Seeking a reflective equilibrium in the face of disagreement

Abstract

How is someone who seeks a reflective equilibrium to respond upon learning that others disagree with her? Regrettably, not much attention has been devoted to that question despite the extensive general discussion about the epistemic significance of disagreement that has taken place in recent years. This paper helps fill the lacuna by exploring possible connections between the relevant bodies of literature. More specifically, I claim that how users of the method of reflective equilibrium should respond to disagreement is crucial to the assessment of the method. According to a common objection, it is flawed because it may lead competent and rational inquirers to arrive at widely divergent and inconsistent equilibria. What I argue is that defensible assumptions about the significance of disagreement permit an advocate of the method to resist that objection.

1 Introduction

How is someone who seeks a reflective equilibrium to respond when learning that others disagree with her? Should the inquirer let such information guide the adjustments she needs to make and, if so, how and when? Regrettably, not much attention has been devoted to those questions despite the intense discussion about the general significance of disagreement in recent years, for example in the shape of the peer disagreement debate.¹ How to respond to disagreement is a contested issue but it is commonly agreed that it has at least some relevance to the justification of our beliefs. To explore what role it should be assigned in the method of reflective equilibrium is therefore an important task.

Part of the contribution I want to make with this paper is simply to help fill that lacuna and to bring out possible connections between the debate about the general significance of disagreement and that of the method of reflective equilibrium (hereafter MRE). However, I shall not attempt a full answer to the question of how MRE-inquirers are to accommodate disagreement as I shall discuss it in a special context. The context is provided by a prominent objection to MRE which I call “the non-convergence objection”.

The point of departure of this objection is the idea that, as Sarah McGrath and Thomas Kelly put it, “a good method for investigating a given domain should lead rational inquirers who impeccably follow that method to converge in their views over time” (2010, p. 341). Kelly’s and McGrath’s point is that this does not seem to hold for MRE and conclude therefore that having reached a reflective equilibrium is not enough for the resulting beliefs to be justified.²

The argument Kelly and McGrath suggest thus involves the following premise.

Non-convergence. Rational inquirers who flawlessly follow MRE would not (in relevant conditions) converge in their beliefs about the investigated domain over time.³

It is easy enough to see how MRE raises concerns about non-convergence. To follow MRE is to adjust one’s moral views to make three sets of elements form a coherent set, namely one’s considered moral judgments, one’s relevant philosophical background theories and a set of moral principles.⁴ However, nothing in the method rules out that users start with different judgments or background theories or that they use different criteria in assessing how coherence is best achieved. For example, some may take the simplicity of a principle to be a more important feature than others in assessing how well it fits with one’s considered judgments. In view of such initial differences, there may seem good reason to expect that MRE-inquirers will end up with radically divergent and incompatible equilibria, even assuming that they are rational and employ the method competently.

The question I shall address is if this picture changes once one considers how rational inquirers should react to the initial differences that potentially might lead them in different directions. How it is reasonable to respond to such differences is what the debate about the epistemic significance of disagreement is all about, of course, and several of the positions that have emerged have implications that appear relevant to Non-convergence. For example, on some views, if we find that our position is a minority one then we have, under certain circumstances, reason to replace it with the majority view.⁵ This is likely to promote convergence. Moreover, the peer disagreement debate concerns how we are to respond when finding that our belief about some matter is rejected by an epistemic peer, where a peer is, roughly, a person who is just as well equipped as us, for example in terms of evidence and inferential skills, to ascertain the truth of the disputed issue.⁶ On certain views, such as most conspicuously that known as “The Equal Weight View” or “Conciliationism”, the appropriate response to a peer disagreement is to conciliate, where this means that we decrease our confidence in the contested belief or abandon it altogether.⁷ Since such revisions reduce the tensions between the systems of the disagreeing peers they may also, arguably, diminish the risks of them ending up with diverging equilibria.

The purpose of this paper is to examine the way of resisting the non-convergence objection just indicated. I shall do so by developing an argument to the effect that rational MRE-inquirers will not, in relevant circumstances, reach radically divergent equilibria, and by defending that argument against certain objections. The argument relies on Conciliationism, which is a controversial position. It is contested by views that instead permit “steadfastness” in some cases of peer disagreement (i.e., that one retains the disputed belief). Some of the objections which I shall address invoke such competitors. My primary strategy in responding to them is to argue that the views in question are not applicable to the cases which are most relevant to Non-convergence.

However, although my main thesis hence is that the argument does indeed provide a compelling response to the most common version of the non-convergence objection, I will also make the further point that it does so at the cost of making MRE vulnerable to another, less familiar version. As I have indicated, the point of departure of the traditional version is the complaint that competent MRE-users might reach incompatible conclusions. This is supposed to make it inadequate by casting doubts on its reliability. The less familiar version focuses on a different type of shortcoming, namely that of not allowing its users to reach any conclusions at all (which is an outcome one may fear that the application of Conciliationism might lead to). That version of the non-convergence objection raises other questions than those that pertain to the traditional version, such as when or in what circumstances a method’s failure to deliver results should be seen as a decisive problem. An upshot of the paper is therefore that the discussion about MRE and its alleged failure to generate convergence needs to be refocused.

The plan is as follows. The non-convergence objection is reconstructed in more detail in Sect. 2, where I also distinguish between the two versions I just mentioned. In Sects. 3 and 4, I develop and state the argument to the effect that rational MRE-inquirers will not, in relevant circumstances, reach radically divergent equilibria. In Sects. 5 and 6, I discuss objections to the argument. The paper is summarized in Sect. 7, where I also make some remarks about the implications of my conclusion and about topics for further discussion.

Before continuing, however, I want to stress one limitation. There are different views about what we may achieve by following MRE. My discussion will proceed on the assumption that the goal of MRE (or at least one of its goals) is moral knowledge and epistemically justified moral beliefs. That is how MRE is usually construed when the non-convergence objection is considered. It should be noted, however, that there are those who explicitly deny that MRE is best understood as aiming at epistemic justification and associate it rather with other aims, such as moral understanding or what has been named “practical” justification.⁸ On those alternative views, the alleged lack of convergence among thinkers who are competently seeking a reflective equilibrium may be less of a concern, but that is not a question I shall pursue.

2 The non-convergence objection

Why do Kelly and McGrath assume that a method aimed at generating knowledge or justification is deficient unless it leads “rational inquirers who impeccably follow that method to converge in their views over time”? An obvious answer is that methods which are supposed to generate knowledge and justification are adequate only if they are truth tracking in the sense that they reliably produce mostly true beliefs. That view offers a simple explanation of why a method’s failure to generate convergence renders it inadequate, for if it fails because it easily permits its users to end up with incompatible beliefs then this shows that it lacks the required reliability. That a method lacks reliability suggests in turn that, even if the conclusions some individual inquirer ends up with by using a method are in fact true, that is just a coincidence. On several familiar views in epistemology, this is enough to conclude that they do not qualify as being knowledge.⁹

This argument from reliability thus provides a strategy for those who want to derive a skeptical conclusion about MRE from its alleged inability to generate convergence. One thing to note about it, however, is that it presupposes a type of non-convergence which is strictly speaking stronger than the one Kelly and McGrath take to be sufficient to invalidate MRE. The appeal to reliability shows that MRE is deficient to the extent that MRE easily permits its users to reach conflicting conclusions However, that a method allows its users to end up in such a situation is just one way in which it could fail to generate convergence. Another is through not leading them to reach any conclusions at all. For if it does not lead them to end up with any conclusions at all then it also does not lead them to converge. In other words, that a method allows its users to reach conflicting conclusions entails that it fails to ensure convergence, but the converse does not hold.

The stronger type of non-convergence can be stated as follows:

Divergence: Rational inquirers who flawlessly and successfully follow MRE would (in relevant conditions) reach radically divergent equilibria (i.e., internally coherent but inconsistent systems of beliefs).¹⁰

Versions of the non-convergence objection which proceed via the argument from reliability are committed to Divergence, which is also the most common way to construe the objection. For example, the conclusion Kelly and McGrath pursue is that having reached a reflective equilibrium is not enough for justification. That is a hard conclusion to reach by arguing just that users of MRE will not reach any equilibria at all.

In addition to versions of the non-convergence objection which appeal to Divergence, one may also imagine those that invoke MRE’s alleged failure to secure convergence in ways that do not entail Divergence. That is the upshot of the less familiar version that I mentioned in Sect. 1, according to which a method may be inadequate in virtue of disallowing its users, in certain circumstances, to reach any conclusions at all (i.e., neither conflicting nor congruent ones). Which circumstances that holds for is not entirely clear, however. Of course, it may be disappointing to learn that a method which is supposed to help us obtain knowledge may in fact take us nowhere. But whether that possibility is decisive depends on whether the problem lies with the method itself or with something external to it, such as the availability of relevant evidence or the nature of that evidence. After all, that the evidential resources available to us at one point is insufficient to reach a confident conclusion through the method does not exclude that we will get sufficient resources at a later point. Think for example of the situation in physics where we still have found no viable way to reconcile the general theory of relativity with quantum mechanics. Surely, this situation does not motivate rejection of the methods that physicists have used to reach where they presently are at.

I shall briefly return to the possibility of developing a plausible objection that invokes forms of non-convergence that do not entail Divergence at the end. However, the main bulk of the paper is about divergence-based versions. Still, there is one observation about versions which do not appeal to Divergence that is relevant also to divergence-based ones.

What the physics example suggests is that if a method does not enable its users to reach decisive conclusions, then this discredits the method only if the circumstances in which the users pursue their inquiry are favorable or even optimal. The same plausibly holds for divergence-based arguments against MRE. Thus, if competent users of MRE reach divergent equilibria then this discredits MRE only to the extent that the conditions in which they have pursued their investigations were favorable. I shall not give a comprehensive list of which those circumstances are, but I shall make one crucial assumption. That is, I shall assume that users of MRE are in favorable conditions only if they have at least some awareness of the disagreement that obtain about the beliefs they try to bring into an equilibrium. It is too much to ask of MRE-inquirers that they must gather detailed knowledge of all the disagreements that pertain to their beliefs, such as knowledge of the exact numbers who hold conflicting beliefs on some matter. However, substantial unawareness of the disagreement would count, I submit, as ignorance of potentially crucial evidence.¹¹ To be thus unaware is not to pursue a reflective equilibrium in favorable circumstances, and the possibility of a scenario where people who are thus unaware reach divergent equilibria is not enough to discredit MRE.

3 Responding to disagreement

The argument I shall develop seeks the conclusion that MRE-inquirers will not, in relevant circumstances, end up with divergent ones. It relies on certain views about how it is rational to respond to disagreement and the purpose of this section is to spell them out.

As for the epistemic significance of disagreement, it is important to see that disagreement can be relevant in more ways than one. A central distinction is that between “first-order” evidence and “higher-order” evidence. First-order evidence in relation to a belief is evidence that pertains directly to its truth. Thus, first-order evidence against a belief suggests that it is false, which means that it threatens to undermine its justification in a very direct way. Higher-order evidence against a belief, by contrast, is supposed to undermine justification without pertaining directly to its truth and instead by indicating that the grounds for it are inadequate. Suppose that we learn, for example, that the observations on which we based our belief that there was a cat in the garden last night were made under the influence of some drug. This may be a reason to reconsider the belief even granted that it does not justify the conclusion that there was no cat.

The above remarks about higher-order evidence do not amount to a full characterization of the concept, and how to define it is controversial, partly because many want to distinguish between higher-order evidence and other types of evidence that affect justification indirectly, such as undercutting defeaters.¹² For my purposes, however, it suffices to note that compelling higher-order evidence against a belief mandates a different type of response than first-order evidence. First-order evidence against the belief that P is a reason to believe not-P. Higher-order evidence against the belief, by contrast, is not a reason to believe not-P but rather to suspend judgment about P or to reduce one’s confidence in P.

The discussion about peer disagreement is standardly carried out under the assumption that it serves as higher-order evidence. A well-known conciliationist is David Christensen. He uses the following familiar case to illustrate his view (2009, p. 757):

You and your friend have been going out to dinner together regularly for many years. You always tip 20% and split the check…and you each do the requisite calculation in your head upon receiving the check. Most of the time you have agreed, but in the instances when you have not, you have taken out a calculator to check; over the years, you and your friend have been right in these situations equally often. Tonight, you figure out that your shares are $43, and become quite confident of this. But then your friend announces that she is quite confident that your shares are $45. Neither of you has had more wine or coffee, and you do not feel (nor does your friend appear) especially tired or especially perky. How confident should you now be that your shares are $43?

What Christensen and Conciliationism say about this case is that you should “be about as confident in $45 as in $43”. The underlying reasoning could be phrased in terms of higher-order evidence: By learning that your peer rejects your belief that the sum is $43 you get evidence that the methods you used to form that belief, or your competence in applying them, are not sufficiently reliable. You should therefore abandon the belief.

More generally, Conciliationism says that whether you should conciliate does not depend on the properties your opponent in fact has (e.g., on her actual level of competence in making inferences, her access to relevant evidence, and so on) but on what you have reason to believe about those properties. If you have compelling, admissible evidence to the effect that your opponent is a peer, then you have reason to conciliate, but if your evidence instead establishes that she is an inferior (i.e., someone who is less well equipped than you in determining the truth of the disputed claim) then you do not. The qualification “admissible” is needed because it is assumed that your evidence may cancel the requirement to conciliate only if it is appropriately independent of the dispute. The idea is that it would be question-begging to dismiss an opponent as being an inferior on the mere ground that she rejects one’s belief or the arguments it is based on. Christensen articulates that idea as follows:

Independence: In evaluating the epistemic credentials of another person’s belief about P, to determine how (if at all) to modify one’s own belief about P, one should do so in a way that is independent of the reasoning behind one’s own initial belief about P.¹³

The phrase “the reasoning behind one’s initial belief” raises tricky questions. It can be interpreted broadly or narrowly, which affects the applicability of Conciliationism. However, nothing in what follows depends on any specific interpretation. The point I want to stress is just that the type of response commanded by Conciliationism is, in accordance with the fact that it treats disagreement as higher-order evidence, abandonment of the disputed belief or a reduced confidence in it rather than the adoption of its negation.

How does the assumption that MRE-inquirers are to apply Conciliationism affect the evaluation of Divergence? Divergence is confirmed to the extent that they would end up with divergent equilibria at the end of inquiry, and the idea underlying divergence-based arguments is that they are likely to do so in so far as there are initial differences between their systems of commitments. What does Conciliationism say about that possibility? We may set aside scenarios in which the initial differences are few and far between, as the remaining overlap may be enough to yield convergence. The more relevant scenarios are those in which the initial differences are extensive. Now, Conciliationism tells MRE-inquirers to ignore their verdicts on the issues they disagree about, provided that they have reason to regard each other as peers. That is going to leave them with less resources when trying to approach a reflective equilibrium, for example in the form of judgments which they can use to adjudicate between competing principles. This is not likely to promote convergence on a unique equilibrium, of course. It rather suggests that the inquirers will not reach any equilibria at all. However, that verdict is, as I have kept stressing, sufficient to disconfirm Divergence.

I shall elaborate this piece of reasoning in more detail in the next section. Before that, however, it should be noted that the fact that disagreement may thus serve as higher-order evidence does not exclude that it can also serve as first-order evidence, Whether it does so in a particular case depends on intricate questions about the epistemic status of testimony and authority, but there are clearly some scenarios in which the adoption of your opponent’s belief is an adequate response to a disagreement.

For example, if we believe that we suffer from some medical condition but learn that an experienced doctor disagrees and judges that our symptoms are better explained by some other condition, then we have a reason to replace our initial assessment with that of the doctor’s. This is because our opponent is an epistemic superior rather than a peer. Opposition from a superior may thus work as first-order evidence. And if we get a second opinion from another superior, who has reached his verdict independently of the first, then the pressure towards adopting the doctor’s view grows.

The latter observation illustrates not only that disagreement can serve as first-order evidence but also that the numbers count. That conclusion is entailed by Condorcet’s famous jury theorem, which applies to cases where a group ponders a question which can be correctly answered with a yes or a no. The theorem implies that the probability that the majority’s verdict is true approaches certainty as the size of the group increases, provided, importantly, that the verdicts of the individuals are relevantly independent of each other and that each has a chance of being correct which is greater than 0.5.¹⁴ What the theorem suggests, then, is that if we initially had a view which conflicts with the majority one then we should not only abandon it but replace it with the majority view, provided that the relevant conditions are satisfied. Note moreover that this is so even if we have no reason to regard the members of the majority as superiors or even peers. Thus, suppose that our chance of being correct is 0.6 while the chance of each of the members of the majority is merely 0.51. Then, if the majority is sufficiently large, we should still adopt the majority view (provided that the verdicts of its members are relevantly independent). So, an upshot of the theorem is that, in some cases, dissent can serve as first-order evidence even assuming that the dissenters are inferiors.

The competence and independence conditions associated with the jury theorem make it hard to apply it to real-life cases. How can we know that the members of some group have formed their beliefs on for example a moral issue independently of each other? Why assume that their chance of being correct is greater than 0.5? Similar problems pertain to Conciliationism. How often do we know that we are at the same level of competence as others?¹⁵ However, I shall avoid those applicability concerns by continuing to focus on hypothetical scenarios, of the type which I deem most relevant to the assessment of MRE. The scenario in question is meant to approximate the actual situation, but its relevance is independent of that supposition.

4 Disagreement and the pursuit of a reflective equilibrium

The aim of this section is to elaborate the ideas about the implications of Conciliationism for Divergence that were sketched in Sect. 3. I begin by characterizing the relevant scenarios more fully.

The scenarios I shall consider do not include those in which there is a considerable overlap between the relevant inquirers’ initial commitments. I also ignore scenarios in which there is extensive initial disagreement but where the relevant MRE-inquirers are substantially unaware of this. The reason for that further restriction is, as I have mentioned, that the lack of such awareness is plausibly seen as ignorance of relevant evidence and thus entails that the inquirers are not in circumstances which are favorable for the use of MRE.

In other words, the scenarios I shall focus on are those in which there is both extensive background disagreement and awareness of the disagreement. There is some overlap, we may assume, in the form of some more or less platitudinous judgments that are (nearly) universally shared. Common examples are that pain is something bad, that parents have prima facie obligation to take care of their children, that it is pro tanto wrong to satisfy a mild desire if this requires killing many innocent people, and that there is at least some moral reason to offer aid to those in distress if such aid is very easily given and comes at very little expense.¹⁶ However, on many other issues, at all levels of generality, the views are divided. These issues include whether a distribution needs to satisfy some structural requirement to be just (such as some egalitarian condition) or if it is enough that it has come about in a certain way, to what extent one may prioritize one’s near and dear, and if the intention of an action is relevant to its moral status. For each of an inquirer’s views about those issues there are many who share it but also many who reject it, although the inquirers may not know the exact numbers.

The dissenters of each inquirer, moreover, are going to include both those she has reason to regard as peers and also, at least in the case of most inquirers, those she should view as superiors or inferiors. To be sure, some thinkers are going to lack superiors. These are the ones who are most plausibly seen as experts, due to their training, general cognitive skills, and so on. Some of those experts have come to accept general theories about moral rightness or our obligations, such as consequentialism, Kantianism, virtue theory, contractualism, but the distribution of the views is such that for each of the theories most of the others reject it.¹⁷

How, should the MRE-inquirers proceed in this type of environment? Let us begin by considering the first step of the process of reaching an equilibrium. This is the stage at which one is supposed to scrutinize one’s initial moral beliefs to identify those that are most considered and which at later stages are to be used for testing candidate moral principles.

One thing to note about MRE is that standard formulations of it do explicitly acknowledge the relevance of higher-order evidence at this stage. The familiar thought is expressed by Rawls in the following quotes (1971, pp. 47–48):

Now, as already suggested, [considered judgments] enter as those judgments in which our moral capacities are most likely to be displayed without distortion. Thus in deciding which of our judgments to take into account we may reasonably select some and exclude others. For example, we can discard those judgments made with hesitation [...] Similarly, those given when we are upset or frightened, or when we stand to gain one way or the other can be left aside.

Considered judgments are simply those rendered under conditions favorable to the exercise of the sense of justice, and therefore in circumstances where the more common excuses and explanations for making a mistake do not obtain… [T]he relevant judgments are those given under conditions favorable for deliberation and judgment in general.

And David Brink makes similar remarks (1987, p. 89):

[W]e take considered moral beliefs to be moral beliefs which are well informed, which result from good inference patterns, which are held with confidence, which are not the result of obvious distorting influences (such as prejudice, social ideology, or excessive self-concern), and which are based on an imaginative consideration of the interests of those involved in the situations which the moral beliefs concern.

The idea is accordingly that if we learn that an initial judgment is influenced by distorting factors, then we should ignore it when seeking a reflective equilibrium. And what Christensen’s mental math case suggests is, precisely, that opposition by a peer is one type of consideration that may give us reason to doubt the reliability of the process which has caused us to form a judgment.

I therefore assume that peer disagreement can, in accordance with Conciliationism, make it implausible to count a moral judgment as being considered. This means that the diversity in the scenario I am considering is a potential reason for an MRE-inquirer to dismiss each of her initial judgments besides those that are (nearly) universally held. She may perhaps resist that challenge in some cases and still include those judgments on the ground that she has reasonable suspicions about the epistemic credentials of the dissenters or reason to think that the judgments are shared by a majority of the experts. But this is in a realistic scenario only likely to hold when the judgments in question are, just like the near-universal ones, either imprecise or weak in the sense that they support strong conclusions only when combined with other substantive assumptions. Those extra assumptions, as well as attempts to make the pertinent judgments more precise (for example by specifying how strong the reason is to offer aid to those in distress and so on), will be disputed.

The set of considered judgments that an MRE-inquirer will be left with after having applied Conciliationism will therefore be seriously impoverished. It will include only a limited number of weak and/or imprecise ones. This creates a difficulty when she proceeds to the second step of the process of trying to reach a reflective equilibrium, namely that which consists in identifying a moral theory (in the form of a set of principles) which is more promising than the others.

The problem is that when the considered judgments of the inquirer merely include a few weak and/or imprecise ones then they will radically underdetermine theory-choice. The inquirer might still try to justify preferring, say, consequentialism over Kantianism by arguing that it provides a better explanation of the remaining judgments. However, she would then have to invoke some rather specific view about when an explanation is better which ascribes relative weights to explanatory values such as scope, simplicity, and conservativeness. Such views must quite generally play a greater role when the evidential basis is slim. And a problem is that there is extensive disagreement about them as well, so they also face a skeptical challenge. Given Conciliationism, this reduces their ability to provide any justification for choosing one set of principles over its competitors.

The problem is greatly amplified, moreover, by the fact that there is, for each of the moral theories the inquirer may select, a majority of the experts which rejects it. The potential significance of that fact is underwritten by Condorcet’s jury theorem. The inquirer may try to justify opting for one of the theories by arguing that the experts’ verdicts are not relevantly independent of each other or that they do not satisfy the theorem’s competence condition. But how plausible would that be in a scenario meant to approximate the actual one? Moreover, the supposition that even the experts’ chances of being right is 0.5 or less is hardly a reassuring thought for someone who tries to justify her moral beliefs.

So, an MRE-inquirer’s considered judgments radically underdetermine theory-choice. They only support a choice given controversial auxiliary assumptions, and there is for each of the theories she is to consider a majority of experts who reject it. I submit that, under these circumstances, the selection of principles to continue working with cannot, given Conciliationism, be justified on rational grounds.

Recognition of that conclusion will reasonably grind an MRE-inquirer’s pursuit of a reflective equilibrium to a halt. The only option seems to be to make a leap of faith and tentatively adopt a theory on non-rational grounds. However, the arbitrariness of that choice creates a problematic form of path-dependence which will stain the outcome. Thus, the other MRE-inquirers in the scenario might make the same choice or different ones. If they make the same choice, then they will presumably not end up with radically divergent equilibria, so we can set that possibility aside. If they make different choices, however, then they may end up with radically different conclusions. But to remain confident about those conclusions while being aware of the conclusions reached by the others is not consistent with Conciliationism, as it would require them to ignore the fact that they must, in reaching and retaining their conclusions, rely on a multitude of assumptions that are denied by the others.

5 Objections

I turn now to some possible objections to the argument developed in Sect. 4. Note that the scenario I consider there is one which is characterized by extensive disagreement, both about particular cases and more theoretical issues. The first objection questions the coherence of this scenario, by invoking a type of argument for no-cognitivism which has been offered by Hare (1952, esp. 146–149), Stevenson (1963), and, more recently, Horgan and Timmons (e.g. in 1991). On those arguments, systematic differences of the kind I imagine, such as those that may obtain between consequentialists and Kantians, indicate that the seemingly opposing inquirers use the terms (“right”, “wrong”, etc.) by which they express their differences to refer to different properties. This would mean that the differences cannot be construed as conflicts of belief, in which case Conciliationism does not command conciliation.¹⁸

That line of reasoning poses no threat to the argument I pursue, however. The arguments in question rest on metasemantical views that are contestable,¹⁹ but they are in any case not going to help someone who wants to defend Divergence. The claim that certain types of moral disagreement are excluded on grounds which have to do with co-reference does not confirm the thesis that different MRE-inquirers would reach incompatible equilibria. On the contrary, it potentially undermines it, as it permits one to deny Divergence even granted that one MRE-inquirer may end up with a system of moral beliefs phrased in consequentialist terms while another ends up with one phrased in Kantian terms (as it implies that their systems are not in genuine conflict with each other).

A second objection is inspired by some observations by Katia Vavova (2014). She claims that Conciliationism is plausible only under an interpretation that significantly reduces its applicability, which potentially challenges my claims about its implications.

As I have stressed, whether Conciliationism tells us to conciliate depends on what we have reason to think about our opponent’s competence. What must be the case for the principle to be applicable? The point of departure of Vavova’s argumentation is a distinction between two types of cases. The first comprises cases in which our evidence confirms that an opponent is a peer and the second comprises those in which it does not rule out that she is one. That a case is of the first kind obviously implies that it is of the second kind as well, but the converse does not hold. For instance, when our evidence is neutral to the opponent’s peerhood status then it does not rule out that she is a peer despite not allowing us to conclude that she is one. The distinction gives in turn rise to two different versions of Conciliationism:

• A. Insofar as the […] evaluation [of our opponent’s epistemic credentials] fails to give me good reason for confidence that I’m better informed, or more likely to have reasoned from the evidence correctly [i.e., that she is not a peer or a superior], I must revise my belief in the direction of the other person’s.

• B. Insofar as the […] evaluation [of our opponent’s epistemic credentials] gives me good reason to be confident that the other person is equally well-informed, and equally likely to have reasoned from the evidence correctly [i.e., that she is a peer], I must revise my belief in the direction of the other person’s.²⁰

Vavova rejects A because she thinks it leads to global skepticism (318). For suppose that we are faced with an opponent who rejects all our beliefs. Then we presumably have no admissible ground for concluding that we are better informed, etc., than the opponent (recall Independence), and A accordingly tells us to give up our beliefs. What I shall focus on, however, is instead Vavova’s claim that the applicability of B is limited. She illustrates that claim by considering the following cases:

A. You and Ann disagree about the permissibility of abortion. You agree on all other moral matters and are on the same side of the political spectrum.

B. You and Beth also disagree about the permissibility of abortion, but you disagree about a myriad of other moral matters as well. You are on opposite sides of the political spectrum.

C. You and Clarisse also disagree about the permissibility of abortion, but you also disagree about every other moral matter. Clarisse is a homicidal sociopath.

Vavova suggests that the disagreement with Ann is the one that gives you most reason to conciliate while the dispute with Clarisse is the one that give you least reason and moreover that this assessment accords with B (but not with A). The reason is that, while the overlap between your moral background beliefs and Ann’s gives you an independent reason to regard Ann as a peer, there is no such reason in the case of Clarisse:

You take someone to be your peer, remember, if you think that, conditional on you two disagreeing, you’re equally likely to be right. If that judgment must be made on grounds independent of the dispute, and if you have nothing independent of your dispute with Clarisse, then you have no reason to take Clarisse to be your peer. (315)

How does this reasoning pertain to the question of whether MRE-inquirers should conciliate in the scenarios examined in Sect. 4? If we apply it to that situation, then we may reason as follows: For every moral issue the MRE-inquirers disagree about there are many other related moral issues that they also assess differently. Independence rules out that they can appeal to that background disagreement in evaluating the credentials of their opponents. This means that they cannot use it for dismissing the opponents as being their inferiors, but what matters to the application of B is that the background disagreement in addition means that they have no admissible positive evidence to the effect that the others are their peers. B does therefore not require them to conciliate and may therefore seem to lack the implications I associate with Conciliationism in Sect. 4.

There are several problems with this objection. One is that the disagreements in the scenarios which I consider are not supposed to be like that with Clarisse, as I assume that there is a core of moral judgments that are universally or almost universally shared. Instead, the scenarios are more like the disagreement with Beth (which incidentally is not dismissed by Vavova as being one in which B is not applicable).²¹

More importantly, the scenarios do not rule out that the inquirers have other, independent ways of evaluating each other’s peerhood statuses, besides that which consists in considering the extent to which their moral background beliefs overlap. They may note for example that the others have no shortage of skills that are generally favorable for knowledge-gathering, such as inferential skills, imagination, open-mindedness and so on. That surely goes some way towards justifying a positive view about their competence also regarding moral issues. The agreement about certain core judgments, moreover, may entitle them to appeal also to domain-specific skills. For example, consider the judgment that the welfare of affected people has at least some relevance to the moral status of an act. That view suggests that empathy is a relevant domain-specific skill since it allows one to identify the concerns and interests of the affected people. So, if they have evidence to the effect that the others have that ability, which is not ruled out by the extensive background disagreement that is supposed to characterize the relevant scenarios, then they have further grounds for regarding them as peers. In other words, the background disagreement does not after all exclude that the people in the scenarios considered in Sect. 4 may have plenty of relevant evidence for viewing each other as peers, in which case B is applicable after all.

One final comment about the objection: An upshot of Independence is that the stage in the process of reaching a reflective equilibrium in which it seems least plausible to think that B is applicable is when the inquirers come close to achieving the necessary level of coherence. Being in a reflective equilibrium requires that one’s moral beliefs are organized so that each is connected to the others through strong and varied relations of mutual support. This means that when the inquirers are approaching a reflective equilibrium there is a diminishing share of their moral background beliefs that may, given Independence, be used to assess the peerhood statuses of their opponents.²² Once they approach the relevant level of coherence, B will therefore to a decreasing extent command conciliation, given that the evidence concerning their general cognitive skills is not enough to conclude that the opponents are their peers.²³ Even granted that conclusion, however, it does not undermine the argument in Sect. 4. For what is argued there is that MRE-inquirers will not, rationally, get to the stage in which their moral beliefs are so well entrenched in their systems in the first place. Attention to disagreement will halt the process at an earlier stage.

6 Further objections: steadfastness

Another way to object to my argument is to question Conciliationism and to invoke a position which permit steadfastness. There are several options for those who want to pursue that line and I shall go through the most prominent ones.

I shall argue that, in the case of most of those views, although the factors they invoke may support steadfastness in some cases, they do not do so in the cases that are most relevant to Divergence. This has partly to do with the fact that they have primarily been designed to handle rather artificial one-on-one disagreements rather than the messier ones which involve more people that MRE-inquirers have to face. The same can be said for Conciliationism, of course, but I have tried to develop the argument in Sect. 4 in a way that makes it less vulnerable to that feature of Conciliationism.

Note also that it holds for several of the competitors to Conciliationism that they do not entirely rule out that there are cases where conciliation is called for. What they entail is that there are factors in some of the cases where Conciliationism commands conciliation that break the apparent symmetry and justify steadfastness, but those factors are not always present and the significance of others depends on the context. The question is therefore, on those views, not so much if conciliation is never required but when it is.

For example, consider the view sometimes named “Justificationism” which is associated with Jennifer Lackey (2008). Lackey invites us to consider a case where someone denies our belief that 2 plus 2 equals 4. She argues that this disagreement gives us no reason to abandon the belief, partly because of our strong antecedent justification in it but partly also because we generally have access to more information about our own competence than about the competence of the others. The view that steadfastness is permitted on those grounds is a violation of Independence and is thus inconsistent with Conciliationism.

Other competitors to Conciliationism appeal to other symmetry-breakers. On Thomas Kelly’s so called “Total Evidence view”, the fact that one’s belief is opposed by an initially plausible peer-candidate does indeed provide a potential reason to reconsider it, but it must be weighed against the support provided by one’s first-order evidence (see, e.g., Kelly, 2010). In some cases, that support may trump the higher-order evidence. This reasoning also violates Independence, which Kelly in addition offers direct arguments against (Kelly, 2013).

More radical views defend steadfastness by invoking the idea that you may have entirely private evidence, in the form of experiences or intuitions that are in some sense incommunicable and thus inaccessible to an opponent (e.g., van Inwagen, 1996). This type of evidence can, it is supposed, cancel the reason for conciliating even if the situation may appear symmetrical in terms of epistemic credentials from a third-person point of view.

I am willing to grant that those views justify steadfastness in some cases. Nevertheless, they do not threaten the conclusion pursued in Sect. 4.

The important thing to note about the views in question is that the cases where the factors they adduce most plausibly tip the scales are precisely of the type instantiated by Christensen’s and Lackey’s one-on-one math cases. Thus, Lackey does indeed have compelling antecedent justification for her belief, in the form both of her own mathematical theorizing and the testimony of others, including master mathematicians. I concede that this makes it reasonable for her to dismiss the opponent. As for the issue which is disputed in Christensen’s case, moreover, it is not one which has been thoroughly considered by thinkers outside of the situation. Indeed, it is new also to those who are involved. The only evidence that seems relevant is the shaky one which presupposes trust in their mental math-capabilities under the circumstances described. To privilege one’s own belief in such cases does not appear irrational, given that some amount self-trust is necessary for any inquiry or the general fact that one has access to more facts which are relevant to one’s own reliability and competence.

What an MRE-inquirer faces, however, is something entirely different. The definition of a considered judgment does not preclude that one has antecedent justification for it (recall Brink’s point in the above quote about “good inference patterns”). But, unless it belongs to the set of core judgments which there is extensive agreement about, the available justification will have to invoke other claims which are also challenged. For example, someone’s justification for judging it wrong to push the big man in the footbridge version of the trolley case (Thomson, 1985) may rely on the view that it is wrong to use a person merely as a means. That view is also contested, not least among experts, which undermines the strength of the justification it can provide and thus its ability to defeat the challenge provided by the dissent.

Moreover, unlike in Christensen’s mental math case, the judgments whose status MRE-inquirers are to reflect on pertain to questions that have been thoroughly examined by countless thinkers, including the most plausible expert-candidates. These are judgments about the relative importance one may assign to one’s own interests, when persons deserve punishment, and so forth. Nevertheless, disagreement remains. In that context, appeals to self-trust or private evidence must, I submit, weigh relatively lightly, especially in view of the fact that, from a third-person perspective, the situation is symmetrical.

That said, there is one steadfast view which may seem to present a more serious challenge. That is the “the Right Reasons view” which in all cases of peer disagreement denies that conciliation is called for, at least when the disagreement is seen as higher-order evidence (see Titelbaum, 2015 for a defense). More specifically, what the Right Reasons View holds is that the justification of a belief is determined by the thinker’s first-order evidence alone, and that higher-order evidence is irrelevant.

The Right Reasons view is furthermore congenial with the so-called “Level-Splitting view”. On that position, we may be justified in holding the (first-order) belief that P while at the same time being justified in holding the (second-order) belief that the belief that P is unjustified. Suppose we get compelling higher-order evidence (for example in the form of peer disagreement) to the effect that the justificatory process behind one of our beliefs is flawed. The Level-Splitting view agrees with the Right Reasons view in rejecting that such evidence gives us a reason to abandon the belief but nevertheless tries to accommodate its significance by in addition holding that it might give us a reason to regard it as unjustified.²⁴

The Level-Splitting view has, just like the Right Reasons view, some very counterintuitive implications.²⁵ However, there are independent reasons for questioning their relevance in the present context. The Right Reasons and Level-Splitting views deny that evidence to the effect that a belief that P has been formed in unfavorable conditions gives the thinker a reason to drop it (although it may be a reason for her to adopt the second-order belief that her belief that P is unjustified). This conflicts directly with the element of MRE according to which we should discard initial moral judgments that are not “given under conditions favorable for deliberation and judgment in general” (Rawls, 1971, 48). The Right Reasons and Level-Splitting views are therefore inconsistent with MRE, which means that the objection they give rise to is separate from the non-convergence objection. The non-convergence objection concerns whether we by following MRE and by applying its epistemological commitments would end up with radically divergent equilibria. Clearly, where we would end up by applying commitments that are inconsistent with MRE does not settle that issue. How to respond to the objection which the Right Reasons and Level-Splitting view provide is accordingly beyond the scope of this paper.

7 Discussion

I have argued that advocates of MRE can plausibly deny that competent MRE-inquirers will end up with radically divergent equilibria in relevant circumstances. That conclusion undermines divergence-based versions of the non-convergence objection.

A remaining question is how to handle versions which are not divergence-based. On those versions, the problem with MRE is that it fails to secure convergence regardless of whether this is so because MRE will lead inquirers to reach divergent equilibria or none at all, since both outcomes count as such a failure. Thus, since the upshot of my argument is that MRE will not lead to divergence because it will not lead them to reach any equilibria at all, it provides little consolation in that context. Indeed, when faced with versions of the non-convergence objection that are not divergence-based, an advocate of MRE is likely to react to my argument by exclaiming: With such friends, who needs enemies!?

In other words, the fact that a method will not lead competent users to reach any result at all may be thought just as embarrassing as that it may lead them to reach inconsistent ones. Personally, however, I am not so sure, on the grounds indicated at the end of Sect. 2. What makes it difficult for rational MRE-inquirers to reach decisive conclusions is that the evidence, including that of the higher-order type, is fragmented and points in different directions. The problem thus lies with the evidence, or if you like “the world”, rather than with the method. In fact, if the method were to generate definite conclusions in such a situation, that would, in my view, speak against it rather than in favor of it.

That stance raises the issue of whether it really is plausible for MRE-inquirers to see higher-order evidence, in the form of information about other peoples’ assessments, as being relevant to answering moral issues. The view that it is not is sometimes voiced, as the Right Reasons View illustrates. Consider for example the debate about the role of intuitions in ethics. Some stress in that debate that it is not the fact that it seems to someone to be wrong, say, to push the big man in the footbridge case that may serve as evidence for or against a moral theory but rather the judgment that constitutes the content of the attitude, namely the claim that the act is wrong (e.g., Singer, 2005, see also Tersman, 2008 for a discussion). However, those who hold that view owe us an answer to why we are entitled to rely on that claim. It is hard to imagine a good answer which does not appeal to the fact that it is the object of a seeming or some other attitude. Answers that do involve such an appeal, moreover, are in turn hard to reconcile with a ban against seeing the attitudes of others as evidence.

Finally, if I am right in claiming that widespread background disagreement makes it difficult for rational users of MRE to reach a reflective equilibrium, does this mean that moral philosophers should stop pursuing that goal? That is not my view. We could still pursue it by bracketing the evidence which consist in information about disagreement. What we achieve by reaching a reflective equilibrium under such circumstances may not be a set of beliefs we are justified in being confident about, but there may be other worthwhile aims of the theorizing. On the view in the debate about philosophical progress named “Equilibrism”, for example, to identify a number of distinct and thought through equilibria is in itself a valuable goal, in spite of the lack of evidential resources to select one of them as the superior one. What we thereby gain is at least a more detailed picture of the theoretical landscape, which in turn promotes a deeper understanding of the questions addressed by the theorizing.²⁶