Who’s Afraid of Disagreement about Disagreement?

ABSTRACT

This paper is not concerned with the (amply discussed) question as to the rational response to peer disagreement. Instead, it addresses a (considerably less often debated) problem to which many views about the (epistemic) significance of disagreement are vulnerable (to some extent or another): self-undermining. I reject several answers that have been proposed in the literature, defend one that has been offered (by meeting objections to it), and show that in its light, the prevalent assumption that the ‘equal-weight view’, a prominent view about disagreement, rationally requires us to suspend judgement about contentious matters, is seen to be too pessimistic.

KEYWORDS:

• Peer disagreement
• conciliationism
• self-undermining
• Right Reasons view
• scepticism

1. Introduction

The problem of disagreement concerns the correct response to learning that your ‘epistemic peer’ disagrees with you about some issue (tomorrow’s weather, the permissibility of abortion, the existence of universals).¹ This paper is not concerned with this (amply discussed) question. Instead, it addresses a (considerably less often debated) problem to which many views about the (epistemic) significance of disagreement are vulnerable (to some extent or another): the problem of self-undermining (SUP). The problem arises for any view that prescribes conciliation, the adoption of a credence for the disputed proposition that is (somewhat) closer to that of our disagreeing peer than was our original credence.² It is engendered by the fact that there is peer-disagreement about the correct response to disagreement.

In fact, SUP consists of two distinct challenges. The first pertains to the very possibility of a satisfactory conciliationist response to disagreement about disagreement (henceforth, DAD). The second is the worry that regardless of the strength of the arguments in support of a conciliationist view about disagreement, a satisfactory response to DAD will have us reducing our confidence in it; may even leave it with no justification at all. The two problems are logically independent. A cogent response to SUP may render conciliationism (or at least some versions of it) unjustified. And, conversely, a response to DAD that does not require us to reduce our confidence in our view about disagreement needn’t be cogent. This latter fact is not just an abstract logical point: I will reject views that prescribe steadfastness for the conciliatory view itself in response to DAD, and defend a view that requires conciliation (section 2). So there will be a need to contend with the second concern, diminished justification, on behalf of the view I end up defending (section 3).

Two views about disagreement are totally invulnerable to SUP. According to the first, which I label absolute steadfastness, henceforth, AS, one shouldn’t change one’s credence at all when encountering even many disagreeing peers (or even superiors); disagreement never has any epistemic significance.³ The other position on peer disagreement that is invulnerable to SUP is the Right Reasons view (RR), according to which the peer that responded correctly to the original evidence should retain his original credence (Titelbaum, 2015).⁴ RR doesn’t require universal steadfastness, because it does not prescribe for those who (unknowingly) responded incorrectly to the evidence. But it never prescribes a reduction in one’s credence in the face of peer disagreement (indeed, even a disagreement with many peers or with an epistemic superior). So it, too, is invulnerable to SUP, because it prescribes a retention of one’s confidence in RR.

There remains a plethora of (more or less) conciliatory views, which require some degree of (epistemic) compromise, and to which, correlatively, SUP poses a challenge. To facilitate the discussion, I will focus on SUP as it arises for the Equal Weight view (EW), which requires us (typically)⁵ to ‘split the difference’ with our peer, assign to the disputed proposition a credence that is the average of our credence and his.⁶ SUP is here most acute. Thus, Plantinga (2000, 446) says that the proponent of EW ‘shoots himself in the foot … Under the conditions that do in fact obtain – namely, his knowledge that others don’t accept it – he can’t properly accept it’. And Elga (2010, 179) says ‘your view on disagreement requires you to give up your view on disagreement’. But what I say is also applicable, mutatis mutandis, to more moderately conciliationist views.

2. Major Proposals for Contending with SUP

In considering proposals as to how DAD should be accommodated, it is helpful to distinguish between those that are premised on the supposition that EW prescribes inconsistently, sometimes giving incompatible directives, and those that deny it. I will consider them in turn (addressing the inconsistency claim when I discuss the latter). Elga, who argues for the inconsistency claim (2010), proposes, in response to the inconsistency charge, a modified version of EW, EW*, according to which one should always ‘split the difference’ except in DAD. This is an attractive suggestion insofar as it enables the proponent of EW to continue to adhere to it when responding to ‘standard’ disagreements (those whose subject matter is not disagreement). Furthermore, because he remains steadfast with respect to EW*, he can straightforwardly meet the sceptical challenge, according which belief in EW is unjustified. But what is the justification for the exemption of EW from its own strictures?

EW* does not apply to itself, Elga argues, because ‘in order to be consistent, views on disagreement, like any basic epistemological principle, must be dogmatic with respect to their own correctness’ (2010, 184–185). There are three objections to this rationale. First, if a basic rule, in analogy with a basic belief, is one which is not justified via an argument (as perhaps is the case for Modus Ponens and the principle of induction), then no view about disagreement is basic: proponents of the (several) views invariably adduce arguments (complex ones) by way of justifying them. Second, a rule must be self-certifying; it is not a matter of stipulation. Elga cites Lewis’ (1971) analogy with a consumer magazine that recommends a rival, and endorses Lewis’ (plausible) claim that we shouldn’t be guided by it. But this doesn’t mean – and neither does Lewis suggest that it does – that the magazine is free to stipulate its superiority. Such a claim must be the verdict of the procedure the magazine follows in general. And if the (self-certifying) verdict isn’t forthcoming (as in Lewis’ example), the magazine fails the test. So we should apply to conciliatory views their own test. And they seem to fail it (to varying degrees): their own test decrees that disagreement should engender a reduction in the confidence of its proponents in its truth. Third, as Christensen argues (2013, p. 89), the exemption is not in keeping with the spirit of EW, which EW* purports to reflect. The reasoning that leads us to accept EW should move us to reduce our confidence in E*: we are fallible, and the peers that reject E* constitute (higher-order) evidence that we may have misinterpreted the evidence we took to support it. ‘There is nothing about this particular topic [DAD] that would make my way of thinking about it special, or especially immune from my usual sort of blunders’ (Christensen 2013, 89). And he asks rhetorically (2013, 89), ‘[o]n what basis could I conclude that I’m the one who got lucky, rather than those who reject [EW*]?’. I conclude that Elga’s proposal fails.

Here is Christensen's first proposal (Christensen 2013). He also thinks that EW prescribes inconsistently, but only when conjoined with another (epistemic) assumption, level-connections. It requires a certain fit between 1^st- and 2^nd-order beliefs. For instance, one shouldn’t ascribe to a proposition a credence that one thinks is irrational. Equally, ‘rational doubts about the correctness of a certain epistemic principle should weaken the extent to which that principle governs one’s beliefs in general’ (Christensen 2013, 85). Without this assumption, Christensen points out, we could respond to DAD by retaining our confidence in EW, and responding to future disagreements as it dictates. Having given up level-connections, we would no longer be committed to a competing prescription. But this, Christensen argues, is doubly unsatisfactory. First, level-connections seems plausible in itself. Second, without it, we lose support for EW, which follows from the conjunction of level-connections and another seemingly plausible principle, which Christensen labels ‘Respecting evidence of our epistemic errors’: when ‘one encounters good evidence that one’s initial level of confidence in P is higher than that supported by one’s first-order evidence, … one will give significant credence to the claim that one’s initial level of credence is too high’. So retaining EW in the face of DAD leaves it without justification (on pain of contradiction). So, Christensen concludes (2013, 92), DAD forces on the proponent of EW a ‘not entirely comfortable’ epistemic dilemma, engendered by two conflicting ideals that motivate EW: respecting evidence of our epistemic errors and level-connections.

Perhaps Christensen is right to suggest (2013, 93) that choosing to violate level-connection incurs a lesser intellectual cost. But surely it would be more satisfactory for the conciliationist to be able to respond to DAD without having to confront this ‘uncomfortable’ choice.⁷ So let us consider proposals that do not have us being embroiled in it.

In a later paper, Christensen argues that ‘[t]he standard Self-Defeat Objection shows, correctly, that complying with [EW] has the consequence that an agent will … have a certain credence in P, and also … believe that some different credence in P is more rational’ (Christensen 2021, 2208). But, Christensen now thinks, initial appearances to the contrary, this (akratic) combination of credences does not involve an ‘uncomfortable’ choice. I will contest this claim.

Christensen adduces several putative examples of ‘rational epistemic akrasia’: cases in which an agent is rational in ascribing to a proposition a credence that he thinks is irrational. DAD, he argues, is another case in point: the proponent of EW is rational to ‘split the difference’ on some proposition while thinking his post-splitting credence is irrational. His violation of level-connections is (perfectly) rational (not an ideal he is violating).

We may grant, at least for the sake of the argument, some of the examples Christensen adduces. They involve subjects who mistakenly, but rationally, accept some view about disagreement on which rationality and accuracy (the analogue of truth in the case of credences) may diverge (The possibility of justifiably believing an incorrect epistemic principle is denied by Titelbaum (2015). But he endorses RR. And the proponent of EW, who is more open to the possibility that our fallibility should be taken into account in epistemology, may well find it congenial.). Here is such a case, adapted from Christensen. Suppose a proponent of EW disagrees about the correct view of disagreement with someone who is by far his philosophical superior. The superior thinks that in a disagreement, one shouldn’t be influenced by the interlocutor’s view (on the disputed proposition), since that would constitute ‘an unbecoming abdication of intellectual autonomy, and thus compromises the rationality of the affected beliefs’ (2021, 2203).⁸ Suppose, further, that they also disagree about some (philosophical) proposition, P. The proponent of EW disbelieves it, and his superior believes it. The proponent of EW, we may plausibly assume, thinks that in a disagreement with an epistemic superior, one should assign one’s interlocutor significantly greater weight than one does to oneself. So he will significantly move towards his superior with respect to both. Now, the evidence that makes it rational for him to doubt the rationality of his confidence about the truth of P does not make it rational for him to doubt the truth of P. (The prescriptions of the autonomy view of rationality may perceivably fail with respect to accuracy.)

So rational akrasia may be plausible in the sorts of case Christensen considers, cases in which a conciliationist disagrees with a someone who gives weight to considerations that are not truth-related (like autonomy).⁹ But Christensen omits to consider an important case of DAD, which is less hospitable to the possibility of rational akrasia. In fact, there are many such cases. They involve two conciliationists who disagree about the proper degree of conciliation a peer-disagreement requires. (Christensen treats Conciliationism as a single position, rivaled by non-conciliationism.) Suppose, to make things definite, our conciliationist espouses EW and his peer thinks the appropriate degree of conciliation is .2. If our conciliationist ‘splits the difference’ with his peer in both disagreements, he will (as before) be believing akratically, because (as before) he will deem his 1^st-order credence to be irrational. But here, his akratic belief cannot be made acceptable in the way Christensen employed in the congenial cases, because there is no (perceived) divergence between rationality and accuracy. The opponent does not attach a value to non-accuracy considerations. So whereas in the disagreement with the peer who values autonomy at the expense of accuracy in DAD, the ‘autonomous’ peer can admit with equanimity that his view of disagreement engenders a gap between accuracy and rationality, the conciliationist opponent won’t. All versions of conciliationism are aiming at accuracy.¹⁰

The next proposal I will consider is Pittard’s (2015). He aims to do what he (plausibly) thinks Elga fails to do: exclude the application of EW to itself in a principled way. Consider the case in which a proponent of EW disagrees with a proponent of AS, one who thinks steadfastness is always required in response to a disagreement. Now, to be sure, Pittard argues, EW requires splitting the difference with respect to our credences. But the underlying rationale for this is displaying proper epistemic deference. And this rationale also requires conciliating with respect to our reasoning in the face of a disagreement about the proper way to reason. In general, Pittard notes, the two coincide: when we conciliate with respect to the weather (say), we are both responding and adjusting our credence as EW enjoins us to do. But the two injunctions come apart when the dispute concerns disagreement. Here, when we reduce the credence we assign to EW, we are deferring with respect to our credences. But we are not being deferential with respect to our response: our disputant thinks we should remain steadfast. And to comply with this injunction, we should remain steadfast with respect to EW. This symmetry means, Pittard concludes, that remaining steadfast is as deferential as splitting the difference. And because there is no better way of being deferential, remaining steadfast is permitted. Since the credence we ascribe to EW even after discovering the disagreement remains unchanged, EW does not issue contradictory prescriptions. (It would, Pittard agrees, if one were to conciliate on it.)

To facilitate the discussion of Pittard’s proposal, let us suppose that our disputant in DAD is a proponent of AS (absolute steadfastness). Although this is not a plausible view, it facilitates the discussion. (As before (see n. 10), the alternative, RR, is problematic in this context, because it only prescribes for a person who responded correctly to the evidence. So it is not clear how Pittard’s suggestion is supposed to work against such a disagreeing peer.)

I will now offer three objections, starting from the least weighty, to Pittard’s proposal. First, the strategy isn’t going to work against a disputant who thinks one should always entirely defer. By retaining one’s credence in EW, one is not deferring to him at either level. But this is not a serious worry, since no one holds this view, at least no one who is our philosophical peer, let alone our superior. So one need not defer to such a disputant.

Second, Pittard is assuming that in the face of the epistemic dilemma, one can rationally refuse to conciliate on Conciliationism (and continue to abide by it). But, equally, one can conciliate on conciliationism (and be guided, in future agreements, by the principle thus arrived at). So this suggestion relies on (epistemic) justification being permissive. To those who think justification is impermissive (White, 2005); that there is always just one correct response to the evidence, this will count heavily against the proposal.¹¹

Here is the third (and most important) objection to Pittard’s proposal. He suggests (2015, p. 9) that retaining one’s credence about EW has equal claim to deference to one’s (‘steadfaster’) disputant as does reducing it to 0.5. But this seems mistaken. If it were a one-off occurrence (the last dispute one will ever confront), the claim would be plausible. But in fact, one’s credence about EW is a commitment to a life-long policy of responding to disagreements, whereas reasoning as your disputant (regarding disagreement) requires happens just once (when you confront DAD). So overall, Pittard’s strategy is hardly deferential; it is much closer to steadfastness. And this means it is not true to the spirit of conciliationism.

It is now time to consider the final suggestion, one according to which the proponent of EW is to conciliate in DAD. This is seemingly the most natural suggestion, straightforwardly in keeping with the spirit of EW. But if it is to be viable, it must reject (in a principled way) the claim (Elga, 2010) that such a response renders the conciliatory view contradictory. In discussing Elga’s argument (2010, 179–181) for the claim that EW prescribes inconsistently (in the actual circumstances),¹² I follow Christensen’s (2013) rendition, which is more perspicuous. Suppose I assign some credence to a proposition, P, and my peer assigns a different one. Suppose, further, that I also face DAD: some peer and I ascribe to EW different credences. Now, EW prescribes ‘splitting the difference’ with respect to P. But it also enjoins me to ‘split the difference’ with respectto EW. Assuming I was (almost) certain about EW, I now reduce my confidence in its truth significantly, as a result of which I am now committed to responding to the disagreement about P in a way that is different from that prescribed by EW: I must give some weight to the rival view about disagreement, one that requires a lesser degree of conciliation or none at all. So, the argument concludes, EW engenders two incompatible prescriptions.

Christensen notes (2013, p. 86) that the argument can be applied against any view that recommends some conciliation (i.e. falls short of AS). But this (‘partners in guilt’) fact is not a reason for dismissing the argument. If we are to reject it with a clean intellectual conscience, we must find the fault in it. I submit the reasoning is fallacious because it treats DAD on a par with other disagreements, so that they have to be contended with concurrently. In fact, a disagreement about EW, even if it occurs concomitantly with a disagreement about some other proposition, takes precedence. Here is why. One should only use a tool one thinks is a good one. If I suspect the brakes of my car are not working properly, I ought to check them before I drive my car (and not after 10 miles, say). The same holds for doxastic tools, rules for forming beliefs. For instance, if I discover that a logical principle I have been invoking is fallacious, I ought to eschew it. Indeed, I should correct (whenever possible) my past invocations of it. Closer to home, EW is a tool for contending with disagreements. And it, too, takes priority over its invocations. If I come to believe that my confidence in the correctness of a rule is too high, I must adopt a better (by my present lights) rule, and only then confront disagreements, abiding by it. Indeed, I should correct, whenever possible, my past responses to peer-disagreements, in conformity with the updated rule. Once I have adopted a new rule for contending with disagreements, the prescription of EW is no longer relevant. And since it doesn’t exert a rational pull in another direction, we can, pace Christensen (2013), retain level-connections even upon encountering DAD.

Having allayed the worry that conciliating on EW in DAD engenders incompatible prescriptions, we can consider the adequacy of the (conciliatory) suggestion. This is required because by itself, the consistency of a proposal does not guarantee its adequacy. To this end, consider, first, the claim (Weatherson, 2013) that ‘splitting the difference’ with respect to EW is unstable. Here is Christensen’s rendition of argument, which is more perspicuous (2013, 79–80). Suppose the proponent of EW encounters a peer who accepts AS. In response to the disagreement, he ‘splits the difference’, ending up with a credence 0.5 for EW. But with this new credence, he now thinks he shouldn’t have conciliated so much: he gave EW greater weight than its current credence suggests to be appropriate. Instead, he ‘should adopt a credence in between what the two rules recommend: a mixture of the recommended credences’. But with this new credence, he now thinks he hasn’t conciliated enough … And the process continues”. As Christensen notes, the argument, if cogent, can be applied to any (even somewhat) conciliatory view, so AS is seemingly the only stable view. But, in fact, the instability can be eliminated. Weiner (2007) suggests a way of determining a stable credence for EW in the face of DAD, one in which there is no rational pressure on its proponent to change his credence in response to the disagreement. In fact, the stable credence for EW, supposing the proponents of EW and AS respectively are initially certain of their views, is ²/₃.¹³, ¹⁴

Here are two objections to Weiner’s suggestion. First, Weatherson (2013, 60–61) and Decker (2014, 1108–1109) think the ‘difference-splitting’ procedure should be carried again and again, forcing the proponent of EW to end up ascribing to EW a credence nearly equal to that of his disputant (who accepts AS), and, in particular, disbelieving it. Thus, Decker argues (2014, 1108) that the first conciliation produces ‘a new evidential situation’, requiring the proponent of EW to move towards his peer again. At no point is the proponent of EW allowed to stop the sequence of conciliations, so the sequence of credences converges to that of his peer (who accepts AS, and never budges). But the reasoning is mistaken. We are not getting new relevant information after first conciliating on EW: we have already taken into account the fact that our peer disbelieves EW on the basis of shared evidence when we first conciliated. So just as we wouldn’t change our credence in response to reading the same newspaper again, we shouldn’t modify our credence beyond the first conciliation.

The second objection to Weiner’s suggestion is Christensen’s (2013, 81): ‘to the extent that [the proponent of EW] thinks [EW] is correct, she’ll be led to lower her credence in [EW]; and to the extent that she thinks [the steadfast view] is correct (and thus [EW] is incorrect), she’ll be led to raise her credence in EW! And this fact is perfectly transparent to [her]. It’s not at all clear that, in this situation, … following [Wiener’s suggestion] would be reasonable’. And he continues: ‘From the agent’s perspective, [she] is supposed to think in the following way about [her] doxastic adaptation to the disagreement: “Well, suppose that [EW] is true. In that case, I shouldn’t be very confident of it. And [EW] is probably correct. So I shouldn’t be too confident in it” … Shouldn’t [she] arrange her beliefs instead so that, to the extent that [EW] is likely to be true, she has high credence in [it]? There seems to be something fundamentally incoherent in her reasoning in a way that manifestly reverses this relationship’.

The objection doesn’t target only EW. It targets any conciliationist response to DAD that enjoins us to lower our credence in our conciliationist view, CV, to a greater extent the more confident we are of CV. But, I will now argue, the objection is specious, misrepresenting its target. True, a view that prescribes a reduction in the credence we attach to a proposition the higher the initial credence (for instance, ‘If P’s credence is x, reduce it to x/2’) is unreasonable. But this is not what Wiener is proposing. Instead, he suggests that we should conciliate more on EW the higher its credence only when there is appropriate evidence - a disagreeing peer. This may be ‘odd’, but harmlessly so: it stems from the content of EW. It tells us to conciliate in case of a disagreement, including a disagreement about it. And, naturally, everything else being equal, the stronger our confidence in its correctness, the bigger the conciliation required. I conclude that Weiner’s proposal survives the objection. And it remains to consider, on its behalf, the second component of SUP: the possibility of justifying the belief in EW in the face of DAD.

3. Scepticism?

EW seems to be threatened by scepticism in two different ways. The first is that it seemingly renders itself unjustifiable (in the face of DAD). The second is that it renders contentious views unjustified. Both problems seem to be more serious the greater the conciliation a view requires, being most severe for EW. So I will discuss them (in turn) as they arise for it.

Here’s Decker’s (2014, 1132–3, original italics) formulation of the first problem: ‘Informed philosophers are not justified in believing conciliationism’. Perhaps the thought is that the proponent of EW, assuming he and his steadfast interlocutor are equally certain about their respective views, is required to ‘split the difference’ with his interlocutor, ending up ascribing to EW credence 0.5. That this is too quick can be seen by distinguishing between two cases. In the first, the proponent of EW starts out by being (nearly) certain of EW; in the second – he is far less certain (but still more confident that it is true than he is of its falsity; this is why he merits the label ‘proponent of EW’). In the first case, we can straightforwardly rebut the sceptical claim by invoking Weiner’s proposal (2007). The rational post-DAD credence for EW is ²/₃, which may well suffice for belief. And surely this is right, in the spirit of EW. The proposals that attempt to justify steadfastness with respect to EW strike me as rationalisations or wishful thinking.

In the second case, the proponent of EW starts out ascribing to it a credence>¹/₂, but considerably less than 1. This case is rarely (if ever) discussed. Thus, in the case discussed by Christensen (2013), both sides in DAD initially ascribe to their view probability .99. But it is, in fact, the more reasonable starting point for our consideration of DAD, and, in particular, the sceptical threat it is supposed to engender. The proponent of RR thinks that if one adopts an attitude towards a claim that is knowable a priori, certainty is the only rational attitude.¹⁵ And neither disagreements nor any other higher-order evidence have any epistemic significance. But from the perspective of a conciliationist, whose epistemological views are informed by the thought that justification must be sensitive to higher-order evidence about our mistakes in respecting the evidence, this supposition is very implausible. And this thought is not restricted to views about disagreement; it pertains to any (non-trivial) philosophical view: dualism as against physicalism in the philosophy of mind, subjectivism as against realism in meta-ethics, etc. We all know that philosophical questions in the past have engendered recalcitrant disagreements, and that they are hard, if not impossible, to resolve satisfactorily. Many of us change our mind about a view we initially think to be very plausible. Feldman suggests, in this (cautious) vein, that in ‘hard cases’, of which philosophical disagreements are an instance, ‘an honest look at what the evidence supports … reveals that our evidence is decidedly modest to begin with … [even] if our individual reflections … provide some justification for the beliefs that may seem correct to us’ (2007, 212, my italics).¹⁶ So it is in keeping with the conciliationist spirit to suppose that someone who finds the arguments in support of EW (much) more persuasive than the arguments against it should be diffident about it even before he encounters disagreeing peers.

We can now consider scepticism – the second component of SUP – from this new perspective. Weiner proposes to generalise his result for the equilibrium point allowing for the possibility that the proponents of EW and AS ascribe to their views credences that fall (even significantly) short of 1. He shows that if they ascribe to EW credences x, y, respectively, the equilibrium credence for the proponent of EW is 2×/(x-y + 2). Let us suppose y = 0: the proponent of AS is certain of his view. This simplifies the discussion, and is plausible in its own right, in keeping with the (dogmatic) spirit of AS. Here are some noteworthy points. First, the reduction in the credence of EW engendered by DAD is smaller the smaller its initial value.¹⁷ (It is 0 when x = 0, and attains a maximum (²/₃) when x = 1.) Second, if x > ²/₃, the post-DAD credence of EW will be > ¹/₂. And for some values, its proponent may well count as believing it. So DAD does not inevitably require suspension of belief in EW.

So far, I have been engaged in ‘damage control’, arguing that the implications of DAD aren’t as bad as they are often thought to be. Here, by way of a final conclusion, are some good news DAD brings in its wake, pertaining to the prevalent claim that EW requires suspension of belief about contentious issues (notably, in philosophy). Thus, Feldman (2006, 235) says that ‘the cases that seem to be cases of reasonable disagreement are cases in which the reasonable attitude is really suspension of judgment’. And Christensen concurs (2013, 77): ‘On [conciliatory] positions’, he says, ‘if I hold some philosophical view, for example, and find myself in disagreement with [peers] … I often should become much less confident in my philosophical views, perhaps, in categorical-belief terms, withholding belief on the topic’.

Elga attempts to rebut the claim. He argues that “in messy real-world cases, the disputed issues are tangled in clusters of controversy. As a result, though agents in those examples may count their associates as thoughtful, well-informed, quick-witted, and so on, they often do not count those associates as peers (2007, 493). But the supposition that peer disagreement in philosophy are typically like this is doubtful: many disagreements are between philosophers who share many background assumptions linked to the disputed proposition, enabling them to judge one another vis-à-vis peerhood. But we now have a much less contentious way of contending with the sceptical conclusion. Instead of rebutting it, we can render it far less extreme. Suppose (implausibly) the proponent of EW (call him DC) is initially certain of the correctness of his view, as is his peer, who endorses AS. Suppose they disagree about P, ascribing to it two opposing extreme credences, 0 and 1. In light of my conclusions about DAD, DC is no longer required to ‘split the difference’ on P. Instead, he should first conciliate on EW, ending up with ²/₃ credence. His (weighted average) new credence for P is ²/₃, which may well suffice for belief.

Now suppose, again drawing on a previously gleaned insight, that DC arrives at DAD with a credence 0.8 assigned to EW (0.2 to AS), much more plausible than certainty. Suppose his rival is confident of the correctness of AS. Now DC will respond to DAD by lowering his credence in EW to 0.57, in accordance with Weiner’s generalisation.¹⁸ And when he now comes to accommodate the disagreement on P, his new credence will be .572: a reduction, but less dramatic. And surely, this is as it should be (given fallibilist intuitions). If his initial credence in EW is .7 (still reasonable), the new one will be .518, and the new credence for P - .59. This seems to me as it should be (from the point of view of the proponent of EW): reflecting the (somewhat inauspicious) epistemic reality, and only a conciliationist who has lost the courage of his convictions will refuse to endorse it.

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

The publication of this work was made possible with funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No [870883].

Notes

1. I am using one of the several ways in which the term ‘epistemic peer’ is used in the literature. It applies to a person one (justifiably) takes to be as likely as oneself to get it right about the contested proposition. Pittard (2015), along with others, takes a peer to be a person ‘who seems as qualified as ourselves to assess the proposition’s plausibility’. Nothing hangs on the precise way the term is construed.

2. I follow Christensen (2007), Elga (2007) and Kelly (2010) in construing a disagreement as the adoption by two people of different credences (degrees of belief) with respect to some proposition. The alternative (Feldman, 2007; Kornblith, 2010; Sosa, 2010) is to discuss the issue in terms of the tripartite distinction: belief, disbelief and agnosticism. The probabilistic rendition is preferable. As Kelly (2010, 117) notes, when the disagreement is between someone who believes a proposition and someone who is agnostic about it, there is no possibility of implementing a compromise. And, similarly, there is – within this framework – no way of implementing a compromise for several other profiles of stances. And whether or not compromise is the correct response, it should at least be possible to implement.

3. Wedgwood (2007, ch. 11) argues that one is permitted (although not obliged) to have an ‘egocentric epistemic bias’, a modest bias towards one’s own beliefs. Perhaps no one holds the more extreme AS view, which Elga (2010, 176) aptly labels stubborn and uncompromising.

4. Kelly (2005) comes close to RR, but I do not count him as endorsing it. He concedes that (some) conciliation is required when there are many peers or superiors opposing one’s view.

5. The qualification needn’t concern us here. The cases in which EW does not require ‘splitting the difference’ are those in which one has ‘personal information’ that one’s disputant lacks, on the basis of which, one can justifiably believe that one’s reasoning is much more likely than one’s disputant to be right. I know I am not lying, drunk, short of sleep or high on drugs (Christensen 2007, 2011). We can suppose that our philosophical peers also satisfy these conditions, so EW does seem to require ‘splitting the difference’ with respect to it. The other kind of case is one in which the target proposition is (almost) self-evident, so that disbelieving it is (epistemically) beyond the pale: for instance, ‘1 + 1 = 2’. But views pertaining to disagreement, indeed philosophical views in general, aren’t of this type. Finally, the disagreement may be only apparent, verbal (Fumerton, 2010, 95). But again, the dispute concerning disagreement is not merely apparent. Graves (2013, 95, n. 8) discusses the difference between EW and ‘split the difference’ view more fully.

6. Proponents of EW include Christensen (2007), Elga (2007), Kornblith (2010) and Graves (2013).

7. Decker (2014, 1131) adduces another objection to Christensen’s ‘conflicting ideals view’. It enables the proponent of RR to adopt the mirror-image strategy, claiming that he endorses two ideals: respecting the evidence and respecting evidence about our epistemic error. When he remains steadfast in response to a disagreement, he is opting for the first ideal at the expense of the second.

8. The example is more perspicuous and its discussion less cumbersome if it involves a disagreement between a proponent of EW and someone he takes to be considerably superior to him epistemically. But nothing of substance hangs on this modification.

9. This might lead us to suppose that these are not concerned with epistemic justification. This suspicion will be based on Alston’s, Sugden, and Sherwood (1985, 59) conception of the ‘epistemic point of view’ as ‘defined by the aim at maximizing truth and minimizing falsity’. But I am granting for the sake of the argument Christensen’s supposition to the contrary.

10. There is, seemingly, another case that is problematic for Christensen’s (rational akrasia) strategy: a disagreement with a peer who holds a radical version of RR, and thinks the one who responded correctly should never conciliate. This is not a case in which rationality and accuracy diverge: RR is an evidentialist view. So it seems as if Christensen’s strategy won’t work in this case either. But I want to stay clear of such disputants, because it is not clear what RR prescribes for one who has responded incorrectly to the evidence.

11. It has been argued (Douven, 2009, Kelly, 2010) that permissive justification precludes conciliationism, or at least renders it unjustified. The thought is that the two disagreeing peers may have both adopted a rational attitude, so their disagreement does not provide evidence that they responded incorrectly to the evidence. If this line of reasoning were cogent, it would be incoherent for Pittard to offer a conciliatory suggestion pertaining to DAD that relied on permisivism. But I am persuaded by those (Christensen, 2009) who argue that even if justification is thought to be permissive, disagreement could still provide evidence of a mistaken response to the evidence. Since I do not wish to argue for this here, I merely note Pittard’s commitment to permisivism. Those who think it implausible (for reasons unconnected to conciliationism) will take that as a consideration against his proposal.

12. Decker (2014, 1120) argues that the mere possibility that a principle prescribes inconsistently (prescribes inconsistently in some non-actual situation) doesn’t impugn it. This is plausible for contingent rules, as attests Decker’s example. The father's injunction ‘Always do what your mother says’, which will engender incompatible prescriptions when the mother says, ‘Don’t do what your father says’, is perfectly acceptable in ordinary circumstances. But EW, which is an epistemic principle, is necessary, and applicable in every situation. But even if Decker is right, EW seems to prescribe inconsistently in the actual circumstances, because it is, as a matter of fact, disputed (among philosophical peers).

13. This proposal is reminiscent of solutions in game theory. An equilibrium point is a combination of players’ moves in which no player has an incentive to change his choice upon learning of the others’.

14. A stable credence, c, is one for which the following equality holds: c=c∙1/2+(1-c)∙1. The solution is c= ²/₃.

15. The restriction is required because even the proponent of RR will not suppose that we should be certain about all a priori truths, including those that we have never considered.

16. MacAskill et al. (2020) defend the more restricted claim, that we ought to be humble about the correct moral theory, assign significant degree of belief to views other than the one we favour. The reasons they adduce are that ‘ethics is hard’, and that we are liable to form our views in a biased manner (because of vested interests, for example). Our epistemological views are typically not affected by vested interests. Still, epistemology, too, is hard.

17. The difference between the initial and final values is x²/(x + 2). And in the relevant interval, [0,1] (x is a credence), it increases monotonically (its derivative is positive).

18. EW_new = 2∙EW_old/(2∙EW_old +2)=(2∙0.8)/(2∙0.8 + 2).