“Nobody ever thinks they're stupid. It's part of the stupidity.”
—Detective Ray Cole, The WireIt is frustrating to discover that you disagree with other equally smart people. Most epistemologists now agree that usually upon discovering this, something should give (Lackey and Christensen Reference Lackey and Christensen2013). Often what should give is your confidence in the view under dispute. If you discover that your epistemic peer disagrees with you about how to divide a restaurant check, you would be foolish to remain exactly as confident as you had been. Sometimes, however, it seems wiser to revise your view about whether your disputant is indeed your equal. Discovering that she believes something that strikes you as odd can lead you to question whether she really is as epistemically savvy as you had thought. Charity and etiquette may require you to presume that others are your equals, but lots of evidence to the contrary can displace such niceties.
In the face of peer disagreement, then, you might revise either your view about the matter under dispute, or your view about whether your disputant is really your epistemic peer (or a bit of each, or neither). But which of these options is best? If you have lots of evidence that your interlocutor is your epistemic peer, and only a little evidence for the matter under dispute, then it appears you would be wiser to revise your belief in the latter. But if you have only a little evidence that your interlocutor is your epistemic peer, and a lot of evidence for the matter under dispute, then it appears that you would be wiser to update your view about her comparative epistemic competence. Your initial assessment of her comparative epistemic competence is not immune to revision (cf. Worsnip Reference Worsnip2014). Thinking otherwise would itself betray remarkable hubris.
So the rational response to peer disagreement seems to depend in part upon how to evaluate the epistemic competence of others. This is easier said than done. Although others are not completely mysterious, you may not know your disputant well enough to form a very accurate or stable opinion about her epistemic virtues and vices.Footnote 1 So it is often far from obvious whether to revise your opinion or to downgrade your disputant.
Unfortunately, judging whether someone else is your epistemic peer involves more than merely accurately assessing her epistemic competence. It also involves accurately assessing your own epistemic competence. And recent psychological research confirms what Socrates, Bertrand Russell and others have surmised: when comparing herself to others, the typical person is quite bad at determining how epistemically competent she is. Most people seriously overestimate their own cognitive abilities.
Here I argue that these empirical findings have radical implications for the epistemology of peer disagreement.Footnote 2 I conclude that much of the time when you are disagreeing with someone you regard as your epistemic peer, you should not merely move halfway to her judgment, as The Equal Weight View has it (Elga Reference Elga2007). That is not conciliatory enough. Surprisingly often, you should be at least weakly confident that you are wrong, and that your disputant is right.
1. SETTING UP THE QUESTION
The vast literature on peer disagreement contains different ways of setting up the problem. Here I will explain what question I hope to answer.
First, a few words about epistemic peerhood. Some understand an epistemic peer to be someone who has both the same intellectual characteristics and the same evidence you have. Others understand it to be someone whose intellectual characteristics and whose evidence are, though perhaps not the same, equally good. Here ‘equally good’ means equally likely to lead to an accurate opinion about the matter under dispute. Two people might not have the same characteristics and evidence as each other, but still may be epistemic peers because their characteristics and evidence are equally likely to lead to an accurate opinion. Still others understand an epistemic peer to be someone whose intellectual characteristics and whose evidence is (merely) roughly equivalent to your own. And Kelly (Reference Kelly, Goldman and Whitcomb2010) takes a peer to be anyone who shares the same evidence and for whom you lack a reason for thinking that either of you is better advantaged in evaluating this evidence, making it possible for you to be peers with people whose characteristics you know nothing about. All of these ways attempt to explain what it is to be an epistemic peer.Footnote 3
Unless otherwise specified, I'll operate with the ‘equally good’ conception of peerhood. Two people are epistemic peers just in case their intellectual characteristics and evidence are equally good. Working with this notion, we can phrase the question of peer disagreement in several different ways. We might ask:
(OBJECTIVE) Given that S is in fact your epistemic peer (whether you are aware of this or not), what should you believe when you discover that you disagree with S?
(EVIDENCE) Given that all the available possessed evidence indicates that S is your epistemic peer, what should you believe when you discover that you disagree with S?
(SUBJECTIVE) Given that you believe that S is your epistemic peer, what should you believe when you discover that you disagree with S?
Here I attempt only to answer SUBJECTIVE. Here's why. When you disagree about whether p with another person who you think is your epistemic peer, you have two relevant opinions: (1) change your mind about whether p, and/or (2) change your mind about whether she is your epistemic peer. It would be odd to just assume that you know or are responding reasonably to the available evidence about one of these two matters, but not the other. More specifically, except in rare special cases, it would be odd to doubt your opinion about the matter under dispute, and yet be completely confident that your disputant is indeed your peer. Rather, you should note the fact that you have both opinions, and that for either you might not be responding appropriately to the evidence available. Perhaps your evidence doesn't support p. Perhaps your evidence doesn't support the claim that your disputant is your epistemic peer. It's possible that either of your opinions is in some way unreasonable in light of the evidence available to you. Asking SUBJECTIVE gives us the room to consider both of these possibilities. Asking OBJECTIVE or EVIDENCE does not.
So, suppose you believe that you and S are epistemic peers, and you discover that you and S disagree about whether p. Should you continue believing p, should you believe not-p, or should you suspend judgment about whether p? Or perhaps put better: should you maintain confidence in p, or should you reduce (perhaps maximally) confidence in p?
As mentioned already, discovering that you disagree with another person can seem to put pressure on your belief about the matter under dispute. It can also put pressure on your belief that your disputant is indeed your equal. We might assume that it puts pressure in only one direction: that of downgrading your disputant. But perhaps discovering that you disagree with her instead puts some pressure on you to downgrade yourself?
2. THE DUNNING–KRUGER EFFECT
Psychologists have found that people typically overestimate their own skills and competences (Adams and Adams Reference Adams and Adams1961; Brown Reference Brown1986). Most believe that they are above average drivers (McCormick et al. Reference McCormick, Walkey and Green1986). Most believe that they are better able to detect lies than others can (DePaulo et al. Reference DePaulo, Charlton, Cooper, Lindsay and Muhlenbruck1997). Most people believe that they have above average leadership skills. Most people believe that they are in the top 10% at “getting along with others” (College Board 1976–1977). Most motorcyclists believe that they are less likely to get into an accident than the average cyclist is (Rutter et al. Reference Rutter, Quine and Albery1998). And perhaps most troubling of all, 94% of college professors report that they do above average work (Cross Reference Cross1977).
Many scholars have replicated and broadened these findings. Few people like to believe bad things about themselves. People tend to suffer from various self-serving biases. Thus people tend to attend to evidence that supports a favorable self-image, and discount evidence to the contrary.Footnote 4 Most of us think that we are better than we really are (Dunning et al. Reference Dunning, Johnson, Ehrlinger and Kruger2003). People “tend to hold overinflated views of their skills that cannot be justified by their objective performance” (Dunning et al. Reference Dunning, Johnson, Ehrlinger and Kruger2003). If you think that you are epistemically savvier than most, you are not alone.
Just how bad are we at estimating our own epistemic competence? To take just one example, “[s]tudies show that people's views of their own intelligence correlate about only .2 to .3 with their performance on intelligence tests and other academic tasks” (Hansford and Hattie Reference Hansford and Hattie1982). This correlation is only slightly better than random. This is representative: in one meta-analysis of studies of the relationship between self-perception and performance, researchers found that the self-perception correlated with objective performance on average at 0.29 (Mabe and West Reference Mabe and West1982).
Thus when evaluating whether another person is your epistemic peer, you need to judge both how competent she is and how competent you are.Footnote 5 If you get them both right, or, even more remarkably, get them both wrong but in exactly the same way and to the same degree, your judgment about whether she is your peer will be spot on. Otherwise, your assessment of your comparative epistemic competence will be off, and the psychological evidence cited above suggests that you are more likely to think you are her peer when she is in fact your epistemic superior.
How should you react to these findings? Perhaps the wise thing to do is dial back your confidence in your judgments about whether you are indeed someone else's epistemic peer, allowing for the possibility that you are overestimating how savvy you are. And so you might be tempted to think:
Aware of the evidence of these self-serving biases, I realize I need to correct my evaluation of my own epistemic competence in order to counteract the bias from which I, like others, undoubtedly suffer. If all my other evidence seems to support that my epistemic competence is at, say, the 80th percentile, I'll dial that down so that I judge that I (and thus my epistemic peers) are actually at, say, the 60th percentile, which is probably close to the truth. Pulling this off would be impressive. It is not easy to accurately counteract self-serving biases. But I suppose some people are indeed better at accurately counteracting self-serving biases than others are. In fact, I would bet that I am one of those people. That is, I suppose that I am better at accurately counteracting self-serving biases than most other people are!
The flaw in that bit of reasoning is easy for a detached observer to spot. The ability to accurately counteract various self-serving biases is itself a desirable quality. And so the average person is likely inclined to think that she has more of this desirable quality than do most others. The average person thinks she is above average about this too. She may believe that her keen awareness of self-serving biases puts her in a good position to neutralize it. And so the average person will ironically overestimate the degree to which she can accurately offset her self-serving biases. But that is just another manifestation of bias itself (Friedrich Reference Friedrich1996).
Furthermore, those who are indeed incompetent in a particular area suffer from a double whammy. It appears that the more incompetent you are at something, the more incompetent you are at judging your own competence at that very thing. The very skills the possession of which enables you to perform well are often also the skills that enable you to evaluate your own competence accurately. This phenomenon has come to be known as the Dunning–Kruger effect (Kruger and Dunning Reference Kruger and Dunning1999; Dunning et al. Reference Dunning, Heath and Suls2004; Dunning Reference Dunning2005; Ehrlinger et al. Reference Ehrlinger, Johnson, Banner, Dunning and Kruger2008).
The Dunning–Kruger effect is thus strongest for those who are least competent. The least competent tend to grossly overestimate their own competence, while average people tend to more mildly overestimate their own competence. The most competent, by contrast, tend to have a much more accurate view of their own competence, although here we still find individuals who suffer either from impostor syndrome (thinking that they aren't as excellent as they really are) or megalomania (thinking that they are not merely excellent but the very best, even better than they in fact are).Footnote 6
Are philosophers, who may have some independent reason to think that they are above average epistemic agents, thus largely immune to the Dunning–Kruger effect, because they probably lie at the upper end of the distribution of epistemic competence? Perhaps for some disputes, they are. If a philosopher judges that she and a particular nonphilosopher are peers with respect to, say, evaluating the validity of arguments, then she probably has no reason to think she is overestimating her own competence at evaluating such matters. Evaluating the validity of arguments is in the philosopher's wheelhouse. But if she disagrees with another philosopher about an argument's validity, or she disagrees with a nonphilosopher about, say, whether the Yankees will win the World Series, then she is not entitled to think she is largely immune to the Dunning–Kruger effect. Knowing that you are expert about one thing does not make you immune to grossly overestimating your expertise about other things. So philosophers should not believe that the Dunning–Kruger effect fails to apply to them.
Of course, nearly no one is in a position to determine how competent she is by first accurately categorizing herself as an overestimator, an underestimator, or an accurate estimator of her own epistemic competence, and then using the findings of Dunning and Kruger to assess if and how much she should adjust this estimate. We have little independent access to how inaccurate our own assessments of our epistemic competence are, at least outside of special cases (more below), and so we typically do not have a better grasp on whether we are more or less accurate than most people are.
Self-serving biases are thus troublesome. It is probably mistaken to think that while many other people grossly overestimate their own epistemic competence, one is oneself mostly innocent of such errors. It is likewise mistaken to think that at least one can figure out just how strong one's bias is, and then somehow manage to correct for it. But the Dunning–Kruger effect reveals that many people think they are much better than they indeed are, making it highly unlikely that they can figure out just how badly off they are to begin with. So, the thought that only other people are very guilty of the self-serving bias to any significant degree is not uncommon. Many people wrongly think they are immune to these problems; if you think that you are immune, you are in plentiful company. You might be right, but your confidence is poor evidence that you are.
How often do people misgauge their susceptibility to self-serving biases? In one notable study, only 24% of respondents who had just learned about the self-serving bias would acknowledge that this bias was true of themselves (Pronin et al. Reference Pronin, Lin and Ross2002)! The authors conclude:
These data suggest that even the immediate experience of having displayed a particular bias, and then being given an explicit description of it (one which presents the bias as a common human tendency), was insufficient to prompt confessions of susceptibility equal to that of one's peers. (375)
The results of our three studies suggest that knowledge of particular biases in human judgment and inference, and the ability to recognize the impact of those biases on others, neither prevents one from succumbing nor makes one aware of having done so. (378)
Most perversely of all, there is even some evidence for the claim that being more familiar with the nature of self-serving biases exacerbates its effects in one's own case (377). If right, learning about self-serving biases is likely to corrupt your competence at judging your own comparative competence.
So to think that the self-serving biases in general, and the Dunning–Kruger effect in particular, affect only other people is likely to be a manifestation of these very biases. Most of us have a bias blind spot that we are unable to eliminate by reasoning (Ballantyne Reference Ballantyne2015). It is very difficult to know oneself; moreover, it is difficult to know how ignorant of oneself one is. At least, there is much psychological evidence that suggests as much.
3. “PEER” DISAGREEMENT
What, if anything, does this evidence imply about what you should believe, when you believe you are disagreeing with an epistemic peer? In this section, I answer this in three different stages. First, I will answer it roughly and informally. Second, I will answer it a bit more carefully and formally. Finally, in the next section, I'll refine the answer in light of some natural objections.
Suppose you and your Department chair disagree about whether Bush will win the next election.Footnote 7 You believe he will. Your chair believes he will not. Prior to this dispute, suppose you believed your chair to be your epistemic peer with respect to such electoral issues. When you disagree with someone whom you believe to be your epistemic peer, you fallibly believe both (a) something about the matter under dispute, and (b) that your disputant is your epistemic peer about such matters. You might be wrong about the matter under dispute. You might be wrong about whether your disputant is indeed your epistemic peer. But upon discovering that your chair disagrees with you about whether Bush will win, you may feel tempted to downgrade your chair.
However, the psychological evidence about self-serving biases implies that you very well may be overestimating your own epistemic competence in this case. Among the many domains about which we are fallible, we are particularly bad at identifying genuine epistemic peers. Moreover, our error tends to be in one direction: that of overestimating our own epistemic competence as compared to others. If you are overestimating your own epistemic competence as compared to others, then when you think that you are disagreeing with an epistemic peer, there is a decent chance that you are disagreeing with someone who is really an epistemic superior. And if your view differs from an epistemic superior's, then it is highly unlikely that you should downgrade her merely because your judgment differs from hers. Instead, it seems that you should revise your judgment about the matter under dispute, especially once you have evidence that you are indeed vulnerable to overestimating your own epistemic competence. So, due to evidence about self-serving biases, there's at least some reason to think you should be somewhat conciliatory when disagreeing with a presumed epistemic peer.Footnote 8
Here is a quick objection to what I've been proposing. The Dunning–Kruger effect and other self-serving biases indeed puts pressure on our rationality, but at a different place. We typically overestimate our own epistemic competence, and thereby underestimate the comparative epistemic competence of others. But this just means that we need to be more generous about whom we deem to be epistemic peers. Perhaps we should be more inclined to grant peerhood to those who otherwise strike us as inferior, and to grant superiority to those who otherwise strike us as peers. But once we have made those adjustments, dealing with self-serving biases requires nothing further from us: we can consider the question of whether peer disagreement itself is grounds for revising belief head on. Any social psychological insights about our irrationality can be handled prior to that epistemological question. That is the quick objection.
This objection works only if it is sensible to assume that we can reliably judge who is and who is not an epistemic peer. For if we were able to do that, and we knew that we were, it would be of interest to think about what disagreeing epistemic peers require of us. But this is much like wondering what we would be rationally required to believe if we could read others' minds. Such an ability would open up a completely new world for us. It might be delightful to think about how we should go about believing things, were we able to do this. But given that we are not actually able to do this, we really do have to think about things like sincerity, trust and trustworthiness when investigating matters epistemological. We should not pretend that we are epistemically better than we really are.
Recall, then, that merely being aware of self-serving biases does not thereby enable you to compensate for their distortions. In fact, there is some evidence that awareness of these biases magnifies their distortions. Therefore, it is foolish to believe that now armed with this bit of social psychology, you can accurately gauge who indeed are your epistemic peers. The relevant question instead is: in light of (1) the general fact that you are not omniscient, and (2) the particular fact that you are unable to reliably determine who your epistemic peers are, what should you believe in the light of apparent peer disagreement? The answer seems to be: conciliate, even more than the equal weight view recommends.
That's the rough and informal characterization of the argument. Here it is again, but a bit more formally. We might view the evidence of the Dunning–Kruger effect and associated self-serving biases as higher-order evidence pertaining to the question whether Bush will win the election.Footnote 9 Consider the following exercise.
$$\hskip-6pc{\rm Let \; E} = \hbox{the (first-order) evidence about Bush and the election}.$$
$$\hskip-6pc{\hskip-30pt\rm Let}\,b = \hbox {the proposition that Bush will win the election}.$$
In light of E, you believe b. That is, in light of the evidence available to you, you believe that Bush will win the election.
$${\hskip-25pt\rm Let \; C} = \hbox{the evidence about your chair's various epistemic characteristics}.$$
$${\hskip-35pt\rm Let \; U} = \hbox{the evidence about your own various epistemic characteristics}.$$
$${\rm Let} \; p = \hbox{the proposition that my chair is my epistemic peer about such matters}.$$
In light of C and U together, you believe p. That is, in light of the evidence available to you about your own and your chair's epistemic characteristics, you believe that you two are equally competent when it comes to things like predicting the results of the next election. Now you go speak to your chair about national politics. You hear her say that Bush will lose the election. This is evidence about what she believes about the matter.
$${\rm Let \; X} = \hbox{the evidence about your chair's belief about whether Bush will win the election}.$$
When you gain X, what does your total evidence about whether b now support? Your total evidence is now E + C + U + X. Steadfasters think that if E alone really does support b, then you might have reason to continue believing b (Kelly Reference Kelly, Hawthorne and Gendler-Szabo2005). On this view, adding C + U + X to E need not change what you have reason to believe about whether b. If E supports b, so too might E + C + U + X.
Conciliationists, however, maintain that in light of E + C + U + X, you have reason to lower your confidence that b, and perhaps suspend belief about the matter altogether. That is, they think that E + C + U + X supports b less – perhaps far less – than E alone does. I am not going to pick sides just yet. Here, I want to add:
$${\rm Let \; D} = \hbox{the evidence about the Dunning-Kruger effect and associated self - serving biases}.$$
How does adding D to your evidence change things? In light of C + U alone, you believe p. Now, in light of C + U + D, you have less reason to believe p than you did in light of C + U alone. More specifically, D is some evidence that you do not assess the evidential value of U properly. And so D is evidence that C + U alone does not really support p as much as you are inclined to think. If in light of C + U you believed p, then it now seems that in light of C + U + D, you have (more) reason not to believe p, but some reason to believe that your chair is your epistemic superior with respect to the matter.
But is D relevant evidence not only about whether p, but also about whether b? That is, does evidence of self-serving biases directly alter whether you should think that Bush will win the election, over and above its effects upon your view about whether your Chair is your epistemic peer?
If your belief about whether p were appropriately sensitive to the relevant evidence, then D would not also be evidence about whether b, for that would be double-counting the evidential value of D. That is, suppose that your view about whether your Chair is your epistemic peer were appropriately sensitive to the available evidence about that, including evidence about your susceptibility to the Dunning–Kruger effect and other self-serving biases. If your belief about whether your Chair is your peer accurately reflects your evidence about whether she is, then you would not need to worry about these biases when it comes time to determine what to believe when you discover that you disagree with her. You would think that she is your peer just in case that's what all your evidence indicates, and so if you judge that she is indeed your peer, you have reason to take particular disputes at face value.
But if you are like most people, then there is significant evidence that your belief about whether p is not appropriately sensitive to all the relevant evidence about whether p. For you now have some evidence that you may suffer from various self-serving biases, and you now also have further evidence that merely being aware of this evidence does not thereby enable you to make accurate judgments about who is your epistemic peer. As we saw, being aware of self-serving biases does not enable you to judge accurately who are your peers. You thus have some evidence that indicates that when you judge someone else to be your epistemic peer, you need to be more conciliatory than you would need to be were you instead able to assess epistemic peers accurately.
This implies that D is indeed evidence about whether b. Or, more accurately, this implies that E + C + U + X + D is materially different evidence about whether b than E + C + U + X alone is. D gives you more reason to conciliate about whether b than you would otherwise have. Thus it seems that when you believe that you are disagreeing with an epistemic peer, you should be very conciliatory.
How conciliatory? Consider first a view of peer disagreement that is much more conciliatory than the well-known Equal Weight View. Here is a rough and admittedly extreme version of the view:
Initial Deference View. If you believe that your epistemic peer disagrees with you about some matter, you should give up your view and adopt hers.
Since it is typically rational to defer to epistemic superiors, and since disagreeing with someone you believe to be a peer may really be disagreeing with an epistemic superior, perhaps you should typically defer to those you believe to be peers. But there are various objections to consider, objections that call for a more subtle and careful answer.
4. OBJECTIONS
4.1 Peers on peers
The Initial Deference View states how to revise your view when you are disagreeing with someone you regard as an epistemic peer. Sometimes, however, when you are disagreeing with someone you regard as an epistemic peer, you also have a view about how she regards your comparative epistemic competence. There are four possibilities: you think (a) she regards you as an epistemic inferior, (b) she regards you as an epistemic peer too, (c) she regards you as an epistemic superior, or (d) she has no view on the matter (or you just can't tell what she thinks).
If you think that she regards you as an epistemic inferior (a), then the Initial Deference View remains at least plausible as it was before. That she views herself to be superior to you is no evidence that she is in fact your peer or your inferior. It's hard to see that her view has much probative value for you. Likewise, case (d) does not change anything; (d) merely makes explicit what was implicit in the run-up argument for the Initial Deference View.
But suppose you think that she also regards you as an epistemic peer (b). The Initial Deference View would then direct you to defer to her, but it would also direct her to defer to you. That certainly cannot be right. We must revise the Initial Deference View:
Once Revised Deference View. If you believe that your epistemic peer disagrees with you about some matter, and you do not have very good reason to think that she regards you as an epistemic peer, you should give up your view and adopt hers.
The Once Revised Deference View is thus silent about what to do when you know that you both regard each other as epistemic peers.Footnote 10 There is space here for the arguments of other philosophers to occupy. I take no stand on what you should believe in this situation.Footnote 11
What should you believe when you think that the disputant whom you regard as an epistemic peer regards you as an epistemic superior (c)? If she regards you as an epistemic superior, then, given that she too is probably susceptible to self-serving biases, her opinion of you counts as some evidence that you are at least her equal. And that evidence points against you deferring to her, as the Once Revised Deference View recommends. On the other hand, the psychological evidence about widespread errors in self-assessment may dampen your confidence in her assessment of you no matter what she concludes about your comparative epistemic competence. I thus doubt there is much one can say at a general level about what to believe in cases like (c). As a result, further revisions to the view under development will omit such cases. So we now have
Deference View. If you believe that your epistemic peer disagrees with you about some matter, and you do not have good reason to think that she regards you as either an epistemic peer or an epistemic superior, you should give up your view and adopt hers.
The Dunning–Kruger effect and related biases provide us with grounds for thinking that the Deference View is true, or closer to the truth than rival views.
The Deference View mentions whether you have “very good reason to think” that your disputant regards you an epistemic peer or superior. I am not sure if this is the best way to put the matter. We could weaken it or strengthen it.
We could strengthen this language by replacing “very good reason to think” with something factive, such as “know”. Then the Deference View would not tell you to defer when you know that your disputant believes that you are her epistemic peer or superior. But sometimes we do not know that, even if we believe it with very good reason. What we have reason to believe seems to depend upon what else we have reason to believe, not only upon what we know. If you have very good reason to believe that your disputant believes you are her peer or superior, then it seems that you need not defer to her, even if it turns out that you (say, due to misleading evidence) do not know this. Perhaps this is good reason to preserve the “very good reason to think” expression in the Deference View.
We could instead weaken this language by replacing “very good reason to think” with something like “believe”. Then the Deference View would not tell you to defer when you (merely) believe that your disputant believes that you are her epistemic peer or superior, even if you believe this for no good reason. Such a replacement brings the second clause of the Deference View in line with the first clause, which likewise is phrased in terms of what you believe. The symmetry seems appropriate.
I want to note one problem with this last proposal. Sometimes when your disputant tells you that she believes that you are her epistemic peer, she does not really believe that. There is a good chance she is just being polite to you. Many people who believe they are epistemically superior to others do not tell other people that they think this, for obvious reasons. It is very nice to be told by someone you regard as your epistemic peer that she feels the same about you. Furthermore, I surmise that many people who believe they are epistemically superior to others are not even fully honest with themselves, and so do not believe that they believe that they themselves are superior. They want to believe that they are more epistemically egalitarian than they in fact are. Hubris is not a quality one wants to believe one possesses. All told, it seems unwise to conclude that your disputant regards you as an epistemic peer simply on the grounds that she tells you that she does so.
This implies that we shouldn't simply take disputants' claims about peerhood at face value. And yet, I am guessing that we tend to do so. That is, I am guessing (and it is only a guess) that most people are only too ready to believe that they and their disputants regard one another as epistemic peers. If my guess is right, it would be mistaken to weaken the language of the Deference View as we are considering, for merely believing that your disputant believes that you are peers is not strongly positively correlated with whether she indeed believes this.
There are many considerations here to balance, and I do not know how to balance all of them. I suspect, however, that we should rarely be confident that we know how others actually regard our own comparative epistemic powers. And so I also suspect that often we should be less confident about this than about the matter in question prior to learning about the disagreement. But these are just suspicions, not arguments.
4.2 Differing domains
So far, I have discussed the relation between the Dunning–Kruger effect and peer disagreement at a very general level, independently of any particular domain. But there is some evidence that the effect is more pronounced in some domains than in others. Self-serving biases do not infect our judgments about epistemic peers across all domains equally. As Dunning et al. Reference Dunning, Heath and Suls2004 write, “people's general evaluations of their skills and character – such as whether they are good leaders or verbally skilled – tend not to be tethered very tightly to objective performances in tasks that should reflect those skills and character traits” (71). But such distortions are not inevitable. For specific domains for which there are objective track records, the Dunning–Kruger effect may nearly dissipate. Thus if you think that I am your epistemic peer with respect to picking horses or solving math problems, and you think this purely because you know that our past answers have been equally accurate, then you should be able to assess accurately whether we are peers. Note that even here, however, the possibility of distortion has not evaporated: one might very well think to oneself “Well, my track record would have been better had it not been for the fact that (insert excuse here) …”
The Dunning–Kruger effect is more pronounced in domains where competence cannot be objectively characterized. When there are no standard ways to measure expertise, people tend to seriously overestimate their own abilities and skills: “I realize that I may be neither a good free throw shooter nor a good chess player, but I do think that I am a very good driver, friend, parent, and that I have a wonderful sense of humor.” Thus we should be skeptical of the reasoning Descartes employs when he writes:
Good sense is, of all things among men, the most equally distributed; for every one thinks himself so abundantly provided with it, that those even who are the most difficult to satisfy in everything else, do not usually desire a larger measure of this quality than they already possess. (Discourse on Method, opening lines)
Hobbes similarly writes:
That which may perhaps make such equality incredible is but a vain conceit of one's own wisdom, which almost all men think they have in a greater degree than the vulgar; that is, than all men but themselves, and a few others, whom by fame, or for concurring with themselves, they approve. For such is the nature of men that howsoever they may acknowledge many others to be more witty, or more eloquent or more learned, yet they will hardly believe there be many so wise as themselves; for they see their own wit at hand, and other men's at a distance. But this proveth rather that men are in that point equal, than unequal. For there is not ordinarily a greater sign of the equal distribution of anything than that every man is contented with his share. (Leviathan, ch. 13)
Both philosophers appear to conclude that people have equal amounts of “good sense” or “wisdom”, on the grounds that each person, not thinking that others have more, is satisfied with her own share of it. Perhaps it is roughly true that most people do not think that other people have more wisdom. And perhaps it is also roughly true that each is satisfied with how wise she believes herself to be. But is the best explanation for this that people are in fact equally wise, as Descartes and Hobbes argue? Socrates would mock this explanation.Footnote 12
The better explanation is, as social psychology teaches, that we overestimate our own competence, especially about matters that are difficult to measure objectively. And epistemic prowess in particular, and “good sense” generally, surely falls into this category: it is practically impossible to measure wisdom objectively. So although there may be some domains for which self-serving biases are minimal, those worried about whether they are savvy enough to generally counteract them should not fool themselves. It would be doubly irrational to think that you are immune.
Here is another way the severity of self-serving biases might vary across domains. Some psychologists have found that self-serving biases are significantly less pronounced when making perceptual rather than conceptual judgments. People regularly overestimate their own ability to make accurate conceptual judgments, but they do not (grossly) overestimate their own ability to make accurate perceptual judgments:
[E]vidence suggests that this overconfidence phenomenon occurs most often when participants are making conceptual decisions, such as answering general knowledge questions (Lichtenstein, Fischoff, and Phillips, 1982) regarding one's knowledge of history, geography, or literature (Fischhoff, Slovic, and Lichtenstein, 1977), or when answering problem-solving questions (Crawford and Stankov, 1996). (Kvidera and Koutstaal Reference Kvidera and Koutstaal2008: 254)
This suggests that you have more reason to be conciliatory in disputes about conceptual questions than about perceptual questions.Footnote 13 This may be related to the point made earlier about track records. It is easier to determine your track record for the accuracy of your perceptual judgments than it is for your conceptual judgments. Thus (for whatever reason) perhaps you have reason to remain steadfast in disputes about which horse won a race, what color a particular dress is, or whether two lines are the same length. But when disagreeing with a person you believe to be your peer about matters moral, religious, political or philosophical, you still have reason to be conciliatory. Perhaps the Deference View should be restricted to conceptual disputes.Footnote 14 We should modify the Deference View, then, in light of these various domain-related considerations.
Domain-sensitive Deference View. If you believe that your epistemic peer disagrees with you about some matter, and you do not have very good reason to think that she regards you as either an epistemic peer or an epistemic superior, and you do not have objective track records to establish that you are indeed peers, and the matter under dispute is more conceptual than perceptual, then you should give up your view and adopt hers.
The Domain-sensitive Deference View is more plausible than the simpler Deference View. But it too can be improved upon.
4.3 How deferential?
So far I have argued that at least sometimes the Dunning–Kruger effect should lead you to be more conciliatory than the Equal Weight View suggests. And I have raised the possibility that you have reason to outright defer to those whom you believe to be your epistemic peers. But can that really be correct?
Consider one of Kelly's best objections to the Equal Weight View (Reference Kelly, Goldman and Whitcomb2010). He persuasively argues that splitting the difference with your epistemic peer would be wise only if the only relevant evidence were the higher-order evidence about your and your peer's initial judgments about the matter under dispute. But the total relevant evidence comprises more than this higher-order evidence: the original evidence bearing on the claim does not disappear after you discover that you and your peer disagree. The total evidence includes both this original evidence and the higher-order evidence. You should not toss out the original evidence when the dispute emerges. But that is essentially what the Equal Weight View directs you to do.
Something analogous may be going on with The Domain-sensitive Deference View (and its predecessors). The Domain-sensitive Deference View says that in at least some disputes, you have most reason to go with your disputant's view rather your own initial view. But this seems to put too much weight on your disputant's view. It essentially instructs you to toss out everything but your disputant's view. But that cannot be right, for reasons similar to those Kelly identified when discussing the Equal Weight View.
The correct view, then, seems to lie somewhere between the Equal Weight View and the Domain-sensitive Deference View. You should put extra weight on your disputant's view, but you should not toss out all the other evidence. Exactly how much extra weight you should put on your disputant's view is probably impossible to say, at least at any level of generality. It may depend upon whether the dispute is about something conceptual rather than perceptual. It may depend upon whether there is anything like an objective track record available. Further psychological research may uncover other parameters that we would do well to take into account.Footnote 15
Here's a quick argument for the thought that you should put only a little bit of extra weight on the view of your disputant. Many studies about the Dunning–Kruger effect and associated self-serving biases ask subjects to compare themselves to the average individual in some category: the average college student, the average professor, etc. These studies show that most people seriously overestimate their own skills and competences. But some psychologists have found that while many of us unjustifiably think that we are better than others, “the better-than-average effect is diminished as the social comparison becomes less ambiguous, or more objectively defined” (Alicke et al. Reference Alicke, Klotz, Breitenbecher, Yurak and Vredenburg1995). Thus, when you compare yourself to “the average individual”, it is likely that you think that you are significantly superior to her, even if you are not. When you compare yourself to a particular individual (e.g. your Chair, your partner, your neighbor), however, you won't overestimate yourself as much. You still are likely somewhat overconfident, but not by as much as when you compare yourself to a faceless abstraction. This suggests that it would be a mistake to be completely deferential when disputing with a flesh-and-blood person whom you deem to be your epistemic peer. If you, like most, overestimate your own competence a little bit, then a little extra weight on your disputant's view may suffice.
In light of these worries, I think that The Domain-sensitive Deference View is too deferential. The Dunning–Kruger effect and associated self-serving biases are not evidence that you should completely defer to those you take to be your epistemic peers. But they are evidence that you have reason to put more weight on their views than you put on your own. You should combine your view with hers, giving less weight to your own than to hers. This is what a humble person would do. Hence we have
The Humble View. If you believe that your epistemic peer disagrees with you about some matter, and you do not have very good reason to think that she regards you as either an epistemic peer or an epistemic superior, and you do not have objective track records to establish that you are indeed peers, and the matter under dispute is more conceptual than perceptual, then you have reason to combine your view with hers, placing more weight upon hers than upon your own initial view.
The Humble View says that you have reason to put more weight on your disputant's view, because you have evidence that you overestimate your own intellectual competence and skill. You have reason to be significantly but not completely conciliatory.Footnote 16 , Footnote 17
Finally, there are different ways of being significantly but not completely conciliatory, and I'm not sure which is the best way to conceptualize what there is most reason to believe. One way is to give up your original view and adopt your disputant's view, but with less confidence in the view than she herself has. Perhaps she is very confident that Bush will lose the election, and so you now only tentatively think that Bush will lose the election. You believe what she believes, but you do so more hesitantly. We might represent the relevant doxastic states here by using credences. Suppose that your initial credence in the proposition “Bush will win the election” is 0.7, while your disputant's is 0.2. Moving to a credence of, say, 0.4 might then accord with the Humble View. This is one way to put extra weight on your disputant's view. Call this way credal conciliationism.
A rather different way to do so is to interpolate between your original view and your disputant's view in a significantly conciliatory fashion. Suppose that you initially think that Bush will receive 54% of the popular vote, while your disputant whom you regard as your peer thinks that Bush will receive 46% of the vote. The Equal Weight View would direct you to believe that Bush will receive 50% of the vote, thereby splitting the difference. The Humble View, however, directs you to put extra weight on the view of your disputant. You could do this by coming to believe that Bush will receive, say, 48% of the vote, which is closer to your disputant's view than to your own initial view. Believing something in between what your disputant believes and what the Equal Weight View tells you to believe is thus a way of putting extra weight on the view of your disputant. Since your new belief is, in a sense, in between your disputant's view and what the Equal Weight View recommends, we might call this way to follow the Humble View interpolative conciliationism.
Obviously, only some kinds of disputes lend themselves to interpolative conciliationism. If you disagree with someone about whether the number of planets in the solar system is odd or even, there is no room for interpolation. There is just nothing in between the odd and the even. But when disputes do lend themselves to interpolation, then it is unclear whether you have most reason to conciliate credally or interpolatively. I hope to say more about this in future work. Until then, the Humble View is not as directive as one might wish.
One final worry about the Humble View. If we are indeed largely unable to learn from the fact that we are biased in the ways I have described, then why think that philosophers are able to learn from the Humble View? Am I now being too confident that I (and perhaps others) can implement this view about how to form beliefs?Footnote 18
Perhaps so! But this points to a much larger issue. In many ways, we are not perfect, and we know this. But we are also often convinced that certain ways of coping with and responding to our imperfections would be wise. But if we can't believe and act well, why think that we can implement these coping responses well, or even as well as anything else we aim to do? Shouldn't we just throw up our hands, realizing that any efforts to manage our shortcomings are likely to be equally doomed? Maybe so, but I doubt it. The proof of the Humble View will be in its (attempted) implementation. We don't fail at everything. Will this be one of the things we succeed at? I am cautiously hopeful.
5. GETTING META
The Dunning–Kruger effect and associated biases provide grounds for thinking that most people overestimate their own comparative epistemic competence. This shapes what we have reason to believe when we discover that we disagree with others. In many circumstances, we have reason to be very conciliatory to those we regard as epistemic peers.
But the Dunning–Kruger effect also points to a larger issue. When we disagree with other people about how to split a check or whether it will rain tomorrow, we bring many background views to bear upon the disagreement (Elga Reference Elga2007). You might believe that it will rain tomorrow, but you also believe things about your own competence at predicting the weather, your disputant's competence at predicting the weather, your disputant's ability to know how she evaluates your epistemic competence, your disputant's sincerity in conveying that opinion to you, and your own comparative abilities to neutralize various biases and distortions. No doubt there are still other cognitive biases evidence of which would affect what we have reason to believe. It is difficult to say anything concrete about how to weigh all this up. The Humble View is an initial stab at this, but if, as I expect, many presumed epistemic peers are not fully convinced, consistency requires me to seriously doubt that it is correct.Footnote 19
