equal weight view. I will briefly take a small detour to look at David Enoch’s view and try to show why it is if not unreasonable, far more problematic than Kelly’s view. Roughly speaking, if Enoch’s view is right, then even under widespread disagreement, people need not significantly moderate their views[2]. Enoch’s argument, roughly, is that the equal weight view relies on treating the dissenting parties as truth measuring devices or truthometers. However, Enoch argues that it is impossible to treat ourselves as truthometers since we cannot escape the first person perspective. Enoch says that given that ought implies can, we need not treat ourselves as truthometers. Enoch draws a distinction between two types of responses to impossible goals. One type of response is to discard the goal as anything inherently worthy of pursuit. The other is to pursue the goal and try to approach the ideal as far as we can. He supposes that the issue of treating ourselves as truthometers fits the former approach. However he does not argue for this. The lack of any argument for why we should not simply try to treat ourselves as truthometers as far as possible implies that at the least we should be agnostic about whether to treat ourselves as truthometers as far as we can. However, the initial reasons for treating ourselves as truthometers presumably still exist: The endorsement of a proposition P by an extremely reliable epistemic agent is indicative of the truth of P and the similar rejection of q is likewise indicative of q’s falsity. Enoch’s reply to the equal weight view, therefore, fails.

Where Enoch goes wrong is in his embrace of epistemic permissiveness. Despite his initial setup where he sets aside epistemic permissiveness and endorses the uniqueness thesis for the purposes of the paper, his argument amounts to a rejection of positive justification[3] as a goal on the grounds of impossibility. The distinction that Enoch may be failing to appreciate is that whereas it is understandable that people will use the reasons that they do possess, they are not necessarily justified in doing so. So, even if Enoch criticises Kelly’s view because it doesn’t give people who face peer-disagreement advice on what to do, this is not a problem for Kelly’s view. That’s because Kelly’s view is not a decision procedure, but an account of epistemic rightness. The above remarks are admittedly insufficient to do full justice to Enoch’s argument which is more complicated than I have detailed. Nevertheless, the above remarks are intended to suggest very roughly that Enoch’s thesis is at best incomplete and that Kelly’s views are the best way to go.

Leaving aside Enoch’s view, let us take for granted that Kelly’s view is mostly correct and return to the possible counter example to Kelly’s Total Evidence view. Richard Fumerton raises a possible counter example to moderating your beliefs when confronted with peer disagreement[4]. Fumerton brings up the case of the Monty hall problem.[5] Fumerton says that prior to explanation, even though most people would disagree with him as to what the optimal strategy is, he says that he should stick to his guns even if most of his colleagues whom he considers to be peers disagree with him. The reason for this is because he knows that he is privy to an argument which he knows that he has assessed correctly, which if he were to tell his colleagues would also convince them. There is an intuitive plausibility about Fumerton’s case, and could provide a possible counter-example to Kelly’s view if Kelly’s view could not accommodate the intuition. Fortunately, this is not the case. Note that in Fumerton’s case, the first order evidence is so strong that it is just inconceivable that anyone, once presented with the evidence could reasonably disagree.

The possible counter-example is relevant for a very important reason. Kelly’s argument pre-supposes that evidence is shared. However, the type of peerhood we pre-supposed is not as strong as Kelly’s. It does not require evidence to be shared. However, the strength of Kelly’s view does not rely on the actual existence of peer disagreement, but rather on the possibility that someone could reasonably provide a counter argument/ evidence against your position. According to Kelly, the real threat comes from the possibility of reasonable disagreement. The possibility of reasonable disagreement indicates that the evidence may not be as strong as initially thought. The presence of actual peer disagreement is just dispositive about the possibility of such disagreement existing.

Thus far, we have shown that Kelly’s view is better than Enoch’s view. However, we have yet to show that Kelly’s view is better than the equal weight view. To recap, the equal weight view says that the fact of peer disagreement necessarily swamps other kinds of evidence. Enoch’s view says that our first person perspective necessarily swamps the fact of peer disagreement. Kelly’s view basically is that which kind of evidence dominates depends on the relative strength of each kind of evidence. If the initial primary evidence is very weak, then the fact of peer disagreement is evidence that it is weak and should appropriately weaken our confidence in that assertion. Kelly’s point is not that epistemic peers can reasonably disagree, rather it is that the situation is asymmetric. The person who has worse first order evidence should moderate his judgement more than the person with better first order evidence. This does not necessarily yield a practical guide as it may be difficult in practice to assess the strength of one’s first order evidence. However, that is not a problem for the view. It is not a practical guide, just an account of what is the epistemically right thing to do.

The problem with the equal weight view is that it cannot really account for why the fact of disagreement swamps the other evidence. The equal weight theorist argues that to still consider our initial evidence is to double count it. This is not true. To make the charge of double counting stick, it must be the case that the conclusion that your peer draws from the evidence is just as reasonable as the conclusion that you draw. Of course, this is not necessarily true. Just because  someone is your peer and is generally speaking as reasonable and as competent as you are, it does not follow that on any one particular case, his conclusions from the evidence are as reasonable as yours are. It might be that  his are more reasonable, or that yours are, or that neither yours nor your peer’s are. Of course as more and more peers disagree with you, the possibility that all of them are being less reasonable than you on this one case decreases.

Is the question of peer disagreement anything more than an abstract problem? Certainly! First of all, settling the question of the proper response to peer disagreement helps us answer the question of how to respond to disagreement with people who are experts and with people who are relative ignoramuses in a particular field. Another question has to do with how to deal with disagreement in a number of investigative communities like juries, electorates, scientific communities and philosophical communities. For example there seems to be a conundrum raised in the context of voting in a democracy.