How to Explain Some Remarkable Features of Evaluative Epistemology By Phillip Galligan B.A., Stanford University, 2008 Thesis Submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the Department of Philosophy at Brown University PROVIDENCE, RHODE ISLAND MAY 2016 © Copyright 2016 by Phillip Galligan This dissertation by Phillip Galligan is accepted in its present form by the Department of Philosophy as satisfying the dissertation requirement for the degree of Doctor of Philosophy Date: __________ _____________________________ James Dreier, Advisor Recommendation to the Graduate Council Date: __________ _____________________________ David Christensen, Reader Date: __________ _____________________________ Joshua Schechter, Reader Date: __________ _____________________________ Russ Shafer-Landau, Reader Approved by the Graduate Council Date: __________ _____________________________ Peter Weber, Dean of the Graduate School iii Curriculum Vitae Phillip Galligan was born in Minneapolis, Minnesota on August 10, 1985. He attended Saint Paul public schools for elementary, middle, and high school, and then Stanford University, where he received a B.A. in Philosophy in 2008. His honors thesis was titled “Nature and Reform in Montaigne.” He then went on to pursue a Ph.D. at Brown University in the Department of Philosophy. He has published one paper, titled “Shame, Publicity, and Self-Esteem.” iv Preface and Acknowledgements I would like to thank quite a few people. Beginning with the most obvious: Thank you to my advisor, Jamie Dreier, not just for his help with the particular content of this dissertation (although he deserves plenty of credit for that) but also for so adeptly pointing out what was wrong with many other ideas that I was, at one time or another, tempted to espouse. Without such regular and generous helpings of criticism, it seems very unlikely that I would have written a dissertation I could be proud of – or even worse, I might have written one that I was proud of for no good reason. It has been a great privilege to spend seven years trying out my thoughts with him. Many thanks also to the other members of my committee: David Christensen, Josh Schechter, and Russ Shafer-Landau. There are some places where it is obvious how much these three papers have been influenced by their work. In particular, Josh’s paper “The Reliability Challenge and the Epistemology of Logic,” along with a course that he and David co-taught, were enormously helpful to me in writing Chapter 1; and I would certainly not have noticed the puzzle that is central to Chapter 2 if I hadn’t learned so much from him about the epistemology of disagreement. Just as importantly, however, it has been valuable for me to have such an excellent group of philosophical role models. It seems to me that a philosopher should express himself as plainly as possible, be open to the possibility that he is wrong, admit it when he doesn’t understand, and treat philosophy as a shared project rather than a contest. Many philosophers are not good at these things, but none of them were on my committee. Of course not everyone who contributed to the content of this dissertation, or to my philosophical development, was a member of that committee. I am grateful to Bernard Reginster for many interesting and enjoyable conversations about Nietzsche and philosophical psychology, without which I would not have even the one publication. I’ve never taken a course with Chris Hill, but somehow a large portion of the nicest things anyone has said to me as a graduate student have nonetheless come from him, and I very much appreciate the encouragement. I have also benefitted from some excellent classmates: Alex King was a great partner to begin studying metaethics with, and Brett Topey has often helped me to refine my thoughts or to understand something about an area of philosophy I am not so familiar with (or just not very good at). Especially in the last few years, it has often been on my mind how lucky I am to have a sufficiently robust support system to complete a Ph.D. at all. Like seemingly all doctoral students, over the last seven years there have been many times that I’ve been deeply discouraged, frustrated, or just plain sad. But it has always been a great consolation to have awesome people who care about me, even when most of them are far away and unaware that they are doing anything to cheer me up. At the very top of that list is my mom, Kathy Galligan, followed by the rest of my family in no particular order: my brother Mac, my sister Eloise, my brother-in-law C.J., and my father David. The distinction between my friends and my family is in many cases merely a biological one though, and I am also very grateful to: Alex Willen, Alice McGlave, Andy Hall, Chintan Patel, Claudia Huntington, David Barrasso, David Green, Jesse Dickinson, Joey Shurtleff, Ken Newman, Lane Barrasso, Marshall Miller, Matt Wilberg, Monica Cendejas, Nilay Patel, Ojus Doshi, Prentice Miller, Rachel Apostoles, Sarah Scheller, Sibylle Barrasso, and Vivek Buch. v Table of Contents Chapter 1: Reliability Challenges and Etiological Undermining in Metaethics 1 Chapter 2: The Peculiar Significance of Evaluative Testimony 23 Chapter 3: Contextualism About Value 48 vi CHAPTER 1 RELIABILITY CHALLENGES AND ETIOLOGICAL UNDERMINING IN METAETHICS 1. Introduction Recently arguments against objectivist theories of value based on the origins of our normative and evaluative beliefs have received a lot of attention. 1 Sharon Street’s (2006) “Darwinian” argument is the most prominent, but it isn’t clear exactly how that argument goes. Street begins with the observation that many of our basic evaluative beliefs (that survival is good, that we should be especially concerned for the welfare of our friends and relatives, and so on) are clearly products of evolutionary design; and the conclusion is that if those beliefs were about an objective, mind-independent subject matter, then there would be something wrong with them - but it’s not clear precisely what the wrongness is supposed to be, or what the intermediate steps in the argument are. There are actually two distinct arguments that might proceed from etiological premises like Street’s, though they might sometimes be run together. One contends that if evaluative facts were objective, then the causes of our evaluative beliefs would have nothing to do with them and it would be an enormous coincidence if most of those beliefs happened to be true. This type of argument has been discussed quite a bit over the last decade, but ultimately it isn’t as strong as it might initially seem; it has no force against some objectivist views, and even against non-naturalist realism - the easiest target for such an argument - it is not very decisive. Still, there is something unsatisfying about the usual objectivist replies to this first argument. My conjecture is that that’s because there is another sort of etiological argument which is not always clearly distinguished from the one above, and which those replies do not address. Whereas the previous argument challenges the objectivist to explain why our evaluative beliefs tend to be true, this is a more properly 1 Usually such arguments are taken to be directed at realists, but actually they are at least as forceful against quasi- realists, and in a way also against views that are too metaphysically noncommittal to be comfortably classified. What they all have in common is that they claim (or in the case of quasi-realism, at least hold that it is metaphysically innocuous to claim) that value is objective. 1 epistemological argument. The question is: How is it that our beliefs are justified, given what we know about their causes? Many of the philosophers working in this area doubt that etiological considerations pose much of a problem for any metaethical theory, but once we understand how etiological undermining works in general we can see why it is a non-trivial task for a theory to avoid the conclusion that an implausibly high percentage of our evaluative beliefs are unjustified. It isn’t that there is no remotely tenable position for an objectivist to take on this issue though - in fact, I explain how some objectivist views can be elaborated to provide pretty satisfying explanations of why our beliefs are secure. Naturalist theories do significantly better than non-naturalist ones in this respect though, and the necessity of providing answers to these questions poses slightly more complicated but nonetheless serious problems for quasi-realism and (supposedly) metaphysically noncommittal forms of objectivism as well. 2. Implausible coincidences and explanatory deficits I won’t attempt a really thorough exegesis of Street’s style of etiological argument, but it will be useful for illustrative purposes to say a bit about it. She begins by noting that there is a strong correlation between what we take to be valuable and the goals and behaviors that would presumably have enhanced our distant ancestors’ chances of successfully propagating their genes. Street provides the following examples: (1) The fact that something would promote one’s survival is a reason in favor of it. (2) The fact that something would promote the interests of a family member is a reason to do it. (3) We have greater obligations to help our own children than we do to help complete strangers. (4) The fact that someone has treated one well is a reason to treat that person well in return. (5) The fact that someone is altruistic is a reason to admire, praise, and reward him or her. (6) The fact that someone has done one deliberate harm is a reason to shun that person or seek his or her punishment. (2006: 115) Presumably the explanation for this strong correlation is that evolution has had a major impact on our 2 evaluative beliefs. 2 This is supposed to be a problem for what Street calls “realism,” but since that is not a very perspicuous term for some of her intended targets, I will use “objectivism” instead. 3 Objectivism: The evaluative facts are mind-independent, which is to say that they do not obtain in virtue of our mental states toward the objects and properties that are their constituents. 4 For example, according to objectivism whether knowledge is valuable for its own sake doesn’t depend on what attitudes we have toward knowledge; if knowledge is valuable, then it would be valuable regardless of how we felt about it. By contrast, an anti-realist might maintain that if knowledge is good, that’s true in virtue of our esteeming it highly. Now we’re ready for the gist of Street’s argument: The challenge for realist theories of value is to explain the relation between these evolutionary influences on our evaluative attitudes, on the one hand, and the independent evaluative truths that realism posits, on the other. Realism, I argue, can give no satisfactory account of this relation. On the one hand, the realist may claim that there is no relation between evolutionary influences on our evaluative attitudes and independent evaluative truths. But this claim leads to the implausible skeptical result that most of our evaluative judgements are off track due to the distorting pressure of Darwinian forces. The realist’s other option is to claim that there is a relation between evolutionary influences and independent evaluative truths, namely that natural selection favored ancestors who were able to grasp those truths. But this account, I argue, is unacceptable on scientific grounds. (2006: 109) This passage mentions almost all of the main elements of the argument; it just isn’t clear how to take Street’s descriptions of them or exactly how they are related to each to other. 5 As I understand it, her argument is that objectivists face a dilemma when it comes to the correlation between the evaluative facts and the beliefs that evolution has led us to hold regarding them: Either admit that there is no explanatory relationship between the two, or offer such an explanation. But if there were no such relationship, then it would be an enormous coincidence if evolution had happened to 2 Even setting aside observations like these, it seems overwhelmingly plausible that evolution has exerted a strong influence on our evaluative beliefs. An animal’s adaptive fitness depends heavily on its behavioral dispositions, so evolution has certainly affected those, and whatever the precise relationship is between a person’s thinking something is valuable and being disposed to act accordingly, it’s hard to see how evolution could have influenced the latter while leaving the former untouched. 3 Street is certainly aware that her argument applies to non-realist views, since elsewhere (2011) she argues that quasi-realists have the same problems. 4 See Street (2006: 110). This way of characterizing realism is drawn from Russ Shafer-Landau (2003: 15). 5 Everyone in this literature seems to agree that the premise about evolution is not really essential; other causes of our beliefs (e.g., cultural ones) could in principle serve just as well. Focusing on evolution makes sense though, since it has had such powerful and often identifiable effects on us, so I will follow Street in this respect (again like most of the literature). 3 supply us with mostly true beliefs - a coincidence so big that we shouldn’t believe it has occurred. So on the first horn of the dilemma, objectivism has the implausible consequence that most of our evaluative beliefs are probably not true. And on the other horn, objectivism is wedded to the empirical theory that the truth of our evaluative beliefs somehow helps to explain why we have them, but it is hard to see how such an explanation could work. So we have a reductio: objectivism is committed to either an implausible skeptical thesis or to a terrible (or maybe just incoherent) theory about why we believe what we do. This argument has attracted a lot of criticism. First, several philosophers have pointed out 6 that it is sometimes reasonable to accept that a big coincidence has occurred; indeed, it would sometimes be crazy to do otherwise. For instance, for most of the NFL’s history the stock market has gone down when an AFC team has won the Super Bowl and up when an NFC team has won. That information shouldn’t convince anyone that there is a significant explanatory relationship between the behavior of the stock market and which conference wins the Super Bowl - should just accept that it’s a remarkable coincidence. Of course, it doesn’t follow that the coincidence between the evaluative truth and what evolution has led us to believe is similarly innocuous, and a theory that posits a large coincidences is usually less credible than one that doesn’t, other things being equal. Whether an implied coincidence is objectionable or not seems to depend on whether alternative theories that are otherwise plausible would be able to explain the phenomena better. 7 In this case, Street would argue that her own constructivist view provides such an alternative explanation. According to constructivism, the evaluative facts are mind-independent; they are determined by (“a function of”) our attitudes. In that case it is no coincidence whatsoever that evolution has mostly led us to have true evaluative beliefs. By causing us to develop positive attitudes toward, for instance, the continued survival of our friends and relatives, evolution indirectly made it the case that their survival is valuable. And we can tell what our own attitudes are by introspection, which completes the explanation of 6Dworkin (1996), Enoch (2011), White (2010). 7Joshua Schechter makes the related observation that the “strikingness” of a phenomenon depends on whether “there is a salient theory that would predict or explain it.” (2010: n. 33) 4 why there is a strong correlation between the evaluative facts and what evolution has led us to believe. (2996: 154) On the other hand though, this is only bad news for the objectivist to the extent that constructivism provides a plausible alternative overall. Maybe it does, but if we can’t determine whether objectivism is committed to an implausible coincidence without conducting a holistic comparison of its merits against those of its rivals, then this is not the decisive, freestanding argument Street presents it as. Instead, it looks more like what Joshua Schechter (2010) calls a “reliability challenge.” In general, a reliability challenge is the task of explaining why it is, in some domain, that we tend to believe true propositions and not false ones. It is hard to believe that it is ever a brute fact or complete coincidence that we are reliable about some subject, so failing to explain why our beliefs tend to be accurate costs a theory some plausibility. 8 Such challenges are particularly difficult for objectivist theories of a priori domains. 9 So maybe no reductio threatens, but if objectivists have no answer to the reliability challenge their theory would nonetheless lose some plausibility. David Enoch’s reply on behalf of his “robust” (non-naturalist) realism is largely to argue that while it is something of a coincidence that our evaluative beliefs tend to be true, it is not as big as Street thinks. Partly that is because we can credit our non- evaluative faculties for much of our reliability; our general ability to detect and iron out tensions among our beliefs can go a long way toward explaining why we are reliable, as long as our basic evaluative beliefs are not too far off to begin with. The other part of Enoch’s strategy is to maintain that we could not easily have had very different evaluative views, and since we are in fact pretty reliable about value, we could not easily have been wrong. This argument actually depends pretty heavily on our knowledge that we are products of 8 The paradigm of a reliability challenge is the Benacerraf/Field problem for platonism in the philosophy of mathematics - see Benacerraf (1973) and the introduction to Field (1989). Street herself notes in a later article that her argument “bears major similarity to” the Benacerraf/Field problem (2016: n. 33), and David Enoch (2011:163- 65) interprets her this way as well. 9 See Schechter (2010: 439-40). 5 evolution, because evolution is what explains why it was almost inevitable that many of our evaluative beliefs would be true; evolution could hardly have led us astray as to the value of survival and reproduction, and the means to those ends, since if we had been indifferent to them we would not have lasted long. And these starting points, plus the aforementioned capacities for working out inconsistencies, explain why we are more or less reliable. (Enoch 2011: chapter 7) We could question how different things would really have to have been for evolution to give us very different evaluative beliefs. Could we not have evolved under circumstances that favored dispositions more like those of bees, and consequently attached little value to our individual lives? Or in circumstances in which it was adaptive to be indifferent to the welfare of our relatives, like sharks? And very few species are as social and egalitarian as we think we ought to be; why couldn’t we have been more like the others in these respects? Another issue is whether Enoch’s sort of defense is unsatisfactorily narrow. It’s only recently that these resources from evolutionary theory have become available for answering the reliability challenge; what does Enoch think the right response would have been for previous generations? It seems strange to think that a few hundred years ago it would have been rational to be less confident that our evaluative beliefs were mostly true, or that the evaluative facts were objective, because we didn’t know about evolution yet. Still, I think Enoch’s is a pretty effective reply to the reliability challenge. It concedes that to some degree we are lucky to have mostly accurate evaluative beliefs, but that degree is much smaller than it first appeared, and it’s hard to see how people who disagree over the size of the remaining coincidence could settle their dispute. There is something else that is unsatisfying about this sort of reply, and it has something to do with Enoch’s reliance on the evaluative assumption that survival is good, but strictly as an answer to the reliability challenge goes a long way toward minimizing the objectivist’s explanatory deficit. 6 3. Etiological Undermining Sometimes it can sound as though Street is thinking of a different sort of argument against realism though, and that may explain why she finds defenses like Enoch’s inadequate. In particular, when she contends (2016) that it is begging the question for a realist to rely on evaluative assumptions - as Enoch does - in replying to her argument, this suggests a more epistemological objection to realism. But on the whole Street seems best interpreted in one of the ways discussed in the previous section, as advancing either a reliability challenge or an argument that realism implies our evaluative beliefs are probably mostly false, so this section will develop an alternative argument for the most part independently of her work. 10 The general idea will be that objectivism combines with what we know about the extent of evolution’s influence on our evaluative beliefs to yield the conclusion that those beliefs are unjustified, whether they are likely to be true or not. It will help to have an uncontroversial example of a belief that is undermined by the subject’s information about its origins. Soccer: Matt is the very proud father of Nathan. Nathan is only seven years old, and he’s just started playing soccer, but Matt thinks he has a natural talent for it. It occurs to him, however, that parents are notorious for overestimating their children - as the parents of Nathan’s teammates seem to be doing - and furthermore that Matt has always wanted to have athletic kids. So there is a good chance (Matt thinks) that his assessment of his son’s soccer skills is the product of a halo effect, wishful thinking, and/or whatever other biases parents are susceptible to regarding their children. 11 These reflections should lead Matt to conclude that he is not in a good position to evaluate Nathan’s potential as a soccer player, and consequently to reduce confidence in his belief. (Which explains why he should hold off on hiring an expensive private coach, turning the garden into a practice facility, and so on.) That isn’t because the evidence of bias is evidence against the proposition that Nathan has a 10 Katia Vavova (2014) disagrees with me here; she interprets Street as making an argument like the one I am about to discuss, the conclusion of which is that realism (objectivism) implies that our evaluative beliefs are largely unjustified. Matt Bedke (2014) gives an etiological argument that if our beliefs were about a non-natural subject matter then they would be unjustified, but he relies on a different sort of undermining principle than I do. 11 The halo effect is a tendency to attribute additional positive qualities to a person one likes for unrelated reasons. See Alfred Mele (2001) for an interesting philosophical account of cognitive biases like these. 7 talent for soccer, to be weighed against whatever reasons Matt has for believing in his son’s ability. Instead, Matt’s suspicions about what might have influenced his belief undermine his justification for it, weakening the degree to which the considerations in favor of the proposition that Nathan is talented support Matt in believing it, but without providing any reason for him to believe the contrary. They militate in favor of agnosticism, not disbelief. 12 We might call factors like the halo effect or wishful thinking “irrelevant influences,” and the threat they pose “etiological undermining” - the undermining of the justification for a belief on the basis of information regarding its causal history. The general form of an etiological undermining argument is this: Generic Schema Call an influence on a subject S's belief that p “irreducibly irrelevant” to that belief just in case it does not affect S's doxastic state regarding p via S's appreciation of considerations that bear on the question with which p is constitutively concerned. Irrelevance: S is justified in believing that S's belief that p is influenced by a factor that is irreducibly irrelevant. Etiology: If S is justified in believing that S's belief that p has been influenced by a factor(s) that is irreducibly irrelevant, then S is justified in believing that S is unjustified in believing that p in proportion to what S is justified in believing the magnitude of the irrelevant influence(s) to be. So S is justified in believing that S is unjustified in believing that p. No Cause For Doubt: If S is justified in believing that S is unjustified in believing that p, then S is unjustified in believing that p in proportion to the degree that S is justified in believing that S is unjustified in believing that p. Conclusion: S is unjustified in believing that p. A couple of key notions need explaining. What does it mean for a belief to be constitutively concerned with a question? And what does the requirement that a legitimate influence operate via our appreciation of considerations that bear on such a question amount to? A belief that p is constitutively concerned with the question of whether p. This means that part of 12 If Matt has in fact fallen victim to some sort of bias, then whether he is aware of the influence or not, his belief is already less than fully justified. This paper focuses on subjects who have evidence of such influences though. The classic examples of this type of undermining, like the religious person who knows she would have had different beliefs if she had grown up in a different family or the philosopher who believes in the analytic/synthetic distinction but knows that he wouldn’t have if he had gone to a different grad school - both of which are from G.A. Cohen (2000: chapter 1) involve disagreements. That can make it difficult to disentangle the epistemic significance of information about the causes of one’s belief from information about what other people think - White (2010) even maintains that the former is wholly reducible to the latter. But while disagreement is a common source of evidence regarding why we believe as we do, it isn’t the only source, and cases like Soccer show that causal information can impact what we are justified in believing on its own. 8 what it is for an attitude to be a belief that p is for it to be the product of an attempt to determine whether p. Pamela Hieronymi has exploited this insight in attempting to sharpen the distinction between the “right kind” and the “wrong kind” of reasons for belief that p, where the right kind includes, for instance, evidence that p, and the wrong kind includes reasons merely to want to believe that p, like the fact that a powerful demon will punish you if you don't believe it. Her suggestion is that the right kind of reasons are constitutive in the following way: “The constitutive reasons are those considerations that (are taken to) bear on a question or set of questions, the settling of which amounts to having the attitude.” (2005: 457) I've adopted her distinction as a way of elaborating the notion of an epistemic reason. This is useful in order to distinguish influences the knowledge of which undermines justification for a belief from those the knowledge of which enhances that justification. Knowledge of an influence is compatible with or enhances a subject’s justification for his belief that p when he has reason to believe it affects his doxastic state only because it is partly responsible for his appreciation of considerations that bear on whether p. That may be because the influence is itself such a consideration, because it causes the subject to be aware of such a consideration, or because it contributes to making the subject appropriately receptive to such a consideration. An influence on a belief that p may be legitimate even if its relationship to a consideration that bears on the question of whether p is indirect. The fact that I have read the Wikipedia page on David Hume does not immediately bear on the question of whether Hume was born in 1711, but I am confident that this fact has affected my belief about Hume only because the article says Hume was born in 1711, which is a consideration that I take to bear on the question of when Hume was born, since Wikipedia is presumably a reliable source of basic information about Hume. The fact that Jay Bilas has been a professional college basketball analyst for a long time does not bear on the question of whether the University of Kansas will win the NCAA men’s basketball tournament, but if this fact influences his belief that Kansas will win only because his experience has made him an excellent evaluator of a team's prospects, then knowledge of the influence should not reduce his confidence in that belief. An etiological argument relies on the subject's having reason to believe that there is some influence on his belief that 9 does not operate through any such channels. We can also capture the gist of an etiological argument this way: It’s not rational to maintain a certain level of confidence that p if one is also justified in believing that that level of confidence is irrational. To learn that one’s belief has been affected by an irrelevant influence is to learn that one’s confidence is not what it would have been if it were rational. And that rationally requires modifying one’s credence. 13 Usually the process of reasoning that someone in Matt’s position would go through is much less complicated than what’s spelled out in the schema. That’s because it’s generally obvious whether a suspected influence would be irrelevant or not, so the main question is just whether it has had any effect on the subject’s belief. Matt is unlikely to wonder whether the fact that he would prefer that Nathan be an excellent soccer player is a good reason to believe that he is; instead Matt will focus on whether his desire has influenced his belief. The question of whether an evaluative belief is undermined by information about its etiology is more complicated because theories of the nature of value have different implications regarding which influences are irrelevant. This sort of etiological argument contends that if objectivism were true, then our evaluative beliefs would be severely undermined by what we know about their causes. But in fact it is quite implausible that we should have so little confidence in our evaluative beliefs, let alone for this particular reason. So realism has a very implausible consequence, and we should reject it. Here’s a more formal version: Evaluative Undermining (EU) Evolution: Suppose that S is justified in believing that S's belief that some evaluative proposition V is strongly influenced by what has historically been fitness-enhancing. Irrelevance: Suppose that S is justified in believing that the influence of what has historically been fitness-enhancing on S's belief that V is irreducibly irrelevant. Then S is justified in believing that S's belief that V is strongly influenced by a factor that is irreducibly irrelevant. 13 It’s a little more complicated than this, actually. Suppose S is agnostic about whether p, and then discovers that she is subject to an irrelevant influence. If S has no idea whether that influence increased or decreased her credence in p, then it looks as though this information shouldn’t make her any more or less confident, so her final credence should be the same as her initial credence. In general though, etiological arguments push us toward agnosticism unless we can tell what the valence of the irrelevant influence is. 10 Etiology: If S is justified in believing that S's belief that p has been influenced by a factor(s) that is irreducibly irrelevant, then S is justified in believing that S is unjustified in believing that p in proportion to what S is justified in believing the magnitude of the irrelevant influence(s) to be. So S is justified in believing that S is largely unjustified in believing that V. No Cause For Doubt: If S is justified in believing that S is unjustified in believing that p, then S is unjustified in believing that p in proportion to the degree that S is justified in believing that S is unjustified in believing that p. Conclusion: S is unjustified in believing that V to a high degree. The possibility that many of our evaluative beliefs might be undermined in this way looks like a problem for objectivism because Irrelevance is by far the weakest premise. It’s hard to deny that evolution has had a strong influence on our judgments, and the rest of the argument is just adapted from the general schema. So we’d better deny that evolution has been an irrelevant influence - that is, maintain that information regarding what is fitness-enhancing has some genuine epistemic bearing on what is good - and it may look as though no objectivist can do that. A constructivist like Street has no such problem. If what is valuable is somehow a function of our attitudes, then evolution is related to the content of our evaluative beliefs in the same way that it is related to the content of our beliefs about what’s pleasant. Evolution has played a large role in causing us to have the attitudes we do, and since the facts about what is valuable - like those about what’s pleasant - are determined by our attitudes, evolution has indirectly helped to determine what the evaluative facts are. In general, this makes the influence of evolution on our evaluative beliefs quite innocuous, since in most cases we can be confident that it has operated only by way of good epistemic reasons. 14 Later on I will be arguing that EU actually has no force against some forms of objectivism - but not for the reasons the opponents of etiological debunking arguments usually cite. Whereas objectivism is usually defended against etiological concerns on the grounds that any such argument must depend on a principle that has implausibly skeptical implications in other domains, or that it must ultimately be question-begging, I don’t think there is any such flaw here. Instead objectivists should accept that arguments of this form are potentially sound and cogent; they should just supply substantive explanations (different ones, according to the details of their theories) of why Irrelevance is false. Before I get to that 14 See Street (2016: 12). 11 though, I will explain why this form of argument doesn’t have the structural problems many philosophers have thought must afflict any argument from etiological premises to metaethical conclusions. It might seem as though EU depends on the view that a belief that p can only be justified if it is explained by the fact that p. That would be very bad news for the argument, since this view seems to imply that we can’t have any justified beliefs about the future or about necessary truths, and that no false belief could ever be justified. 15 But EU doesn’t presuppose any such thing. The relationship between causal history and justification that it relies on is just this: A belief is justified only to the extent that it is the product of good epistemic reasons. And a belief that p might be produced by good reasons for the subject to believe that p regardless of whether p itself explains that belief - I believe that it will rain tomorrow because that’s what the forecast says, which is a good reason and so justifies my belief even though the fact (if it is a fact) that it will rain tomorrow can’t possibly have had any effect on my belief. Some other popular lines of objection are related to skepticism. One of these is simply that any argument from etiological premises to the conclusion that our evaluative beliefs are unjustified (or likely to be false) would also support skepticism about, e.g., the external world. For instance, you might think that what the would-be debunker demands is an “external,” non-evaluative justification for our beliefs, relying on a principle like Independent Vindication: S is justified in believing that p only if S is justified in believing that S is justified in believing that p. This would explain why our evaluative beliefs aren’t justified; if we aren’t justified in believing that those beliefs have whatever sort of causal history would justify them, then this principle entails that they are themselves unjustified. But Independent Vindication is indeed a skeptical principle, since it would require an infinite chain of justifications of anyone who is justified in believing a single thing. That is not a very plausible requirement, and anyway it is hardly distinctive of evaluative beliefs that they can’t 15Benacerraf (1973: 671-72) relies on an epistemological view like this in using a relative of EU against platonism about mathematical facts. White (2010: 582-83) and Dworkin (1996: 119-20) suggest that an attraction to such an epistemology may be behind etiological arguments, and both provide counterexamples. 12 satisfy it. 16 EU doesn’t employ any skeptical principles though. No Cause For Doubt looks a little bit like one, but crucially, what NCFD says is that if S is positively justified in believing that his belief is not justified, then that belief is unjustified. This is far from an impossible requirement; our beliefs only fall short of it when we have evidence of our own irrationality. In trying to formulate a plausible reductio of objectivism based on the threat of etiological undermining, Vavova progresses from a principle very similar to Independent Vindication to one that is almost identical to NCFD. She thinks the latter is a good principle (she names it “Good”), but can’t see how to use it against objectivism. How can we show that objectivism entails that our beliefs aren’t justified? If we had some way of establishing that evolution is likely to have led us astray then maybe we could show that knowledge of evolution’s influence would require us to reduce confidence in our beliefs. But we cannot determine if we are likely to be mistaken about morality if we can make no assumptions at all about what morality is like. To hold that the moral truths do not coincide with the adaptive judgments, we must assume something about what those moral truths are, or are like. (2014: 91) ...and if the would-be debunker has to assume that many of our evaluative beliefs are false in order to make his argument work, he is unlikely to persuade many people. EU certainly doesn’t make any such assumption, so how is it supposed to work? The key point here is that undermining a belief doesn’t require showing that that belief is likely to be false. It is sufficient to give the subject evidence that her credence is not rational, which can be done just by pointing out an influence that is not (reducible to) a good reason. Vavova might reply that whether something is a good reason for an evaluative belief is itself a first-order, substantive issue, so this approach would have to be question-begging. That isn’t right though. Suppose I believe in the analytic/synthetic distinction, but you don’t, and you present me with strong evidence that to some extent my belief is just due to the greater charisma of the philosophers on the pro- distinction side. This ought to be enough to convince me to revise my credence; but the closest thing to a 16 Vavova thinks Street relies on a very similar principle, and points out that it leads to a very general skepticism. 13 substantive assumption about the analytic/synthetic distinction that it relies on is that the greater charisma of those who accept the distinction is not a good reason for me to accept it myself. Even people with opposing views on a substantive question can usually agree to a large extent on which influences are irrelevant. In the next section we will also see some examples of higher-order, metaethical beliefs that help to justify the lower-order ones. Finally, and relatedly, it might seem that etiological information can’t have the undermining effect EU alleges because only an internal tension among our existing evaluative beliefs can provide us with reason to revise them, and etiological considerations generally have no bearing on whether there is any such tension. 17 Call this the autonomist view of evaluative epistemology. There are at least two ways of understanding this view, however, and the more plausible one actually doesn’t ground any objection to EU. Strong Autonomism: An evaluative belief can neither be rebutted nor undermined by non- evaluative considerations. Weak Autonomism: An evaluative belief can’t be rebutted by non-evaluative considerations, but it can be undermined by them. The view that what we know about the causal histories of our beliefs threatens to undermine our justification for them is only at odds with Strong Autonomism. And I’m not sure what there is to be said in favor of that version over the weaker one - presumably autonomism is derived from the “fact/value gap,” but the weak version doesn’t license us to draw first-order evaluative conclusions from purely descriptive premises any more than the strong version does. Furthermore, Strong Autonomism seems to be vulnerable to independent counterexamples. For instance, if I have evidence that my belief that material wealth is not very valuable is the product of a “sour grapes” mechanism which causes us to reject what we have little hope of attaining, that does seem to require that I revise my belief. So the plausible version of autonomism is the one that is compatible with the possibility of etiological undermining. 18 17 This view is behind much of Dworkin’s (1996) defense of the objectivity of morality, and some of Roger White’s (2010) arguments are in the same spirit. See Kelly and McGrath (2010) for arguments that there is more to normative epistemology than this. 18 Of course, in a manner of speaking they do because the proposition that a belief is unjustified is a normative one. But this is just an epistemic assessment of the belief; it has no implications regarding the truth value of that belief’s 14 4. How Objectivists Can Defuse Etiological Concerns I’ve presented Street’s coincidence argument, the reliability challenge, and the threat of etiological undermining as arguments against objectivism in general, and each of the replies I have considered relies on premises that are compatible with any such theory. It turns out that objectivists can do a lot to mitigate the force of the first two arguments without relying on any special features of their particular views, but the threat of etiological undermining discussed in the previous section seems not to be diminished by any such considerations. The differences between forms of objectivism are actually quite important here, however. In this section I will show how naturalists can successfully address all kinds of etiological concerns head on, rather than trying to evade them or do damage control. Non-naturalists do not fare so well against the undermining argument, though the problem they ultimately confront is not as serious as it might initially appear. (And since quasi-realists are essentially quasi-non-naturalists, they face the same problem.) Finally, “relaxed” or metaphysically quietist forms of objectivism earn an incomplete: until they have said more about what they take value to be, we can’t determine whether they are vulnerable to the undermining argument or whether they can do better than Enoch’s moderately concessive answer to the reliability challenge. 4.1 Vindication for Non-Reductive Naturalists Suppose that evaluative terms refer exclusively to natural properties, though not because they are analyzable in purely descriptive language. Instead, they get their reference the same way that natural kind terms do, picking out whatever underlying natural property offers the dominant causal explanation for our usage in the actual world. So just like “water” refers to whatever substance gives rise to our impressions of wetness and clarity, satisfies our thirst, and so on, “good” refers to whatever property stimulates the content. 15 impressions that regulate our application of that word - the natural property, maybe, that is dominantly responsible for arousing attitudes like admiration and desire in us. 19 So as a matter of linguistic competence we know (at least implicitly) that something is good iff it possesses the property that generally causally regulates the relevant attitudes. (Maybe this much is analytic.) That allows us to use our attitudes as evidence of what is good, and this forms the basis for fully satisfying answers to both the reliability challenge and the question of how it is that our beliefs are produced by good epistemic reasons. First, the etiological issue. Evolution has had a strong effect on our evaluative beliefs, but this effect has been achieved by giving us the attitudes we can use as evidence of what is good. So the influence of evolution has gone by way of good epistemic reasons, and of course learning that our beliefs have been determined by good reasons does nothing to undermine our justification for them. And given this account, of course it is no coincidence that we are fairly reliable on evaluative matters; we tend to have true evaluative beliefs because we are reasonably good detectors of our own attitudes (and those of other people in our linguistic communities). Presumably there is some evolutionary explanation for this related to the value of being able to predict other people’s behavior, but whatever the scientific explanation is will not threaten such an account of our reliability. 4.2 ...and for Analytic Naturalism Analytic naturalists 20 have equally satisfying explanations available. On this view, evaluative properties are wholly analyzable in terms of natural ones; if the property of being good is the property of maximizing human happiness, for instance, this is something we can discover a priori, as a conceptual matter (though a full analysis might be exceedingly difficult). The explanation for our reliability as judges of value is then just the explanation for our ability to detect that natural property. And presumably the 19 Richard Boyd (1988) may be the best example of this sort of view, though it is specifically a variety of moral naturalism. 20 Frank Jackson (1998) is a good example, though again his focus is on morality. 16 main way evolution has affected our beliefs is (as above) via determining what we have chosen to use our evaluative terms to refer to, so there is no general reason to worry that it has influenced our beliefs independently of reasons that genuinely support them. 4.3 Non-Naturalism It’s difficult at best to give a very illuminating explanation of what non-naturalism is, but fortunately some of its entailments are much clearer. The only one that is relevant here is: Non-naturalism: Evaluative properties are not identical to physical properties, and so do not stand in any causal or constitutive relations with the physical world. This not supposed to be a definition, but no view that denied it would seem to count as a form of non- naturalism. Enoch’s reply to the reliability challenge (from Section 2) is the best a non-naturalist seems to be able to do on that front; since it is a central commitment of the view that evaluative facts don’t stand in any causal or constitutive relationships to anything in the natural world, such a theorist can’t invoke the sorts of explanatory links between the beliefs evolution favored and the evaluative facts described above. Still, this failure to completely explain our reliability does not, on its own, seem to be a fatal flaw. It might seem as though the undermining argument is a much more serious threat. If our basic evaluative beliefs have a wholly non-natural (and so acausal) subject matter, then how could any possible cause of such a belief be a good reason for it? Usually, a consideration X qualifies as a good reason for believing that Y in virtue of one of the following: (1) X is justifiably believed to be likely to cause (or have caused) Y: The fact that Alice exercises frequently is a good reason to believe that she is healthy. (2) Y is justifiably believed to be likely to cause (or have caused) X: The fact that Jack said so is a good reason for believing that Alice is at the gym. (3) X is justifiably believed to (at least partly) constitute Y: The fact that Alice is good at math is a good reason to believe that she is smart. (4) Y is justifiably believed to (at least partly) constitute X: The fact that Jack is a bachelor is a good reason to believe that he is not married. (5) Or: X and Y are indirectly related in some combination of the foregoing ways - e.g., X is likely to have been caused by Z, and Z is likely to cause Y. 21 21 I don’t claim that this is an exhaustive list of the ways for X to be a good reason for believing that Y; it is open to 17 But non-naturalism rules out all of these relationships between evolution and the evaluative facts. However, even where there is no known explanatory relationship, X can be a reason to believe that Y just because X is justifiably believed to be correlated with Y. For instance, given the unexplained but robust tendency for my coffee shop to be crowded on Tuesday mornings, the fact that it is Tuesday morning is a good reason to believe the coffee shop is crowded. So the good news is that if a correlation between the evaluative facts and the beliefs evolution has caused us to have could be established, then the fact that evolution has favored a certain evaluative belief would be a reason to think that belief is true. The bad news is that we would need an explanation of how we could establish such a correlation, and for that we need to be justified in some evaluative beliefs independently of our justification for any belief with a natural subject matter. This amounts to saying that the non-naturalist would have to adopt an intuitionist epistemology. There are different ways such an account might go, 22 and this is not the place for extensive arguments against them, but suffice it to say that positing a distinct source of non-inferentially justified beliefs about a non-natural subject matter, in addition to the more familiar ones that other theories of value can make do with, is a commitment few philosophers would be happy to take on. 4.4 Quasi-Realism and Relaxed Realism The foregoing sections don’t address every form of objectivism. Quasi-realists take the question of whether some evaluative proposition is objectively true to be a first-order one, so if we understand their theory as a strictly metaethical one then it is not itself a form of objectivism. They do try to make room for objectivism as a reasonable substantive view, however, for instance by translating the claim that anyone attempting to explain why some element in the etiology of a certain belief doesn’t undermine that belief to identify another sort of relationship that makes that element a consideration that bears on the relevant question. I doubt there is any such relationship that could be of use to a non-naturalist though, disconnected as non-natural properties are supposed to be from the ones we have access to in more familiar ways. 22 See Shafer-Landau (2003: chapter 11) and Audi (2004: chapter 2) for views according to which self-evidence is the source of this justification, and Huemer (2005) for a version according to which our intuitions are “intellectual seemings” somewhat like perceptual impressions. 18 something is objectively good as expressing a plan to pursue or promote it even in hypothetical circumstances in which one’s attitudes toward it are not so positive. (Gibbard 2003) This raises questions about our reliability and the epistemic status of our beliefs on objective matters, however. And although quasi-realism once again doesn’t come packaged with any particular answers to these questions (categorizing them as just more first-order issues), we might worry that 1) the questions do need answers and 2) once we have combined our quasi-realism with positions like those in 4.1-4.3, it will be unclear what the point of being a quasi-realist is. (E.g., we could adopt the - first-order? - view that it is analytic that “good” refers to a certain natural property, and so gain the advantages of a view like Jackson’s. But a major part of the motivation for quasi-realism is usually taken to be that it allows us to avoid such allegedly implausible semantic views.) 23 Similarly, “relaxed realists” 24 like Scanlon (2014), Parfit (2011), and Dworkin (1996, 2011) take it that we can retain objectivism without adopting any controversial metaphysical views. But that would leave us with no explanation for, e.g., the rationality of maintaining our evaluative beliefs in the face of etiological undermining arguments. Since the view that this sort of thing is just a brute fact is not very attractive, relaxed forms of realism seem to be incomplete; they just refuse to explain some of the things that a metaethical theory ought to cover. 5. Conclusion This paper distinguishes and attempts to clarify two ways etiological premises can be used to argue against metaethical theories. The first of these, which is the one I have attributed to Sharon Street, contends either: a) that a target theory is committed to viewing the fact that our evaluative beliefs tend to be true as an enormous coincidence; or b) that such a theory at least loses a great deal of plausibility for failing to explain the remarkable fact of our reliability. 23 Street (2011) argues along these lines that quasi-realism does no better than real-realism at allaying etiological concerns. 24 This excellent name is from McGrath (2014). She also contends that there is ultimately no way for an objectivist to dodge the sorts of questions I’ve been discussing. 19 Alternatively, we might argue that a certain theory entails that many of our evaluative beliefs are unjustified because their causes are not good reasons. In some ways this is a more threatening form of argument, but most of the philosophers who have weighed in on it doubt that we can draw any metaethical conclusion in this way. I have argued that my preferred way of articulating this form of argument has none of the structural flaws these philosophers have suggested it must - the general principles it relies on are quite plausible. Some views can meet this threat head on, however, by explaining what makes the causes of our beliefs epistemically relevant to their contents. Naturalist forms of realism do quite well in this respect. Non-naturalists are forced to take an unappealing intuitionist view of the way our basic evaluative beliefs are justified, but that may not be a decisive strike against them, and at least they have something to say about these issues. Quasi-realists and relaxed realists, on the other hand, simply refuse to address them. 20 References Audi, R. (2004). The Good in the Right. Princeton, N.J.: Princeton University Press. Bedke, M. (2014) No Coincidence?* In R. Shafer-Landau (ed.) Oxford Studies in Metaethics 9, 102-25. Benacerraf, P. (1973) Mathematical Truth. The Journal of Philosophy 70, 661-679. Boyd, R. (1988) How to be a Moral Realist. In Sayre-McCord, G. (ed.) Essays on Moral Realism. Ithaca: Cornell University Press., 187-228. Cohen, G. A.( 2000) If You’re an Egalitarian, How Come You’re So Rich? Cambridge, MA: Harvard University Press. Dworkin, R. (1996) Objectivity and Truth: You’d Better Believe It. Philosophy and Public Affairs 25(2), 87-139. ______. (2011) Justice for Hedgehogs. Cambridge, MA: Harvard University Press. Enoch, D. (2011) Taking Morality Seriously: A Defense of Robust Realism. New York: Oxford University Press. Field, H. (1989) Realism, Mathematics & Modality. Oxford: Blackwell. Gibbard, A. (2003) Thinking How to Live. Cambridge, MA: Harvard University Press. Hieronymi, P. (2005) The Wrong Kind of Reason. The Journal of Philosophy 102(9), 437-57. Huemer, M. (2005) Ethical Intuitionism. New York: Palgrave Macmillan. Jackson, F. (1998) From Metaphysics to Ethics: A Defence of Conceptual Analysis. Oxford: Clarendon Press. Kelly, T. and McGrath, S. (2010) Is Reflective Equilibrium Enough? Philosophical Perspectives 24, 325- 59. McGrath, S. (2014) Relax? Don’t Do It! Why Moral Realism Won’t Come Cheap. In Shafer-Landau, R. Oxford Studies in Metaethics 9. Oxford: Clarendon Press, 186-214. Mele, A., (2001) Self-Deception Unmasked. Princeton, NJ: Princeton University Press. Parfit, D. (2011) On What Matters. Oxford: Oxford University Press. Scanlon, T. (2014) Being Realistic About Reasons. Oxford: Oxford University Press. 21 Schechter, J. (2010) The Reliability Challenge and the Epistemology of Logic. Philosophical Perspectives 24(1), 437-464. Shafer-Landau, R. (2003) Moral Realism. New York: Oxford University Press. Street, S. (2006) A Darwinian Dilemma for Realist Theories of Value. Philosophical Studies 127(1), 109-166. ______. (2011) Mind-Independence Without the Mystery: Why Quasi-Realists Can’t Have It Both Ways. In Shafer-Landau, R. (ed.) Oxford Studies in Metaethics 6. Oxford: Clarendon Press, pp. 1-32. ______. (2016) Objectivity and Truth: You’d Better Rethink It. Forthcoming in R. Shafer-Landau, Oxford Studies in Metaethics 11. Oxford: Clarendon Press. Vavova, K. (2014) Debunking Evolutionary Debunking. In Shafer-Landau, R. (ed.) Oxford Studies in Metaethics 9, 76-101. White, R. (2010) You Just Believe That Because… Philosophical Perspectives 24, 573-615. 22 CHAPTER 2 THE PECULIAR SIGNIFICANCE OF EVALUATIVE TESTIMONY 1. Introduction Alice and Ben are hiking in the wilderness. Ben climbs a hill while Alice is taking a break, and when he comes back he tells Alice that he saw a bear in the distance. Alice wants to see the bear, so she goes back up the hill with Ben - but what Ben thinks is a bear just looks to Alice like a vaguely bear- shaped rock. When Ben first informs Alice that he’s seen a bear - assuming she doesn’t have any evidence that he’s joking, or that there’s something wrong with his vision - Alice ought to significantly raise her credence that he saw a bear. And even once she’s discovered that what looks like a bear to Ben just looks like a rock to her, Alice should still give much more credence to the proposition that it’s a bear than would’ve been rational in the absence of Ben’s testimony. Now suppose Caitlin and Dan are discussing the merits of various philosophers’ lives, and it emerges that Dan thinks Nietzsche’s life was better than Montaigne’s. Upon reflection, though, it seems to Caitlin that Montaigne’s life was better. Further discussion reveals that they are at an impasse, as neither she nor Dan can produce an argument for his or her view that the other finds compelling - even after Caitlin has presented all of the information she takes to be relevant to determining which life was better, and Dan has done the same, the conflict remains. Given all the information at their mutual disposal, Montaigne’s life just seems better to Caitlin, while Nietzsche’s just seems better to Dan. If Caitlin’s epistemic predicament is relevantly similar to Alice’s, then initially - before she’s thought about the issue for herself - Caitlin should be inclined to take Dan’s word for it that Nietzsche’s life is better, and even once she’s arrived at her own conflicting judgment, she ought to be much less confident in that judgment than she would’ve been if she hadn’t been aware of what Dan thinks. But is that right? While a lot of care needs to be taken when comparing cases like these, it seems to me that Dan’s 23 testimony does not have the same epistemic significance for Caitlin that Ben’s does for Alice. In fact, depending on the details of her circumstances, Dan’s testimony may not give Caitlin any reason whatsoever to be less confident in her opinion. But it may not be obvious that there is such a deep contrast between Alice’s predicament and Caitlin’s, so I’ll begin by arguing that it is a genuine phenomenon. Then I’ll examine some explanations of the difference which are consistent with maintaining that one or the other must have a false belief. I think these explanations are all badly flawed. Instead, accounting for the special significance of evaluative testimony seems to require adopting a non-absolutist theory of the semantics of evaluative terms. This might take either of two forms: a contextualist view according to which predicates like “is good” function to attribute a variety of different properties, depending on their context of use; or a relativist one according to which such a predicate always attributes the same property, but whether it is true that an object has an evaluative property varies from one perspective to another. 25 2. The epistemology of ordinary, non-evaluative disagreement Again, my suggestion is that in light of Ben’s opinion that the distant object is a bear, Alice should not be very confident that it isn’t a bear, even though it looks like a rock to her. That is, Alice should “conciliate” with Ben - she should be less confident that the object isn’t a bear than it would have been rational for her to be in the absence of Ben’s testimony. Caitlin, however, should not be similarly concerned when Dan disagrees 26 with her over whether Nietzsche or Montaigne had the better life. 25 Two important notes on the scope of this paper: First, my main examples will all involve evaluative beliefs, but I take it that practical normative epistemology is the same in every relevant respect (and, in fact, that other domains like aesthetics and humor are similar as well). It’s hard to see how these domains could require very different treatments, since they are tightly interrelated. What we ought or have reason to do and what would be best are clearly not independent issues, but these domains are also connected with the others; a novel might be good, for instance, in virtue of being funny. Second, with the exception of a short section toward the end, this paper is exclusively about non-moral normative and evaluative epistemology. As I will explain later, I am not sure whether moral epistemology exhibits the peculiarity I am interested in. 26 Throughout this paper I will speak of Caitlin and Dan, and others in analogous circumstances, as “ disagreeing,” but keep in mind that I don’t mean to prejudge the question of whether one member of any such pair must be incorrect. Maybe this makes my usage of the terminology of disagreement idiosyncratic, but nothing depends on whether it is or not. (In defense of that usage though, see Khoo and Knobe (ms.) for some empirical evidence that ordinary speakers take it to be possible for two people to disagree even when their assertions can both be true. Bjornsson and Finlay [2010: 19] and Plunkett and Sundell [2013] also point out some types of disagreement which don’t consist in the overt assertion of incompatible propositions.) 24 The first of these claims is uncontroversial. This is because under most circumstances, like Alice’s, we have reason to believe that other people’s assertions reflect their beliefs, and that their beliefs are likely to be true. Given that Ben believes the object is a bear, and Alice’s background information about the quality of his vision, it is likely - setting aside the evidence of Alice’s own visual experience - that the object is a bear. This evidence may be outweighed by other evidence, but it provides some rational support for the proposition that the object is a bear. It’s a good thing that information about what other people believe generally has this significance; it justifies us in believing all kinds of things that we wouldn’t otherwise know, whether because we don’t have sufficient information, can’t figure things out on our own, or would just prefer not to do the epistemic work. And in cases of disagreement in particular, others’ conflicting opinions provide a useful check for our own fallible judgment. The ability to modify one’s credences appropriately in response to information about what other people think is very often critical to rationality. This is usually a complicated task, though. For one thing, it is often hard to tell who thinks what. Even when someone has directly told us what she believes, we may not be entirely sure that her assertion is sincere, or that our interpretation of what she says is a good representation of what she believes (maybe because what she’s said is ambiguous, or because there is some evidence that she has not articulated herself very well). And much of our information about what other people believe is even less straightforward, because they don’t directly tell us what they think; we have to infer it, with varying degrees of justification, from their behavior and their explicit assertions. The contents of others’ beliefs are not the only things we need to know in order to determine how to revise our own, either. We also need to know how confident they are in what they believe. For instance, suppose you know that I think Socrates was a great admirer of Pericles. This should have a stronger effect on your confidence in that proposition if you know that I am very certain about it than it should if I tell you I only dimly seem to recall this bit of trivia. The rational credence for you to retain in your belief is much higher if you know that I only tentatively seem to recall this fact than if I insist that I am absolutely certain. 25 Further, even if we know perfectly well what someone else believes and how confident he is, it may be difficult to assess how heavily to weigh his opinion as evidence for its content. How likely he is to be correct depends both on how well-informed he is and on his reliability with any relevant reasoning in the circumstances (considering, e.g., his susceptibility to biases on the issue), and our information about both factors is often limited or difficult to assess. Finally, even setting aside these practical problems, there is a major point of philosophical contention regarding how much one ought to conciliate in a case of disagreement. Some epistemologists hold, while others deny, that if A has antecedent reason to believe that B is an equally good judge of whether p, then in a case of disagreement A ought to think B is equally likely to be right, and so A ought to adjust A’s credence in p so that it splits the difference between A’s own antecedent credence and B’s. An example will be helpful. Simple Math: Evan and Francine are splitting their dinner bill evenly, and they have agreed to add a 20% tip. Each independently calculates what they owe, but they arrive at different answers - Evan comes up with $27.50, but Francine gets $29.50 (and each is equally confident in his or her solution). They do calculations like this together a lot, and their track records are very similar; most of the time they come up with the same answer, but in case of disagreement, Evan and Francine have historically been equally likely to be correct. Neither has any evidence that the other is impaired in any way, either (drunk, tired, sick, taking a new medication, etc.). 27 The question is: Upon becoming aware of the disagreement, but before checking their calculations, should each party’s credence in his or her answer split the difference between their antecedent credences? One camp (the “conciliationists”) contends that to do otherwise would be to privilege one’s own belief for no reason. If Evan and Francine have historically performed equally well at solving this sort of problem, and there are no special circumstances, what could justify either of them in believing that they aren’t equally likely to be right this time? But on the other hand, “steadfasters” argue that if Evan’s mathematical reasons support the conclusion that the answer is $27.50, they must also justify him in believing that it isn’t $29.50, and so that Francine’s answer is wrong - for Evan to become agnostic about which of them is right would require him to ignore his first-order epistemic reasons. 28 27 This example is a version of David Christensen’s much-discussed “ Restaurant Case” (in his 2007). 28 Prominent defenders of conciliationism include Christensen (2007, 2009, and 2011), Adam Elga (2007, 2010), and Richard Feldman (2006). 26 This is just the beginning of a deep and complicated debate. But what is important for present purposes is that whatever the right epistemological account of cases like Simple Math is, by itself it won’t explain why there is significant rational pressure for Alice to conciliate over the identity of the distant object but little or no such pressure for Caitlin to change her evaluation of Montaigne’s life. First, the debate over conciliationism is at least primarily (and maybe, for most of the philosophers involved, exclusively) about what we should do when someone disagrees with our reasoning. In Evan’s case, for example, he and Francine reason to different conclusions from the same mathematical information. But neither Alice’s case nor Caitlin’s seems to involve divergent reasoning; those disagreements arise from differences in perceptual impressions and basic evaluative standpoints, respectively. The steadfaster position is intuitively less plausible regarding Alice’s disagreement with Ben, and at least some of the main arguments for it don’t seem to apply either (e.g., reducing confidence doesn’t require Alice to ignore the results of any simple a priori reasoning she has successfully carried out), so there might not be any philosophical controversy over whether someone in Alice’s position should conciliate. More importantly, though, the topic of this debate seems to be orthogonal to ours. What we are looking for is an explanation for why evaluative disagreements generate less pressure for the parties involved to reduce confidence in their views than descriptive disagreements do, and an explanation for this phenomenon should cite some further difference between the two types of disagreement. But to choose a side in the debate between conciliationists and steadfasters is not to identify any such difference; that debate doesn’t distinguish between descriptive and evaluative epistemology. 29 And although in reality it may be difficult to know what others believe and on what basis, in the The terminology of “ conciliation” is from Elga 2010. Tom Kelly (2005 and 2010) is perhaps the most prominent steadfaster, but see also David Enoch (2010) for an interesting alternative version. 29 It could turn out that, for reason R, some steadfaster view is the right account for cases like Evan’s, but not for cases like Alice’s, and that R (something to do with a priority, maybe) also explains why cases like Alice’s are different from cases like Caitlin’s. I won’t really argue against such an explanation here, but I am very skeptical that it can be made to work. This is mainly because, in the version of Caitlin’s case that I’ll develop in the next section, she seems to face no conciliatory pressure at all, and it is hard to believe that any relevantly similar descriptive disagreement could share this feature. It may be plausible, for instance, that Evan shouldn’t quite split the difference between his antecedent credence and Francine’s, but it is much less plausible that he should be totally unmoved when she disagrees with him. (Kelly defended a view according to which such absolute steadfastness is sometimes rational in his [2005], but he avoids this radical commitment in more recent work.) 27 cases I’ll be discussing I want to abstract away from these issues as much as possible. So in Alice’s case, take it as stipulated that she knows Ben isn’t joking about thinking the object is a bear, that he’s just as confident that it’s a bear as she is that it isn’t, that his visual apparatus is generally quite reliable, and that he is not drugged, hallucinating, and so on. I take it that under these circumstances, Alice should be significantly less confident that the object is a rock than she would have been in the absence of Ben’s testimony, but nothing really turns on how much conciliation is required; the crucial point is that she should reduce her credence to at least some degree. 30 3. Should Caitlin conciliate? It seems to me that Caitlin’s epistemic situation is quite different, but that is unlikely to be clear until we have carefully described her predicament, so I’ll begin by setting out the version of Caitlin’s situation that I have in mind in more detail. In case it is still not directly intuitive that Caitlin shouldn’t conciliate with Dan, though, I will also provide a reductio argument for that conclusion - if Caitlin were rationally required to conciliate with Dan, then she would also be required to defer to others in highly counterintuitive ways. 3.1 Caitlin’s predicament as a case of pure evaluative disagreement Most normative and evaluative disagreements are entangled with disagreements over “purely descriptive” issues. For instance, you and I might disagree over whether the US should implement a certain tax policy because we disagree over whether it would increase the rate of job creation, even though we both think that if the policy would have that effect, then it should be implemented. In this case, 30 One final note on the general epistemology of disagreement: The literature tends to focus on “ peer” disagreement, in which each party’s epistemic credentials are known by both to be equally good (prior to the present disagreement). In Simple Math, for instance, I stipulated that Evan and Francine have historically been similarly reliable at calculating how much to tip on their bills, and that neither of them has any evidence that the other is impaired. It’s convenient to focus on simple cases of peer disagreement for illustrative purposes, but this can give the impression that the question of how to respond to peer disagreement is importantly different from the question of how to respond when the person one disagrees with has better or worse epistemic credentials. This impression is misleading though - it’s not plausible that it should make a huge difference to Evan whether Francine has a slightly worse track record than he does, so whatever account we settle on for the epistemology of peer disagreement had better treat the phenomenon as continuous with asymmetric disagreement. I’ll also be sticking to examples in which each party seems to deserve a similar level of epistemic respect, but there is nothing deeply special about these. 28 provided I have at least some reason to think you’re intelligent, well-informed, and sincere, I ought to respond to our disagreement by reducing credence in my original opinion of the proposed policy. But this is because the disagreement calls into question my empirical belief about the effects of the tax policy, which is a purely descriptive matter just like the one between Alice and Ben. What I mean to use Caitlin’s case to examine is how we should respond to purely evaluative disagreement, evaluative disagreement which is not derived from any descriptive disagreement. So we should imagine that she and Dan don’t disagree regarding the non-evaluative properties of Montaigne’s and Nietzsche’s lives. 31 Further, Caitlin should conciliate with Dan even if she merely has reason to believe that their disagreement is due to some purely descriptive disagreement, whether it really is or not. In virtually every normative or evaluative disagreement that actually takes place it is a live possibility that the controversy is at least partly a consequence of descriptive disagreement, because in reality we never have comprehensive knowledge of what anyone else believes, and most of the time our evaluative views seem similar. But we can nonetheless sometimes be confident that our evaluative disagreements would outlast any relevant descriptive ones, as when each party has rehearsed all the arguments he takes to favor his position, and yet the dispute persists. (For instance, disagreements over the significance of animal suffering, the relative values of freedom and security, the rational discount rate to apply to future goods, and many matters of taste seem unlikely to be resolved by the dissemination of any purely descriptive information.) It may be a little hard to imagine that Caitlin could really be sure that her dispute with Dan falls into this intractable category, but it will do for our purposes if we just suppose that she has very good evidence that it does, and is correspondingly confident. And finally, we are generally inclined to continue seeking common ground in an evaluative disagreement unless we have a plausible explanation for why those who disagree with us might just be 31 Some metaethical views, according to which normative and evaluative concepts are reducible to purely descriptive ones, obviously imply that there is no such thing as evaluative disagreement in the absence of descriptive disagreement - at least if disagreeing entails having contradictory beliefs (see notes 24 and 29 for some reasons to doubt that it does). This doesn’t matter for our purposes though. What I am interested in are linguistic exchanges of the form A: “ X is good.” B: “ No it isn’t.” in which neither party seems to generate any epistemic reason for the other to reduce confidence in the belief she has expressed. Analytic naturalism is compatible with the existence of such exchanges. 29 irreconcilably different in some relevant way. So in this example, Caitlin and Dan can each see why - in an etiological sense - things would seem different from the other’s perspective. Pure Evaluative Disagreement: Caitlin and Dan are both historians of philosophy who are particularly interested in Montaigne and Nietzsche, and they’ve studied the biographies of the two philosophers extensively. They’ve also spent a lot of time discussing the merits of the two, and they even agree on what those merits are: Nietzsche was a lonely, unhappy man who didn’t have much success at anything other than philosophy, but a lot of his work was original and profoundly important. Montaigne was not much of an original philosopher (though he was a gifted and hugely erudite writer), but he appears to have been a much happier and more likable person, with a thriving social life and an interesting political career. The well-roundedness and pleasantness of Montaigne’s life just weigh more heavily with Caitlin, whereas Nietzsche’s philosophical achievements seem more important to Dan. Further, Caitlin and Dan can both see what likely explains this basic difference in their sensibilities; Dan comes from a family of talented but neurotic and single-minded artists, whereas Caitlin grew up in a more social, easygoing environment, and it seems likely that this led them to settle on correspondingly different images of what an ideal life is like. Nietzsche’s life matches the image that Dan grew up with pretty well, but Montaigne’s life matches Caitlin’s ideal better. 32 In this case, I think Caitlin’s confidence in her judgment of Montaigne’s life should not be shaken by the discovery that Dan disagrees with her. His testimony gives her no reason, for instance, to reduce her social activities in order to spend more time on her work. 3.2 Unacceptable consequences of maintaining that Caitlin should conciliate I hope this verdict carries some intuitive appeal, now that the case has been properly described and distinguished from relatives in which Caitlin doesn’t have such strong evidence that she is engaged in a purely evaluative disagreement. But we can produce further support for it by drawing out some implications of the alternative view that Caitlin is rationally required to conciliate with Dan. If Dan’s dissenting opinion ought to reduce Caitlin’s confidence in her evaluative judgment, then presumably adding more judges who have independently arrived at the same view that Dan has ought to reduce it further. Opposing Evaluative Consensus: Caitlin and Dan consult the rest of their neighborhood, one by one, on the question of whether Nietzsche or Montaigne had the better life. After extensive discussion and reflection, and even though there is total consensus on the relevant purely descriptive facts about each life, everyone else agrees with Dan; Caitlin is outnumbered 25 to 1. 32 This is the sort of story that Mackie presumably had in mind when he suggested that values “ reflect people’s adherence to and participation in different ways of life.” (1977) 30 If Caitlin’s predicament were like Alice’s, then the rational way of responding to her information about what her neighbors think would be to drastically reduce her confidence that Montaigne had the better life - in the aggregate, she ought to trust their judgment over her own, just as Alice ought to accept the majority opinion if 24 more independent observers agree with Ben that the distant object is a bear. 33 But intuitively it could never be rational for Caitlin to give her personal evaluative judgment similarly little weight. Another way of making the same point is to consider the epistemic value of information about what other people think on an issue about which one has yet to arrive at any opinion of one’s own. I haven’t personally weighed the evidence for and against the proposition that humanity is contributing significantly to climate change, but I am perfectly justified in relying on expert opinion on the subject. Even if I did research the question extensively for myself, the epistemic significance of my own opinion would be small given the degree of attention the issue has received in the scientific community; the knowledge of what so many other intelligent, well-informed people think would swamp the epistemic weight of my own judgment anyway. So I almost might as well skip my own investigation, when I have access to the opinions of a large community of knowledgeable and intelligent people. But consider Google Value: In the year 2040, Google develops a service for crowdsourcing the answers to evaluative questions. Users upload whatever descriptive information they take to be relevant and then elicit one another’s opinions about what’s good and bad, or better and worse. The service is extremely popular; the average query quickly receives thousands of responses, which Google aggregates into percentages of negative and affirmative responses to yes or no questions. They also diligently weed out respondents who are insincere or don’t take others’ questions seriously, and users can even curate their panels themselves - adding and subtracting people to form groups of advisers they have personally vetted. 34 Suppose Caitlin has access to Google Value, and that she has a choice between assessing the merits of Nietzsche’s and Montaigne’s lives on her own or consulting the service. Which would be the better method for Caitlin to use? 33 It’s important that their neighbors all reach this judgment on their own, independently of Dan and of one another. If 24 more people tell Alice that she’s wrong about whether the object is a rock or a bear, but only because they’re taking Ben’s word for it, this would seem to generate little or no additional pressure on Alice to conciliate. 34 This service is a relative of Robert Howell’s (2012) “ Google Morals” which he uses as a device for discussing moral deference. 31 If she were interested in the answer to a purely descriptive question, it would clearly be epistemically rational to defer to the analogous panel. But in the evaluative case it would be bizarre to treat even a large number of carefully selected and well-informed people as a collective guru; there is no very good substitute for one’s own evaluative judgment. 35 Someone who thinks that the evaluative domain is epistemically autonomous - in the sense that we can never draw evaluative conclusions from purely naturalistic premises, so that a separate belief- forming method must be involved - might find it unremarkable that information about what these other people think is of little epistemic significance for Caitlin, for the following reason: Discovering what someone else thinks makes it rational to adjust our credences only to the extent that we have reason to respect his judgment - for instance, it’s only rational for Alice to reduce her confidence that the distant object is a rock to the extent that she has reason to believe that Ben has good vision, isn’t hallucinating, and so on. But what reason does Caitlin have to respect Dan’s evaluative judgment, or the views of people she’s never met in Opposing Evaluative Consensus and Google Value? If she has no such reason, then that explains why she need not reduce confidence in any of these scenarios. Suppose that this autonomist view is correct. It seems to me that we still don’t have an explanation for the major difference between Caitlin’s predicament and Alice’s. Suppose Caitlin knows that, in the past, Dan’s evaluative views have strongly tended to agree with her own (and similarly, in the other cases, that the people whose opinions she has access to have long track records of agreeing with her). Even in these versions of the cases, as long as she is sure their disagreement is purely evaluative, it seems to me that Caitlin ought not to conciliate over who had the better life; the history of agreement makes no epistemic difference. 36 This is, again, strikingly dissimilar to the case of Alice and Ben - the rational pressure for Alice to conciliate varies with her evidence that Ben’s vision is reliable, and a track record of agreement is one source of such evidence. To recap: If we are careful not to confuse cases of purely normative or evaluative disagreement 35 My view is certainly not that information about what other people think is totally useless in evaluative epistemology; examples like the disagreement over whether a certain tax policy is good are sufficient to demonstrate that it isn’t. But the epistemic relevance of evaluative testimony is conditional in a distinctive way, so that our evaluative beliefs should often be less sensitive to what other people think than our straightforwardly descriptive beliefs should be. 36 Again, in remotely realistic versions of such a case, I think she ought to conciliate at least a little bit. But as I will explain later on, this is not for the same reason that Alice ought to conciliate with Ben. 32 with disagreements that involve descriptive conflicts, it seems intuitive that they do not rationally require the participants to conciliate. And this view of the epistemic import of information about what other people think gains further support from the fact that rejecting it entails, implausibly, that others’ evaluative judgments are sometimes vastly more epistemically significant for us than the judgments we make on our own. 4. Should we privilege our own evaluative judgment over others’? From now on I will take it for granted that Dan’s disagreement shouldn’t make Caitlin significantly less confident in her belief that Montaigne’s life was better than Nietzsche’s. What might explain this epistemological datum? In most cases the absence of pressure to conciliate is due to the subject’s having very strong reason to believe the other party is much less reliable. For instance, Alice shouldn’t take Ben’s opinion very seriously if she knows he’s on a hallucinogenic drug, or that he has extremely poor vision. Might Caitlin have some analogous reason to think that, under the circumstances, she is overwhelmingly more likely to be right than Dan is? By stipulation, Caitlin knows that their disagreement is purely evaluative; they don’t disagree about any of the biographical facts of Nietzsche’s and Montaigne’s lives. So if she is justified in disregarding Dan’s opinion on the basis of strong evidence that he is unreliable, she must be justified in believing that the defect resides in his evaluative sensibility. And she must be quite highly justified in believing that Dan is very unreliable, if she is licensed to completely disregard his opinion. We can imagine the case so that Caitlin has no special, particular reason to think Dan is unreliable - in fact, as we just saw, we can assume that in the past Caitlin and Dan have largely or even completely agreed on evaluative matters. This seems not to have any bearing on whether Caitlin should conciliate. So if Caitlin is entitled to believe that, conditional on disagreeing with her, Dan is very unlikely to be correct, this must be for some general reason having nothing especially to do with Dan - she must be entitled to believe that, other things being equal, she is far more likely to be correct than anyone who might disagree 33 with her on evaluative matters! It is hard to imagine what strong, general reason Caitlin could have for believing that she is more likely to be right in a case of evaluative disagreement than anyone who might disagree with her - even those who have always agreed with her in the past. For one thing, what could explain Caitlin’s possessing such a special gift for evaluative judgment? And if it’s rational for Caitlin to have so much faith in her own reliability, then it must be equally rational for Dan to have a similarly extreme level of trust in his - but surely it doesn’t make sense to think one’s own judgment is better than everyone else’s while acknowledging that everyone else has just as much reason to think that his judgment is the best. 37 Some philosophers writing about what’s odd, or problematic, or puzzling about moral deference have suggested that the distinctive importance of one’s own judgment in that sphere is explained by the greater moral worth of actions that are based on deep moral understanding or performed for the right reasons, which requires reliance on one’s personal sense of right and wrong. 38 If some such account explains what’s peculiar about moral deference, it might seem that there should be a similar explanation for the relative insignificance of information about what others think in normative and evaluative epistemology. Perhaps there is some extra value in understanding what makes it the case that one ought to do something, or what makes something good, which we can only capture if we rely exclusively on our own judgment. But it seems to me that this is not even a good explanation of whatever is problematic about moral deference, let alone that it can provide a satisfying account of the phenomenon we’ve been 37 It might be thought that something like Ralph Wedgwood’s view of moral epistemology (in his 2010) suggests a solution to our puzzle. On the evaluative analogue of that view, there is a crucial difference between Caitlin’s first-order judgment that Montaigne’s life was better and her knowledge that Dan thinks Nietzsche’s was better: her epistemic reason to believe that Montaigne’s life was better is her direct intuition that Montaigne’s life was better, whereas her reason to doubt it is that Dan has the intuition that Nietzsche’s life was better. These are different kinds of reasons. The former is capable of directly justifying Caitlin’s belief, but the latter has no such immediate relevance, since it depends on Caitlin’s also having reason to believe Dan is likely to be right; and Wedgwood contends that it is rational to treat the two reasons in different ways because it is rational to have a special sort of trust in one’s sources of immediate, basic justification. But to the extent that Caitlin is justified in maintaining her original credence in the face of her disagreement with Dan, she must be justified in believing that Dan is more likely to be wrong - and I don’t see how the difference Wedgwood observes can justify Caitlin in believing that. Further, Caitlin would need to be justified in believing the disparity in their evaluative reliability was very large in order to be justified in maintaining full confidence in her own judgment, and it is not plausible that the mere fact that Dan’s judgment can’t directly justify Caitlin’s belief could account for such an extreme difference. 38 See Alison Hills (2009, 2013). Sarah McGrath (2011) also suggests this is the most promising way of accounting for what’s odd about moral deference. 34 examining. 39 The enhanced moral worth of an action that is performed for the very reasons that make it morally good may well provide some reason to privilege one’s own moral judgment - but not an epistemic reason. For instance, if Gwen thinks it’s wrong to wear clothes made from animal fur and Harry disagrees, it isn’t any more subjectively likely for Gwen that Harry is wrong just because her prospects for accruing moral worth are better if she ignores his dissent. To put things a bit differently, if the puzzle of moral deference is supposed to be a genuinely epistemological issue, a solution to it must explain why, from each person’s perspective, and other things being equal, he or she is much more likely to be right about what’s right and wrong than anyone else. Appealing to the distinctive moral worth of actions done for the right reasons doesn’t accomplish this. 40 In summary, the prospects for an explanation according to which each of us should privilege his or her evaluative judgment over others’ seem very dim, because we would each have to be justified in thinking that, in a case of disagreement (and other things being equal) we were more likely to be right than anyone else - even those who have, by our own lights, been just as reliable as we are in the past. In order to be so justified we would seem to need some compelling explanation for our superior evaluative reliability, and it is hard to imagine what this could be - especially because we would have to be justified in believing it despite knowing that everyone else has equally good reason to think he or she is more reliable than we are, and that at most one of us can be right about this. Furthermore, each of us would have to be justified in believing that his or her own judgment was far better, in order to explain why it’s rational for Caitlin to conciliate so little with Dan despite having no special evidence that he is unreliable. 41 5. Can both parties in an evaluative disagreement be right? In the previous section we rejected some possible explanations for the fact that Caitlin shouldn’t 39 Assuming there is something similarly problematic about moral deference. 40 Maybe the puzzle these authors are interested in is not really an epistemological one. In that case, since the topic of this paper is an issue in epistemology, my point is just that nothing like their proposal will work in the present context. 41 The arguments in this section are relatives of Crispin Wright’s “ Simple Argument”; see his (2006), esp. pages 40-41. 35 conciliate which assumed that the truth of her belief must be incompatible with the truth of Dan’s. My suggestion is that we can do better if we drop this assumption. There are at least two ways we might try explaining how Caitlin and Dan could both be correct: Their beliefs could be about distinct, compatible propositions, or they could both be about a single proposition which is capable of simultaneously being true for Caitlin and false for Dan. 5.1 Relativism The latter view, sometimes called “genuine relativism,” 42 is usually motivated in part by the intuition that cases like Caitlin’s exemplify “faultless disagreement” - proponents find it intuitive 1) that the parties to such a dispute attribute opposite truth values to the same proposition and yet 2) that neither party need be mistaken. This may sound paradoxical in the abstract, but it is an attractive way of describing what’s going on in at least some domains of discourse, notably discussions involving epistemic modals. For instance: Might: Ingrid and Jake are looking for a nice place to play backgammon. Ingrid would prefer to go to a certain coffee shop, but she isn’t sure it’s open - she knows they aren’t open every day, but she doesn’t know the schedule. So Ingrid says, “The coffee shop would be nice, but they might be closed today.” Jake is better informed, however. “No, they’re only closed on Mondays,” he explains. Relativism provides an appealing account of this sort of case. It looks as though Jake and Ingrid are assessing the truth value of the proposition that the coffee shop might be closed; this is the proposition Ingrid asserts, and when Jake says “No” he is expressing his belief that that proposition is not true. But it doesn’t seem as though what Ingrid said was simply false (she certainly would’ve been doing something inappropriate if she had, given the same information, asserted that the shop was open). So Might appears to be a case in which two parties ascribe different truth values to the same proposition, and yet neither is incorrect (although of course the truth value of that proposition for Ingrid flips in the wake of the new information she gets from Jake). 42 This term is from Kolbel (2004). See also Egan (2004) and MacFarlane (2007) for relativist views of predicates of taste and epistemic modals, and especially Egan (2012) for his relativist dispositional theory of value. 36 Similarly, maybe the proposition that Nietzsche’s life was better than Montaigne’s is true for Dan even though it’s false for Caitlin. This would explain why Caitlin can maintain full confidence in her view without any implausibly chauvinistic commitment to the superiority of her evaluative judgment. I won’t argue against relativism here; our topic is the peculiar epistemic significance of evaluative testimony, and if relativism makes sense then it does explain that phenomenon. This account is based on the primitive notion of relative truth though, so it is only illuminating to the extent that we have an independent understanding of that notion - and it is difficult at best to see how a single object could simultaneously have a certain property and lack it, depending on which perspective is operative. An explanation in more familiar terms would be preferable, so for the remainder of the paper I will set relativism aside. 43 5.2 Contextualism A sentence however, can obviously be true in some contexts and false in others, because the proposition it expresses may vary. Contextualism is the view that the contents of evaluative sentences depend on the circumstances of use in this way, and it promises to explain the peculiar (in)significance of information about others’ evaluative beliefs simply by invoking this everyday sort of context-dependence. When Dan says “Nietzsche’s life was worse than Montaigne’s,” Caitlin has strong reason to believe that the relation he is using “worse” to attribute is compatible with the relation she attributes by saying “Montaigne’s life was better than Nietzsche’s,” and since both statements can thus be true, Dan’s testimony doesn’t give her any reason to question the belief she has expressed. 44 45 This may sound very odd; are we supposed to believe that whenever there is an evaluative dispute 43 Hawthorne and Cappelen (2010) make a more sophisticated case against relativism. They argue that relativism is motivated by misleading tests for content identity and an undue pessimism about the resources available to contextualists, and that in any case it is vulnerable to counterexamples. For instance, if evaluative propositions have relative truth-values, then evaluations made by people with conflicting interests apparently ought to contradict each other. But when one backgammon player says “ What a great roll!” and his opponent says, “ Oh no, that’s terrible,” there is no sense of contradiction, and neither is likely to think the other has expressed a false belief. (See chapter 4, esp. 122-26.) 44 Someone in a highly idealized position like Caitlin’s can be unusually confident that this is what’s going on, but even in more familiar circumstances it is frequently a non-negligible possibility. 45 James Dreier (1990) sketched a similar view of moral language, and more recently, as I’ll be discussing in a moment, Stephen Finlay (2014) has worked extensively on this sort of contextualist view of normative and evaluative terms. 37 each party is talking past the other? 46 Most of the time, I think, this isn’t what’s going on - in general, we coordinate our usage of evaluative terminology so that everyone involved in a single conversation uses “good” to pick out the same property, so in the normal case if I assert “X is good” and you say “No it isn’t,” we can’t both be asserting true propositions. But pairs of sentences that might superficially appear to express incompatible propositions are not always really contradictory; two people may unwittingly use the same indexical term to pick out different objects or properties, or not realize that multiple objects satisfy a certain description, or fail to converge on some relevant standard (what height counts as “tall,” whose orientation determines what is “on the left”), and in these cases there need not be any real tension between their assertions. For most context-sensitive expressions, it’s clear what the source of the extensional variation is. For instance, “here” refers to the speaker’s location at the time of utterance (with the boundaries of that location drawn by other features of the context), “now” refers to the time of utterance, and “that dog” refers to whichever dog is most salient in the context. What’s context-dependent about “better” and its cognates? Specifying the parameter is a complicated job, and too big for this paper, but I’ll briefly mention a couple of possibilities. One natural “subjectivist” thought is that “X is good” expresses a proposition about the relationship between X and the speaker’s perspective, where her perspective might consist of a set of some or all of the following: attitudes, desires, emotional dispositions, interests, and preferences. So “X is good” might mean something like “I prefer X [to some salient alternative].” This view clearly explains why Caitlin shouldn’t reduce confidence in her evaluative belief when Dan (superficially appears to) reject it; she’s talking about her preferences, while he’s talking about his preferences, and they have strong evidence that those preferences diverge in this case. But subjectivism is rightly unpopular, for reasons that might appear to be endemic to any contextualist account of evaluative terms: First, subjectivism seems to entail that too much of evaluative discourse involves parties talking 46 G.E. Moore lodged this objection to subjectivism in his (1922: 333-34), but the same basic idea - that the theory can’t account for disagreement - remains a popular criticism of more sophisticated forms of contextualism. 38 past one another; and second, if subjectivism were true, we would expect Caitlin and Dan to cite their own attitudes as evidence that their evaluative beliefs are true - but actually this would be bizarre. The contextualist can do better, though. For instance, according to Stephen Finlay’s “end- relational” semantics for evaluative terms, “X is good” means (ignoring some subtleties) “X makes [contextually salient end, or goal] more probable than [contextually relevant alternative] relative to [contextually determined set of background information].” (2014: chapter 1) This explains the prevalence of contradiction, provided that the parties to such disputes usually have the same ends in mind; they disagree over what makes those ends more probable. 47 And since the achievement of some preferred mental state is not generally our most salient end - for the most part our goals are non-psychological states of affairs - information about our own attitudes is not usually relevant to the evaluative propositions we are interested in. I don’t mean to defend Finlay’s particular view here; that would also be a very large project. But contextualism does offer a very attractive account of the difference between Caitlin’s disagreement with Dan and Alice’s with Ben, and Finlay’s version demonstrates that such a theory need not take on subjectivism’s obvious defects. 48 5.3 Expressivism It might be thought that expressivist theories offer a third way of explaining why evaluative testimony often carries less epistemic weight than purely descriptive testimony. But expressivists are 47 As I alluded to earlier, in footnote 3, the contextualist can also explain that there is a sense in which even parties who are talking past one another may disagree. This is the sense in which disagreeing is an activity, rather than a state of mind. As Hawthorne and Cappelen observe in introducing this distinction, in this sense one can agree to something without realizing what one is agreeing to. Suppose you sign a document, without reading it through. With ‘I agree. NN,’ you have in an important (and often legally binding) way agreed with its content. The case also illustrates that, just as a very standard way of promising is actually to use the verb ‘promise,’ so a very standard way of agreeing is actually to use the verb ‘agree.’ (2010: 60) 48 Contextualists do have other, more serious problems. A sophisticated contextualism can avoid many of the objections that simpler versions like our toy subjectivist theory are vulnerable to, but there may not be any getting around some counterintuitive implications regarding our assessments of the truth-values of one another’s utterances. For instance, if some contextualist theory of evaluative terms is correct, then why is it that when Dan says “ Nietzsche’s life was better” Caitlin will say “ No, it wasn’t,” even when she understands Dan’s usage completely? In that case, Caitlin supposedly knows that Dan is saying something true, so why does she appear to tell Dan he’s wrong? There may be a good contextualist explanation for this linguistic pattern, perhaps involving an extensive pragmatic story, but it is a major difficulty (and perhaps a reason to prefer relativism). Again, the main purpose of this paper is to pose a strong challenge for absolutist theories of evaluative language, not to argue for contextualism in particular; and I certainly don’t claim that the considerations I have produced are decisive on either front. 39 caught in a dilemma here - straightforward non-cognitivist views can offer a distinctive explanation, but are widely regarded as implausible for other reasons, and the quasi-realist theories that have been constructed to solve problems have given up their distinctive explanatory resources in the process. An old-fashioned non-cognitivist like A.J. Ayer could offer a very simple explanation. On his view, sentences like “Montaigne’s life was better than Nietzsche’s” don’t have truth values or express beliefs at all; rather, they express emotions, and so it doesn’t even make sense to ask epistemological questions about them. Alice is required to conciliate with Ben because their disagreement involves incompatible beliefs; Caitlin isn’t required to conciliate with Dan because their disagreement just consists in the expression of different emotions or attitudes. But this sort of expressivism fell out of favor decades ago, for a cluster of reasons. Not only do we call our evaluative attitudes “beliefs” and speak of them as being “true” and “false,” which would seem quite surprising if non-cognitivism were true, but we also use evaluative sentences in deductive arguments and embed them in ways that only seem to make sense if they have truth values. 49 Modern forms of expressivism, like Allan Gibbard’s and Simon Blackburn’s, 50 seek to accommodate these realist-seeming aspects of evaluative discourse. Their “quasi-realist” views accept that our evaluative attitudes are beliefs, that their contents are capable of being true or false, and that they are subject to epistemic norms, which allows them to avoid at least some of the straightforward non- cognitivist’s problems. But as a consequence, the non-cognitivist’s explanation for the special significance of evaluative testimony is no longer available to a quasi-realist. Maybe quasi-realism could be combined with relativism or contextualism to provide a solution, but I don’t see what distinctively expressivist explanation could be given without reverting to non-cognitivism. 51 49 John Searle (1962) and Peter Geach (1965) were the first to raise concerns like these, now collectively known as “ the Frege-Geach problem”; see Schroeder (2008) for a review. 50 See Gibbard (1990, 2003) and Blackburn (1984, 1993, 1998). 51 There also might not be any point to be being a quasi-realist if one has already rejected absolutism about evaluative terms. The primary motivations for quasi-realism - naturalism, internalism, the apparent unanalyzability of evaluative terms, avoidance of dicey epistemological commitments - do not obviously provide any basis for favoring it over a non-absolutist theory of evaluative language according to which it expresses the same sort of mental state (ordinary belief) that purely descriptive language does. 40 6. What about morality? So far I’ve restricted my discussion to the epistemic significance of information about other people’s non-moral normative and evaluative beliefs. This may seem strange, both because philosophers have mainly tended to adopt the converse focus - on morality at the expense of the rest of the normative domain - and because moral epistemology and non-moral evaluative epistemology might seem likely to be similar in every philosophically interesting way. It might even seem that if my suggestions about non- moral evaluative epistemology cannot be extended to explain the rational way of incorporating information about others’ moral beliefs, this is strong evidence that I’ve gone wrong somewhere. I am actually not sure whether moral epistemology is different, in this respect, from evaluative epistemology in general, but it does seem more intuitive that if Caitlin and Dan were having a moral dispute, one of them would have to be saying something false. 52 Fortunately, there seems not be a problem for the sort of contextualist account I’ve gestured at either way: if there is no relevant difference between moral value and non-moral value, then we can just apply the contextualist story more generally; but if there is such a difference, then I think contextualism has the resources to explain it. For illustration, I’ll use a new example: Pure Moral Disagreement: Kate and Larry are having a dispute over whether it is morally permissible to boil a lobster alive in the course of making dinner. They agree that this causes the lobster intense pain, that this method has no significant nutritional benefits, that the lobster will taste much better if they boil it alive, and so on - they are justifiably convinced that their moral disagreement is not derived from any such difference in their descriptive views. It’s just that causing intense pain to an unintelligent animal seems wrong to Kate, but not to Larry. I don’t have any strong intuitions regarding the rationality of conciliation in this case. For the sake of argument, though, suppose that Kate should conciliate even though someone in Caitlin’s circumstances shouldn’t. If the sort of contextualist view I’ve recommended for evaluative terms is correct, then what explains this difference? 52 Although, on the other hand, Khoo and Knobe (ms) present empirical evidence from several studies that suggest ordinary speakers tend to treat it as possible for parties in a conversation who are from very different cultures to have exchanges of the form “ X is wrong”/“ No it isn’t” in which neither party is saying something false. (Incidentally, the authors discuss this pattern of results in the service of their argument that disagreement doesn’t require the assertion of incompatible propositions - subjects also tended to describe the parties in such exchanges as disagreeing - which is some evidence that my usage of that terminology in this paper has not been idiosyncratic.) 41 My suggestion on behalf of the contextualist is that morality may have its own ends, so that “morally good” and “morally ought” (and uses of normative and evaluative language which are implicitly about morality) don’t have the same flexibility that “good” and “ought” generally exhibit. Consequently, saying that boiling the lobster is morally bad may not be like saying that snakes make bad pets; it’s like saying that snakes make bad pets for people who want to have affectionate relationships with their animals. If that’s right, then there isn’t the same room for people to talk past each other in moral discourse because their claims must be relativized to the same ends (that is, morality’s ends - whatever those are). 53 7. Conclusion In the first half of this paper, I argued that information about others’ purely evaluative beliefs has much less significance for what we ourselves should believe than information about their purely descriptive beliefs does. In cases like Pure Evaluative Disagreement, we can see that when someone’s evaluative assertion conveys no new descriptive information to us, her testimony provides us with no reason to modify our evaluative beliefs. This view gains further support from the implausible consequences of denying it; if information about others’ purely evaluative opinions did have the same sort of significance as information about their descriptive beliefs, then there would be circumstances (like Google Value and Opposing Evaluative Consensus) in which it would be rational to defer completely to their evaluative beliefs. But in fact it wouldn’t be rational to rely entirely on the judgments of others in these cases, so this line of reasoning also supports the view that we should retain full confidence in the face of purely evaluative disagreement. This striking difference between descriptive and evaluative epistemology needs an explanation. I argued that we can’t provide an adequate one unless we accept that it’s possible for both parties in a 53 It could be that some cultures have distinct moralities, each with its own system of ends; this would fit the data from Khoo and Knobe (ms). In that case, there would be some room for members of different societies to talk past one another, although it would be much more unusual than the sort of individual divergence that Caitlin and Dan exhibit. (And plausibly there are some things that, as a conceptual matter, any two moral systems must agree on - e.g., that maximizing hand-clasping is not the highest moral priority. See Philippa Foot [1958].) 42 dispute like Caitlin’s and Dan’s to be correct. If Caitlin takes herself to be saying something incompatible with the view Dan has expressed, then maintaining full confidence in her belief would commit her to the view that she must be gifted with an evaluative sensibility far superior to anyone else’s - even those who had always agreed with her in the past - despite her knowledge that everyone else was equally justified in believing that he was the lucky one. Finally, I examined a few different ways we might try to explain how each party in a pure evaluative disagreement might be correct. The genuine relativist’s view explains how this might be, but it does so by relying on a primitive concept of relative truth, and it seems to me that we should avoid employing this mysterious notion if we can make do with more familiar ones. A subjectivist contextualism also explains the absence of conciliatory pressure in situations like Caitlin’s, but this advantage is more than outweighed by other problems, even if we focus only on epistemology - in general, it makes justification for evaluative beliefs too easy to come by and obtainable in intuitively unacceptable ways. But some sophisticated form of contextualism appears particularly well-suited to explaining why information about others’ evaluative beliefs is relevant to our own in just the circumstances that it intuitively is. The possibility of talking past one another explains why others’ opinions can be irrelevant to our evaluative questions, since to the extent that I am justified in believing your evaluative claim concerns a different relation than the one I am interested in, it may give me no reason to doubt my judgment. Usually we are on the same evaluative page, but sometimes there is just a basic divergence in what we classify as valuable, and contextualism can account for the way it is intuitively rational to treat this sort of circumstance. And if that’s right, then it turns out that the feature that has often been seen as contextualism’s fatal flaw is actually an important strength. There is no denying that the contextualist has his work cut out for him if he is to explain why, in other respects, we appear to treat disputes like Caitlin’s with Dan as involving the assertion of incompatible propositions. But as I argued earlier, it looks as though rejecting the appearance of contradiction is the only satisfactory way to explain why Caitlin is not rationally 43 required to lose confidence in her belief - and more generally, why it makes sense for us to rely as heavily on our personal evaluative judgment as we do. 44 References Ayer, A.J. (1952) Language, Truth, and Logic. New York: Dover Publications. Blackburn, S. (1984) Spreading the Word: Groundings in the Philosophy of Language. Oxford: Oxford University Press. _____. (1993) Essays in Quasi-Realism. Oxford: Oxford University Press. _____. (1998) Ruling Passions: A Theory of Practical Reasoning. Oxford University Press. Bjornsson, G. and Finlay, S. (2010) Metaethical Contextualism Defended. Ethics 121(1), 7-36. Cappelen, H., and Hawthorne, J. (2010) Relativism and Monadic Truth. Oxford: Oxford University Press. Christensen, D. (2007) Epistemology of Disagreement: The Good News. Philosophical Review 116, 187- 217. _____. (2009) Disagreement as Evidence: The Epistemology of Controversy. Philosophy Compass 4, 756-67. _____. (2011) Disagreement, Question-Begging and Epistemic Self-Criticism. Philosophers’ Imprint 11(6), 1-22. Dreier, J. (1990) Internalism and Speaker Relativism. Ethics 101(1), 6-26. Egan, A. (2012) Relativist Dispositional Theories of Value. Southern Journal of Philosophy 50(4), 557- 82. _____. (2006) Secondary Qualities and Self-Location. Philosophy and Phenomenological Research 72(1), 97-119. Elga, A. (2007) Reflection and Disagreement. Noûs 41(3), 478–502. _____. (2010) How to Disagree about How to Disagree. In Feldman, R. and Warfield, T. (eds.), Disagreement (pp. 175-86). Oxford: Oxford University Press. Enoch, D. (2010) Not Just a Truthometer: Taking Oneself Seriously (but not Too Seriously) in Cases of Peer Disagreement. Mind 119, 953-97. Feldman, R. (2006) Epistemological Puzzles about Disagreement. In Hetherington, S. (ed.), Epistemology 45 Futures (pp. 216-36). Oxford: Oxford University Press. Finlay, S. (2014) Confusion of Tongues: A Theory of Normative Language. Oxford: Oxford University Press. Foot, P. (1958) Moral Beliefs. Proceedings of the Aristotelian Society 59. 83-104. Geach, P. (1965) Assertion. Philosophical Review 74, 449-65. Gibbard, A. (1990) Wise Choices, Apt Feelings. Cambridge, MA: Harvard University Press. _____. (2003) Thinking How to Live. Cambridge, MA: Harvard University Press. Hills, A. (2009) Moral Testimony and Moral Epistemology. Ethics CXX(1), 94-127. _____. (2013) Faultless Moral Disagreement. Ratio XXVI(4), 410-427. Howell, Rt. (2012) Google Morals, Virtue, and the Asymmetry of Deference. Nous 48(3), 389-415. Kelly, T. (2005) The Epistemic Significance of Disagreement. Oxford Studies in Epistemology I, 167–96. _____. (2010) Peer Disagreement and Higher-Order Evidence. In Feldman, R. and Warfield. T. (eds.) Disagreement (pp. 111-174). Oxford: Oxford University Press. Kolbel, M. (2004) Indexical Relativism versus Genuine Relativism. International Journal of Philosophical Studies 12, 297-313. MacFarlane, J. (2007) Relativism and Disagreement. Philosophical Studies 132, 17-31. Mackie, J.L. (1977) Ethics: Inventing Right and Wrong. Harmondsworth: Penguin. McGrath, S. (2011) Skepticism About Moral Expertise as a Puzzle for Moral Realism. The Journal of Philosophy CVIII(3), 111-37. Moore, G.E. (1922) Philosophical Studies. London: Routledge & K. Paul. Plunkett, D. and Sundell, T. (2013) Disagreement and the Semantics of Normative and Evaluative Terms. Philosopher’s Imprint 13, 1-37. Schroeder, M. (2008) What is the Frege-Geach Problem? Philosophy Compass 3(4), 703–720. Searle, J. (1962) Meaning and Speech Acts. Philosophical Review 71, 423-32. Wedgwood, R. (2010) The Moral Evil Demons. In Feldman, R. and Warfield, T. (eds.) Disagreement (pp. 216-46). Oxford: Oxford University Press. 46 Wright, C. (2006) Intuitionism, Realism, Relativism and Rhubarb. In Greenough, P. and Lynch, M. (eds.) Truth and Realism (pp. 38-60). Oxford: Clarendon Press. 47 CHAPTER 3 CONTEXTUALISM ABOUT VALUE 1. Introduction Contextualism about evaluative terms is the view, or family of views, according to which an evaluative term can be used to attribute a variety of different properties depending on the context in which it occurs, so that a single sentence like “Abraham Lincoln was the best president of the United States” can be used to assert a wide range of distinct propositions. 54 This sort of view faces a prominent objection, and judging by the scarcity of contextualists about value 55 it is widely considered to be a strong one: if there were really so much variation in what we mean by our evaluative terms, then there would be far less disagreement between us than there actually is. And doesn’t anyone who says “Lincoln was our best president” disagree with anyone who says “FDR was our best president”? There appears to be an abundance of such counterexamples. In this paper, I will begin by presenting the objection from “lost disagreement” in more depth, specifically as it applies to the version of contextualism that I favor. My contention will be that while there are difficult puzzles involving evaluative disagreement, far from providing decisive grounds for rejecting contextualism they actually suggest that some member of that family of theories must be right. Contextualism does imply that some exchanges bearing the superficial markings of contradiction in fact involve compatible assertions, but there is strong independent support for that diagnosis. And when there is independent reason to believe that two people really are contradicting one another, contextualists have 54 Until recently it would have been perspicuous to call this a relativist theory of value, but now “relativism” often refers to a different sort of view (also called “New Age relativism” or “genuine relativism”). On that sort of view, there is just one property that a term like “good” attributes, and so each evaluative sentence expresses a single proposition (assuming it contains no other context-sensitive expressions); what varies between contexts is not the content of an evaluative sentence, just its truth-value. So unlike the contextualist, the genuine relativist holds that evaluative propositions are potentially, in the final analysis, simply true for you but false for me. See, for example, Kolbel (2004) and MacFarlane (2007). 55 Stephen Finlay (2014) is the only clear example, although David Copp (1995), James Dreier (1990, 1992), Gil Harman (1975, 1978, 1996), David Velleman (2013), and David Wong (1984, 2006) have views about morality that are more or less closely related. Richard Kraut (2011) argues that there is no such thing as “absolute” goodness – as opposed to, e.g., goodness for X – but his view is a sort of skepticism according to which absolute goodness is, like phlogiston, simply something we have no reason to believe exists. 48 underappreciated resources for explaining why that is. This advantage is just one among many, however, and in the latter half of the paper I explain how a contextualist theory like the one discussed in this paper can offer comparatively simple, plausible explanations of many of the phenomena that mainstream metaethical theories have struggled to account for, including internalism, our knowledge of what is good, and the relationship between value and the natural universe. 2. The problem of lost disagreement A contextualist version of subjectivism provides a good illustration of the problem of lost disagreement. Subjectivism: S’s assertion “X is good” expresses the proposition that S’s value system approves of X. There is more than one problem with Subjectivism, but the most important one for our purposes can be seen in the following example: Taxes: Suppose Aaron and Beth are discussing a bill that would raise taxes on the top earners in order to reduce those on the lowest brackets. Aaron: (1) “That’s a pretty good bill.” Beth: (2) “No, it’s a terrible bill.” Here’s the problem: Intuitively, Aaron and Beth are disagreeing - they are asserting propositions that can’t both be true. But if Subjectivism were true, they wouldn’t be; Aaron would be saying something about his values and Beth would be saying something about her own, in which case their assertions would be perfectly compatible. This is a very implausible consequence, so we should reject Subjectivism. 2.1 Contextualism need not be self-centered It might seem as though any version of contextualism must succumb to essentially the same objection. But while Subjectivism is a version of contextualism, other versions do not share its flaws. Subjectivism has it that “X is good” is formally a lot like “X is here”; each attributes a relation between X and the speaker, which is why Subjectivism implies that it’s so easy for different speakers to 49 talk past each other. But many context-sensitive expressions are much more flexible. “Nearby,” for instance, provides a much more promising model for a contextualist theory of value, because the location it refers to is not so narrowly constrained. For illustration: Nearby: Carl has run out of gas on the highway. He calls Dana, who asks “Is there a gas station nearby?” Later on Dana tells Eric what happened, concluding “Fortunately there was a gas station nearby.” Dana’s initial question concerns the proximity of a gas station to the location of the person she’s speaking to, rather than her own location; and when she relates what happened to Eric she uses “nearby” to talk about the former location of a third party. And of course the location in question need not be occupied by anyone at all, as when an astronomer observes that the orbit of a faraway planet is influenced by a nearby black hole. So “nearby” gives us a model of an expression the content of which varies widely with the context in which it occurs. 56 Subjectivism is mistaken in taking “good” to be like “here,” but a form of contextualism that uses the model of “nearby” is in a straightforward way much less vulnerable to the objection from lost disagreement. Just as each party in Nearby uses that term to refer to a single property (being near Carl’s location at the time that he ran out of gas), we can generally expect speakers in the same context to coordinate their usage of “good” and “terrible” so that each term to refers to a single property (and consequently, nothing can possess both). Assuming Taxes is a normal context, then, as a first pass we can translate the exchange between Aaron and Beth as follows: Aaron: (3) System V approves of the bill. Beth: (4) System V disapproves of the bill. So for instance, suppose V is a Rawlsian value system, the relevant part of which is the view that any change in the distribution of wealth should be to the benefit of the least well-off. 57 Then Aaron might 56 “Nearby” is what David Kaplan (1989) calls a “pure demonstrative,” since unlike “I” or “now” its content varies with the speaker’s intention (perhaps among other things), and David Perry (1997) calls such an expression a “wide discretionary indexical.” My suggestion (below) is that evaluative terms belong in the same categories. 57 See Rawls (2001: 42-43.) 50 assert (3) because he thinks the poorest members of their community would benefit from the proposed tax bill, and Beth might assert (4) because she thinks it would actually harm that same group. We’ve now recovered the disagreement that Subjectivism lost: (1) and (2) express (something like) (3) and (4), and since those propositions are incompatible, we have an account of the disagreement between Aaron and Beth. In general, when evaluative claims express incompatible propositions, those propositions are about what would be best from the standpoint of a single contextually salient system of values. We’ll return to the problem of lost disagreement soon, but first I’ll quickly turn this general contextualist strategy into more of a proper theory. That will involve dealing with a couple of possible problems with (3) and (4). One is that according to the translations above, the propositions Aaron and Beth are asserting are propositions about value systems, and this seems wrong; intuitively they aren’t talking about the values that are relevant in the context, they’re just saying different things about the bill itself. Relatedly, it’s plausible that people who lack the concept of a value system are nonetheless capable of making evaluative judgments (and that those judgments are justified to a degree that they would not be if they had such sophisticated contents). 58 The solution to this problem is to remove the evaluative content from our final analysis of the proposition expressed by an evaluative sentence. It isn’t that (1) is irreducibly about a value system; rather, it’s about what that system values. So what Aaron is saying (the proposition he asserts) is something like the bill will probably increase the welfare of the poorest citizens, whereas Beth is asserting the contrary proposition that the bill will probably decrease the welfare of the poorest citizens. There are an enormous number of different ways that evaluative terms can be used - “X is good for Y,” “It’s good to A,” “It would be best if p,” “B is a bad C,” etc. - not to mention all the uses of normative vocabulary that a complete theory of their semantics ought to be able to explain. (That is, the theory should account for the logical relationships between assertions regarding what is good, what ought 58 Alex Silk (ms) lodges this objection to contextualist theories of morality. 51 to be done, and what reasons there are.) I am not going to present such a theory; I’ll only explicitly give an analysis of sentences attributing value to states of affairs. But it seems to me that once we understand what it is to ascribe value to states of affairs we have a good foundation for understanding the rest of normative and evaluative discourse. 59 60 Without further ado, here’s the proposal: Preferencist Contextualism (PC): “It is good [a good thing] that X” expresses the proposition that the expected satisfaction of [salient preference set S] is higher given X than given [salient alternative A], given [salient background information B]. To illustrate with an epilogue to Taxes, suppose the bill Aaron and Beth had been debating is passed. If Aaron says, “You’ll see, it really is a good thing that it passed,” PC says that the proposition expressed by the second clause of his utterance is that the expected satisfaction of R [the shared Rawlsian preferences] is higher given that the bill passed than given that the bill did not pass, given B [the information available to Aaron and Beth]. 61 The thesis above is the official statement of PC that I will work with, but the general idea can be captured more intuitively this way: There is nothing that is absolutely good or bad, there is only value from a perspective, and which perspective an evaluative sentence invokes varies with context. Often we adopt a single perspective together, and then the property picked out by your evaluative terms is the same one I designate with mine; but we do not always share a perspective, and then we may use the same terms to think and talk about different properties. 62 59 Stephen Finlay argues for this approach in chapter 2 of his (2014). Even if it turned out that we couldn’t explain all the other ways evaluative terms can be used in terms of this one, we will have come a long way if we can defend contextualism about the value of states of affairs. 60 Since “good” is a gradable adjective, it might seem very strange not to begin with an analysis of its comparative - just like the right way to explain what “tall” means is to start with “taller,” rather than the other way around, we should account for “good” in terms of “better.” I agree that goodness is reducible to betterness (for something to be good is for it to be better than a contextually relevant standard); I’ll just be focusing mainly on “good” because it’s more puzzling and more heavily used. See John Broome (1999) on the reducibility claim. 61 Of course, Aaron and Beth’s original dispute was over the value of an object (the bill) rather than the value of any state of affairs. But this is superficial; whatever the value of the bill, it is pretty clearly to be explained in terms of the value of passing it - which is to say, the states of affairs it might bring about. To ascribe value to an object is, in general, to say that a salient state of affairs involving it yields a higher expected satisfaction of a salient set of preferences than a salient alternative, given the relevant background information. And similarly, to ascribe value to an action is to say that the state of affairs in which that action is performed yields a higher expected satisfaction…and so on. 62 PC identifies these perspectives as sets of preferences, but that isn’t an essential commitment of contextualism, 52 What is expected preference satisfaction though? Essentially, it is the probability-weighted average preference satisfaction of a state of affairs. Assign each possible outcome a number corresponding to its position in the relevant hierarchy of preferences (so that the most preferred is assigned the highest number, the second most preferred gets the second highest, etc. - the scale doesn’t matter, as long as the numbers reflect the ranking). To evaluate a target state of affairs T, find the conditional probability of every possible outcome given the contextually relevant background information plus T, multiply the resulting probability of each outcome by the number representing its preference satisfaction, and then add up the products. This is the expected preference satisfaction of T. T is better than an alternative A iff it has a higher expected preference satisfaction; and T is good iff has a higher expected preference satisfaction than the contextually relevant alternative. All this might seem too complicated. Should we really believe that normal, linguistically naive judgments of value are judgments regarding the expected satisfaction of a contextually determined set of preferences? It’s hard to imagine there are even very many economists who naturally think this way. But PC doesn’t imply that ordinary speakers would even find its analysis an intuitive account of what “good” means, let alone that they actually think in the terms that analysis contains. Knowing how to apply a concept doesn’t require being able to give or even recognize the correct analysis of that concept; even standard examples like bachelorhood are more difficult than they appear. It would be a problem if ordinary speakers’ intuitions disagreed with PC’s implications regarding the extension of “good” in numerous contexts, but we shouldn’t expect them to be able to articulate the rule those intuitions are following. And people routinely exhibit the ability to track expected preference satisfaction; every time we accept a guarantee of a pretty good outcome over a long shot at a slightly better one, or a risky proposition with huge potential benefits over a certain one with a low ceiling, we are employing this capacity. So PC doesn’t require that ordinary speakers have any implausibly sophisticated conceptual or and many of the arguments and considerations to follow could be adduced (or at least adapted) in support of other versions. Finlay (2014), for instance, replaces PC’s preferences with ends, so that to call a state of affairs good is to say that the probability of a contextually salient goal is higher given that state of affairs than given a salient alternative, given salient background information. 53 mathematical resources. This is also a good place to note that in an important sense PC is an objectivist view. One popular complaint about anti-realist theories of morality is that they entail that the moral facts are “mind- dependent” in an objectionable way: if actions are wrong only in virtue of eliciting our disapproval, for instance, this has the consequence that if our attitudes toward the objects of moral assessment had been different, then they would have had different moral properties. 63 But that’s quite counterintuitive; even if we all approved of Hitler’s actions, or believed they were morally permissible, that would not make those actions any less wrong. PC doesn’t entail any such mind-dependence thesis though. That’s because while an evaluative thought or assertion gets its content partly from the contextually relevant perspective, it isn’t about that perspective. If I say “It was cold yesterday,” then which proposition I have expressed depends on what standard of temperature I am using, but that proposition’s truth value doesn’t. Suppose by “cold” I meant “below 40 degrees Fahrenheit” (so that what I asserted was that it was below 40 degrees yesterday) and that it was in fact below 40. Obviously whether the temperature was below 40 yesterday doesn’t depend on my standards of coldness. Similarly, suppose I say “It would be terrible if elephants went extinct,” speaking from my own present perspective. Then I’ve expressed the proposition that the expected satisfaction of my preferences (de re) would be lower, given my information, if elephants were extinct than if they were not. It would be a problem if PC implied that this proposition would have been false if I had had different preferences, because I don’t think the value of the existence of elephants depends on my attitudes toward them. But PC doesn’t have that implication. Another token of the same sentence could be used to assert a false proposition, but that has no bearing on the truth value of the proposition I actually asserted. 2.2 Preferencist contextualism and fundamental disagreement 63“Mind-dependent” will do for our purposes, but actually (despite being commonly used in this context) it isn’t a very perspicuous description. What I really mean is “stance dependent” - see Shafer-Landau (2003: 15-16). 54 PC appears to be able to account for the vast majority of evaluative disagreements in the same way that it handles Taxes, by interpreting them as involving incompatible beliefs about what would be likely to satisfy the relevant sets of preferences. But it seems as though not all evaluative disagreement is ultimately empirical - don’t we sometimes disagree over which of two things is better without disagreeing over what purely descriptive properties they possess? In this section I will argue that there are evaluative disagreements that are intractable in this way, but that they aren’t counterexamples to PC. Examples that clearly illustrate the problem are not easy to devise. On the one hand, it’s almost always an open epistemic possibility, for a given pair of disputing parties, that their evaluative disagreement is derived from a purely descriptive one; and in the remaining cases the intuition that the parties are asserting incompatible propositions is generally weak. If I say that a certain philosophical article is excellent, and my friend thinks it’s awful, there are likely to be some discrepancies in our descriptive beliefs that plausibly explain the difference in our evaluations. And if a Yankees fan and a Red Sox fan disagree over whether it’s a good thing that the Yankees have such an effective pitching staff this year, it is not especially tempting to think that one of them must be expressing a false belief. This is good news for PC, since in the former sort of case the theory can explain what the disputants have incompatible beliefs about (e.g., the soundness of the paper’s arguments), and in the latter sort there don’t seem to be any incompatible beliefs to explain. Indeed, cases like the baseball example are sometimes called “faultless disagreements,” and taken to strongly motivate theories according to which both parties are (at least in some sense) correct. 64 Still, there are some cases that appear more problematic for PC. Here’s one: Ambiguous Admission: Frank is a very talented young man, but his abilities pull him in different directions - he has great potential both as a pianist and as an intellectual, but he can’t be world- class at both. This makes the end of his high school career an unusually climactic time for him, because if he goes to a college with a great music program he will pursue musical performance at the expense of his academic studies. Frank is almost perfectly indifferent, so he decides that he will focus on piano if and only if he is admitted to the school of his musical dreams. As it happens, he gets in. Frank’s parents, Gary and Helen, are at home without him when the letter of acceptance 64 E.g., by Kolbel (2004) and MacFarlane (2007). 55 arrives. Gary opens it and says, (5) “Oh good, he got in!” But Helen is disappointed. (6) “No, it would’ve been better if he hadn’t,” she replies. Upon further discussion, it emerges that they both think Frank would’ve been happy and successful either way; the source of their disagreement is that Gary thinks it’s better that Frank become a happy, successful pianist, whereas Helen thinks it would have been better if he had become a happy, successful academic. No amount of empirical agreement will resolve such a dispute, but Gary and Helen may be more inclined to view one another as having false evaluative beliefs than the rival baseball fans above. Since PC holds that evaluative utterances express incompatible beliefs only when they serve to make conflicting purely descriptive claims, this disagreement looks like a counterexample. How can a view like PC explain why it’s appropriate for Helen to say (6) if she doesn’t think Gary is mistaken about any matter of purely descriptive fact? It looks as though Gary’s (5) expresses the proposition (7) It is good that Frank was admitted whereas Helen uses (6) to assert (8) It is not good that Frank was admitted. But PC says Gary and Helen are just talking past one another, because the most sensible interpretation it can supply is that (5) says (9) The expected satisfaction of preference set G is higher given that Frank was admitted than given that Frank was not admitted (and given background information B) whereas (6) says (10) The expected satisfaction of preference set H is lower given that Frank was admitted than given that Frank was not admitted (and given background information B). (9) and (10) are propositions about different sets of preferences, so they are (at least logically) compatible 56 - and that may seem wrong. But there are several things to be said in PC’s defense. 2.2.1. Disagreement doesn’t require assertions with incompatible contents First, we should not be misled by the fact that Gary and Helen disagree. It is tempting to think that two people can only disagree in virtue of asserting (or at least believing) incompatible propositions, and if that were true then PC would be committed to denying that there is any disagreement in Ambiguous Admission. But actually there are good independent reasons to doubt that disagreement always takes that form. “Faultless” disagreements can be interpreted as counterexamples to that view - if two people have a difference of opinion over whether the Eiffel Tower is beautiful, one intuitively plausible way of understanding their situation is that they disagree even though neither of them necessarily has any false beliefs about the tower. And indeed, that assessment has some empirical support. Khoo and Knobe (ms) have studied ordinary speakers’ intuitions regarding whether two parties making claims of the form “X is morally wrong” and “X is not morally wrong” were disagreeing and whether one party must be saying something false or incorrect. Subjects were significantly more willing to classify these exchanges as disagreements than to say that one party must be saying something false. 65 Darley and Goodwin (2008) also found that subjects’ assessments of whether other people disagreed with them on moral matters diverged widely from their views of whether one or the other must be incorrect. 66 And moral cases are intuitively not even 65 Khoo and Knobe also found that subjects’ reluctance to treat these statements as incompatible increased with the cultural distance between the parties involved - intracultural disagreements were more likely to be seen as involving a conflict, disagreements between a human and an extraterrestrial much less so. Knobe and Seth Yalcin (2014) also found that experimental subjects displayed a strong tendency to treat some disputes over epistemic modals as disagreements in which neither party says anything false. 66 Interestingly, Darley and Goodwin also found that the content of a moral judgment had a large effect on whether subjects thought that disagreements about it must involve incompatible assertions. [Subjects] were far more willing to say that there was a correct answer as to the wrongness of robbery (83%), than they were to say the same about the goodness of giving (52%), and highly unlikely to claim a correct answer as the permissibility of assisted death (17%). (2008: 1352) 57 the strongest candidates for instances of disagreement in the absence of conflicting content; disputes over matters of taste, or over the application of a term like “easy” would seem to provide more plausible examples. 67 The results of these studies could be used to support two different claims. They could be taken as direct support for contextualism about value, since these findings suggest that ordinary speakers (at least implicitly) take it that there are normative disagreements in which neither party is saying something false. I do think that these data are difficult to accommodate for invariantists about normative semantics, but my main point in this section is more modest: that there are strong independent reasons not to infer from the mere fact that Gary and Helen disagree that they must be using (5) and (6) to assert incompatible propositions. 68 They do seem to disagree, but that doesn’t show that Ambiguous Admission is a counterexample to PC. 2.2.2 A contextualist-friendly account of the disagreement between Gary and Helen Still, it might be argued that the best explanation for the fact that Gary and Helen disagree over whether Frank’s admission is a good thing is that they are asserting incompatible propositions, so even if cases like this aren’t direct counterexamples to PC, they still provide strong grounds for rejecting it. This section offers an alternative explanation. One thing such an explanation must cover is why it’s felicitous for Helen to say “no” in rejecting Gary’s statement. It might seem that “no” always expresses disbelief, but Khoo and Knobe (ms) also found that their subjects viewed “no” as an appropriate way of expressing disagreement even in cases where they thought both parties could be correct. Elsewhere, in a paper on disagreements involving epistemic modals, Khoo (2015) suggests that in such a context “no” may instead express denial of a 67Or terms like “easy,” which I discuss in 2.2.3. 68Relativism and contextualism share some motivations, but there is an important difference here. Relativists take the view that disagreements may be faultless because the object of their dispute is a proposition that is true for one but false for the other, which means that their view faces the same problem that an invariantist has - if relativism were true, then the subjects in these studies ought to regard those who disagree with them as asserting false propositions. 58 presupposition conveyed by the target assertion. That explains why “no” is sometimes felicitous even when the speaker has no reason to doubt the truth of the preceding claim; she might think that it just conveyed a false presupposition. This sort of view can also handle disagreements like Ambiguous Admission. There are different ways of filling in the details - e.g., which presupposition is rejected and whether this rejection expresses disbelief or something like refusal to accept - but here’s one such account: When Gary says (5) “Oh good, he got in!” the proposition he asserts is (9) The expected satisfaction of preference set G is higher given that Frank was admitted than given that Frank was not admitted (and given background information B). Helen might not be sure exactly which set of preferences Gary has in mind, but she is justified in believing that Gary has either said or presupposed something false. If Gary intends (5) as a claim about what increases the expected satisfaction of both his preferences and Helen’s, then he’s wrong about Helen’s preferences, and consequently about what is good from their joint perspective. Alternatively, Gary might intend (5) as a claim about some other set of preferences, but Helen could nonetheless be confident that Gary was presupposing that her preferences were similar, since it would be rude of him to adopt a perspective for their conversational purposes that he knew she didn’t share. Either way, Helen can use “no” to signal that she denies part of what Gary’s assertion conveys (whether that part is in the content of his assertion or in its presuppositions); and this also explains why it makes sense to describe them as disagreeing. 69 Objection: Suppose Helen knows that Gary knows that from her perspective it isn’t a good thing 69 Maybe this doesn’t explain how Gary disagrees with Helen; but it isn’t so intuitive that there is a disagreement in that direction. If Gary either asserted or presupposed something false about Helen, and her utterance of (6) conveys this to him, and he has no special reason to doubt that she knows her own preferences, then he ought to just immediately concede that he was wrong. Of course, he need not express that by saying “Oh, I guess it isn’t good”; he could instead just say something explicitly about his own perspective, like “Well, I’m glad he got in at least.” See Finlay (2014: chapter 8) on asymmetric evaluative disagreements. 59 that Frank was admitted. It seems as though she can still felicitously utter (6), even though Helen has no reason to think Gary has said or presupposed anything that isn’t true. So the above account just dodges the most troublesome cases. There are a few things to be said on PC’s behalf here. First, while it does appear to make sense for Helen to say “No” in this case, at least for me there is no intuition that one party or the other must have a false belief. Gary and Helen wouldn’t seem to be contradicting one another any more than two swimmers one of whom says “The water is warm” while the other responds “No it isn’t.” These are people who are knowingly talking past each other. This may or may not still seem to be a case of disagreement; I have no strong intuition either way. If this new version of the case isn’t a disagreement, then PC needs no further resources to explain why that is; Gary and Helen don’t disagree because neither thinks the other has said or presupposed anything that isn’t true. But suppose this is still a disagreement; there are at least a couple of contextualist-friendly ways of explaining why that is. One is Stephen Finlay’s (2014, ms) “quasi-expressivist” view of non-content- based evaluative disagreement. On this view, holding conflicting non-cognitive attitudes toward a single object is sufficient for disagreement, and evaluative sentences can be felicitous in virtue of expressing those attitudes even when they contribute no new propositional information. So as long as Gary and Helen are expressing conflicting preferences, they are disagreeing. Alternatively (or possibly in addition) this might be what Plunkett and Sundell (2013) call a “metalinguistic” disagreement. In that case, rather than trying to change one another’s substantive beliefs, Gary and Helen would be negotiating over how to use their evaluative terms - when Gary says that it’s good that Frank was admitted, he is implicitly proposing that they use “good” in such a way that his description is accurate, and the point of rejecting his assertion for Helen is to put forward her own proposal regarding which set of preferences to relativize their usage to. (The swimmers above are engaged in another such metalinguistic disagreement, according to Plunkett and Sundell.) 60 2.2.3 Companions in “guilt” Maybe the foregoing proposal for understanding the disagreement between Gary and Helen isn’t right; it is certainly not the only one a contextualist might try. But perhaps more importantly, there are a couple of good reasons to think that whatever the right account is, it must be consistent with contextualism. One of these is that the puzzling linguistic phenomena we’ve been considering are not distinctive of evaluative disputes, or even of disputes in which it is at all plausible that the predicates involved have absolute meanings. For instance, the view that “easy” attributes the same property in every context is not very attractive; to say that something is easy is always to assert that it is easy when measured against some contextually salient standard. It might be accurate to characterize a certain arrangement of billiard balls as presenting “an easy shot” in the context of a professional tournament, but not in the context of a game between amateurs. Nonetheless, disagreements over what’s easy display the same surface features that ground the argument against contextualism about value that we’ve been discussing. Here’s a simple example: Physics Exam: Igor and Janet are high school students who have just taken a physics exam. (10) “Well, that was easy,” Janet says. (11) “No it wasn’t!” Igor replies. This disagreement might be a substantive one, in which Janet and Igor have incompatible views of whether the exam meets a certain shared standard of easiness - but it need not be. Even if Igor knows that Janet means the test was easy for her, and has no reason to doubt that it was, his reply is felicitous as long as the test wasn’t easy by his lights. The point is not that we can see exactly what’s going on here and then adapt that explanation to solve the disagreement puzzle for contextualism about value; it’s that whatever the explanation is for the felicitousness of Igor’s saying “No it wasn’t,” it appears safe to assume that it will also cover Helen’s reply to Gary. And that explanation will be consistent with contextualism, because “easy” is transparently a context-sensitive predicate. So whether the particular account I proposed in the previous section is correct or not, the puzzling nature of cases like Ambiguous Admission provides no good reason to reject 61 contextualism about value. 70 2.3 The problem of excessive modesty There is another strong independent reason to accept that these cases do not feature parties asserting incompatible beliefs: rejecting that view would saddle us with a serious epistemological problem regarding the significance of information about others’ evaluative views. If Gary and Helen are (sincerely) asserting incompatible propositions, then one of them must have a false belief. 71 That presents a dilemma: Either they should each reduce confidence in their evaluations of Frank’s admission to his music school of choice, or they should not. The first horn is both directly counterintuitive and has further implausible consequences. And there is no plausible explanation for the latter, if Gary and Helen can’t both be right. 72 2.3.1 Neither party should reduce confidence in the face of purely evaluative disagreement Suppose Helen’s disagreement rationally requires Gary to reduce confidence in his belief that it was a good thing that Frank was admitted (even though he knows this disagreement doesn’t derive from any non-evaluative one). This is the right way to respond to most disagreements; to the extent that we respect someone else’s epistemic credentials, we ought to give some credence to the possibility that she is correct and we are mistaken. But even at first glance, in this case it doesn’t seem as though Gary’s confidence in his evaluation should really be shaken by Helen’s dissent - there is some sort of conflict between them, but not the sort that rationally requires anyone to change his mind. And some of the consequences of imputing this epistemic significance to others’ value judgments 70 Cappelen and Hawthorne (2010: chapter 4) employ a somewhat similar argument for contextualism over relativism about predicates of taste; they argue for contextualism about “filling,” and then that terms like “fun” and “disgusting” fit the same mold. 71 Relativism denies this, maintaining that two people can assert incompatible propositions without either of them expressing a false belief, and so avoid the dilemma below. At best though, this is very difficult to understand, and see Cappelen and Hawthorne (2010) for extensive arguments that relativism ultimately doesn’t help to solve the puzzles that are supposed to motivate it anyway. 72 For a much more thorough version of this argument, see “The Special Significance of Evaluative Testimony.” 62 are even more unpalatable. If disagreeing with Helen ought to reduce Gary’s credence that it was a good thing that Frank got in, then presumably the more well-informed people agree with Helen, the lower Gary’s credence should sink. In that case, rationality requires that Gary give up whatever evaluative views he would otherwise be inclined to, provided he knows they are sufficiently unpopular. Indeed, every such unpopular opinion he holds is evidence that his evaluative judgment is faulty in general, so if he is very idiosyncratic then rationality requires him to distrust it, in the same way that someone whose memories regularly differ from those of his associates ought to develop some skepticism about its reliability. In fact, if one’s personal evaluative judgment is no more epistemically weighty, in general, than information about what other people think on the same subjects, then it should apparently be rational to outsource even fundamental evaluative labor. This is what most of us do on scientific matters; we don’t review the evidence ourselves, we just rely on the testimony of people we have reason to believe are well- informed themselves. Intuitively though, it would not be rational to defer in such dramatic ways regarding essentially evaluative questions. And since these implausible consequences follow from the view that Gary should reduce confidence in his belief when confronted with Helen’s disagreement (or vice versa, of course) we should reject it. 2.3.2 ...but not because each party should be sure that the other is wrong The other horn of our dilemma accepts that fundamental disagreements like Gary and Helen’s produce no rational pressure for the parties involved to change their minds. This seems right, for all of the reasons above. The problem is that there is no remotely plausible explanation for why these disputes are so epistemically inconsequential, if one of the participants must have a false belief. There are circumstances in which it makes sense to disregard others’ dissent; one might have strong reason to believe the other party is ill-informed, insincere, or prone to mistakes in her reasoning, and in those cases her testimony is not a good guide to the truth. But fundamental evaluative 63 disagreements are peculiar in that their epistemic insignificance does not depend on the participants having evidence of any of these defeating conditions. Less commonly, it can also be rational to take disagreement lightly when one has evidence that it stems from a conflict in basic impressions, and that the other party’s are far less reliably accurate in general. If Superman seems to hear a knock on the door, my testimony that I didn’t hear anything should not diminish his confidence that someone knocked, as long as he knows that I don’t have his extraordinary senses. But it is hard to believe that each of us can rationally trust her own evaluative faculties over others’ in the same way that it makes sense for Superman to trust his hearing over that of a mere mortal. There is no reasonable explanation for why someone in Gary’s position should think that in case of evaluative disagreement, he is overwhelmingly more likely to be correct than Helen is; the only good explanation for the peculiar epistemic insignificance of this sort of disagreement is that the assertions involved don’t really conflict. So the implication that is supposed to be contextualism’s Achilles’ heel is actually a datum that every metaethical theory should try to accommodate. 2.4 Summary of Section 2 The most serious objection to contextualism about value is that it mistakes some disagreements for cases in which the parties involved are just talking at cross purposes, but a nuanced form of contextualism like PC can explain how we generally coordinate our usage so that all the parties in a single conversation use “good” to attribute the same property. This is not really very remarkable; even though “nearby” can be used to designate an indefinite number of different properties according to the context in which it is used, within a single interaction each party generally applies it in the same way. Contextualist views like PC are nonetheless distinctive in that they imply that fundamental evaluative disagreements involve parties asserting compatible propositions. There might seem to be a contradiction in terms there, but recent empirical research (and careful consideration of certain cases) provides independent support for the thesis that people can disagree without taking each other to be 64 saying anything false. One way for a contextualist to explain what’s going on in these cases (like Ambiguous Admission) is that while each party may strictly be asserting something true, he may still justifiably be interpreted by his conversational partner as presupposing something false or otherwise objectionable, which is why that partner can sensibly reject his claim. And even if this particular contextualist account of disagreement isn’t correct, there are two strong reasons to think that “lost disagreement” can’t ground any serious objection to views like PC. First, this sort of objection to contextualism about value would seem to be an equally good objection to the view that terms like “easy” are context-sensitive, and it’s hard to believe that contextualism isn’t the right way to understand those expressions. And second, the fact that it is rational to maintain confidence in one’s opinion in the face of pure evaluative disagreement can only be adequately explained by a view according to which the parties involved are asserting compatible propositions anyway, so this commitment is far from a reductio of contextualism. 73 3. The positive case for contextualism So far I’ve been arguing that the main objection to contextualism is ultimately not a compelling one; in this section I will turn to the advantages of adopting PC. As before though, similar arguments could be used in support of other versions of contextualism. 73 This is a good place to note that the arguments I’ve just summarized could be taken to support an error theory rather than a contextualist one. Broadly speaking, I have been arguing that contextualism makes good sense of ordinary evaluative disagreement and (in 2.3) that absolutism can’t. But an error theorist might say that while it might make more sense if we treated evaluative terms as context-sensitive, that is not our actual practice; ordinary speakers aren’t fancy contextualists, they are absolutists whose false presupposition that value is a single property sometimes leads them to engage in misguided disputes. We could object on grounds of charity, or because it isn’t plausible that evaluative discourse is so riddled with error, but the most compelling reason to reject the error theory is more direct: it just doesn’t explain the relevant phenomena very well. If we are naive absolutists, then why don’t we change our minds in cases of fundamental evaluative disagreement? On that view, Gary and Helen ought to take it that one of them must be incorrect, and they ought to find that epistemically troubling, but that just isn’t what happens in such cases. The best explanation for this observation is that the parties involved don’t take themselves to be asserting incompatible propositions, and that suggests they are implicitly contextualists. For a different response to the analogous objection against contextualism about morality, see Harman (1978). 65 3.1 An explanation for the extent of evaluative diversity Some of the most prominent contextualists about morality view its ability to explain the actual diversity of moral codes, and the prevalence of intractable disagreement between them, as that theory’s main selling point. 74 There is an analogous (but stronger) argument for contextualism about value. It’s a simple inference to the best explanation, where the explanandum is the fact that there is a lot of fundamental diversity - that is, diversity that is not due to differences in non-evaluative belief - in what people value. In some cases one of us values something that another doesn’t (intrinsically) value at all: patriotism; pride; fame; various forms of art; the welfare of non-human animals or ecosystems. And those are only the starkest differences. Much more commonly, we differ over how the values of two different things compare - the rates at which it makes sense to exchange freedom for security, benefits to ourselves for benefits to others, happiness for achievement, present goods for future ones, authenticity for success. The extent of moral disagreement is a contentious issue, which makes it debatable how strong the IBE argument for moral contextualism is; but the variety of fundamentally incompatible evaluative views is both wider and more obvious. If value is a single property, then the explanation for this diversity must be either that there is some serious flaw in whatever our mechanism for detecting it is, or that sources of distortion are rampant. On the other hand, PC can explain this diversity as a product of the diversity of human preferences - we think of and describe different things as good or better simply because we don’t all care about the same things to the same degrees. 75 3.2 A nuanced account of internalism 74 See Gil Harman (1978, 1996) and David Wong (2011). John Mackie (1977) takes his “argument from relativity” to support an error theory, but it could be taken as an argument for contextualism instead. 75 For an extensive survey of IBE disagreement arguments against objectivist views of morality, see Gowans (2000); and for a reply on behalf of realism see Enoch (2011: chapter 8). In my view the availability of defenses like Enoch’s strongly mitigates the force of these arguments, but since his defense depends on the arguably limited scope of moral disagreement and the hope that relatively little of that disagreement is fundamental, it isn’t much help to an objectivist about value. 66 We can generally infer from S’s saying “It is good that/would be good if p” that S is motivated to ensure that it becomes or remains the case that p. This can seem puzzling though, because while it seems as though such evaluative sentences express beliefs, it is also plausible that beliefs are not inherently related to motivational states - the influence that a belief has on a person’s behavior is mediated by his desires, which are completely independent aspects of a person’s psychology. So what explains this tight connection between motivation and evaluative belief? 76 PC has an attractive answer: Evaluative assertions express beliefs about what is conducive to the satisfaction of a contextually salient set of preferences. If we are rational, preference for a certain end motivates us to seek the known means to that end. So if a rational person shares the perspective (the set of preferences) from which an evaluative assertion is made, he will be motivated to pursue what it classifies as good. It might seem as though this only explains why a person who knows her own preferences is generally motivated to pursue what she believes to be good; if she didn’t, she wouldn’t be able to tell whether or to what degree she shared the set of preferences her belief was about. And it is true that according to PC, our evaluative beliefs only cause appropriate behavior when we have this self- knowledge. But the explanation for the fairly tight relationship between motivation and evaluative belief is not primarily that the latter regularly causes the former. We seek what is good from our perspectives, but the process isn’t normally mediated by thoughts containing normative or evaluative concepts; we simply have preferences among states of affairs, and those preferences motivate us directly. For the most part what explains internalism is that when we do deploy these relatively hifalutin concepts, we generally use them to think about the preferences we actually have - with the consequence that we are usually motivated to pursue what we classify as good. Not always, though, because we can adopt perspectives other than our own. This is another point 76 This is the evaluative analog of the eponymous puzzle in Michael Smith’s The Moral Problem (1994). 67 in PC’s favor, because the relationship between calling something “good” and one’s feelings and motivations regarding it is not as tight as it seems at first glance. Backgammon: Karen and Leo are playing backgammon, and Karen is winning - but then Leo gets a lucky roll that turns the tables. “Oh, that’s good,” Leo says. “Yeah,” Karen glumly agrees. PC can make sense of this sort of “exocentric” use of evaluative language, in which the relevant perspective is not the speaker’s own; Karen is saying that Leo’s roll increased the expected satisfaction of his preferences, not hers. 77 That explains why she can sincerely speak as she does even though she isn’t at all pleased about Leo’s roll. So PC offers a satisfactorily nuanced account of internalism. Usually we speak with reference to our own perspective (or one we share), so that our evaluative sentences express beliefs about what suits our preferences, and to prefer or disprefer something entails being appropriately motivated. 78 But sometimes we speak from another perspective, and in those cases there is no such straightforward relationship. PC thus explains why there is a strong but not exceptionless connection between judging something to be good and being motivated to promote it. 3.3 Naturalism Contextualist views like PC also have a cluster of advantages related to their naturalism. It’s notoriously difficult to explain what naturalism is, in the relevant sense, because it’s the view that everything that value is a natural property, and it’s hard to understand the complementary thesis. But for present purposes it is sufficient that contextualism is not committed to the existence of anything that could not (in principle) be studied scientifically. 79 “Good” is no more the name of a special, elusive 77 The distinction between autocentric and exocentric uses is from Lasersohn (2005). 78 We can sincerely believe that something would be good, from our own perspective, without being even slightly motivated to do anything about it at the time of the judgment (as when we are very tired or depressed). But I don’t see how someone could have such a belief if he didn’t think there were any circumstances in which he would act accordingly. See Michael Stocker (1979) and James Dreier (1990). 79 On the meaning of “natural” in this context, see Enoch (2011: Chapter 5) and Shafer-Landau (2003: 58-65). 68 property than “here” is the name of a very slippery place. This has some ontological appeal. First, it is parsimonious; a naturalist’s universe is composed of fewer basic entities than that of someone who believes there is a property of being good that is fundamentally distinct from all those that our investigation of the purely descriptive ones could ever give us reason to believe in. And second, according to PC there is nothing “queer” about the subject matter of evaluative discourse - it’s just a matter of preferences and probabilities. 80 Similarly, the fact that normative and evaluative properties supervene on natural ones is not the problem for PC that it is for non-naturalist metaethical theories. There is a simple explanation for why two things can’t differ in their evaluative properties without differing in their natural ones: the property that “good” attributes is the natural property of comparatively high expected preference satisfaction. By contrast, if value were not a natural property itself, it would be puzzling why it is so intimately related to properties that are. 81 And not only is PC ontologically and metaphysically straightforward; it also avoids reliance on obscure concepts. I’ve already mentioned how hard it is to understand what a “non-natural” property could be, but non-naturalist realism isn’t the only metaethical theory to lean on a problematic notion. Relativists propose to solve the problem of lost disagreement by maintaining that the parties involved in a fundamental evaluative disagreement really are expressing incompatible beliefs, and yet that what each of 80 It might seem as though taking this view that evaluative properties are nothing special must come at the cost of an inability to explain why we don’t seem to be able to equate them with natural properties - that is, why Moore’s open questions are open. For instance, suppose the property “good” refers to in a certain context is yielding a higher expected satisfaction of a preference set S, given background information B and alternative A. Then PC implies that in that context the question “Is X, which yields a higher expected satisfaction of S than A given B, good?” is closed. But it might seem that such a question would always sound open. That question actually doesn’t sound especially open to me. But even if I am idiosyncratic in this respect, Finlay points out that contextualism can supply a good pragmatic explanation for the appearance of openness: Since “good” is a context-sensitive expression, the audience for such a question has to determine how the speaker is using it in order to understand what is being asked. Interpretations according to which the speaker is either asking for information she could not possibly lack - i.e., in which the question is closed - or incompetent with the words she is using, are virtually never reasonable interpretations. So the audience automatically searches for a preference set the speaker might have in mind that isn’t identical to the one she has explicitly mentioned - which is to say, an interpretation according to which the question is open. See Finlay (2014: 20-22). 81 Dreier (1992) argues for a relative of PC regarding moral terms partly on the basis of supervenience considerations. 69 them believes is true-relative-to-her. Other theories attempt to explain why normative beliefs are so closely tied to motivation by invoking the concept of an inherently motivating belief. 82 3.4 Epistemology The considerations above, especially the ability to account for the prevalence of disagreement, are relatively familiar motivations for contextualism. But the epistemological advantages that a view like PC offers are underappreciated. Of course, the metaphysical advantages covered in the previous section come with an obvious epistemological one. If normative and evaluative language is just a special way of talking about ordinary natural properties and facts, then a competent speaker can be justified in her evaluative beliefs simply in virtue of being justified in some unmysterious empirical ones. I don’t need to know anything of an irreducibly evaluative nature in order to be justified in believing that I should walk home before it starts raining; the fact that I would prefer not to be rained on is all the information I need. This means that a contextualist need not suppose that we have a special faculty, distinct from our physical senses, memory, and so on that accounts for our fundamental evaluative knowledge. That gives PC an(other) edge in parsimony, and allows it to avoid difficult questions like what explains our having such a faculty and how it might work. PC can simply invoke whatever explains our ability to identify preferences and the effective means of satisfying them; the justification for those empirical beliefs transmits to our evaluative ones. 83 The foregoing epistemological advantages are not unique to contextualism; naturalistic forms of realism can offer similarly attractive explanations for our knowledge of value. But no form of realism - assuming we understand realism as entailing that value is a single property - can explain why each of us as individuals rationally ought to rely on his or her own judgment to the remarkable extent that we do. In 82 Even quasi-realism has this implication, since it concedes that our evaluative attitudes are beliefs and yet holds that they are also inherently motivating. See Gibbard (2003). 83 See “Reliability Challenges and Etiological Undermining in Metaethics.” 70 general we should weigh the testimony of other well-informed people quite heavily in forming our own beliefs, but as I argued in Section 2.3, it would not be rational to take their evaluative opinions so seriously. The only good explanation for this fact is that value is not a single property, so that the properties that other people use their evaluative terms to pick out are not necessarily the same properties we are interested in; and this gives contextualism an epistemological advantage that no mainstream metaethical theory can claim. 84 85 4. Conclusion There is supposed to be a decisive objection against any form of contextualism about value - that it mistakenly categorizes some genuine disagreements as instances of parties merely talking past one another. But a flexible version of contextualism like the PC can explain the vast majority of evaluative disagreements in the same way that an ordinary absolutist view can,as cases in which the disputants are asserting incompatible propositions. And in the remaining cases - the fundamental evaluative disagreements - there are both promising explanations available according to which no such conflicting contents are involved and good independent reasons to accept that diagnosis. That is especially interesting because there is so much to be said for contextualism on other counts. The most commonly cited motivation for the analogous view of morality has been contextualism’s ability to explain why there is such a wide range of fundamentally divergent beliefs, and that consideration favors contextualism about value even more heavily. PC also provides an especially satisfying account of the practical, action-guiding character of evaluative judgment without sacrificing the cognitivism that is such an appealing aspect of realist views; and it is as naturalistic and ontologically parsimonious a theory as possible, since it holds that “good” is just a special way of referring to a contextually determined natural property of the kind that economists study. But the most distinctive 84 Relativists can also claim this advantage, but the debate between relativism and contextualism is sufficiently different (and the audience for it is sufficiently small) that I won’t address it here. 85 This is not a complete list of what views like PC have going for them; in fact the most thoroughly developed version of contextualism about value - Finlay’s - is motivated primarily by linguistic considerations (2014: 246-48). 71 advantage of contextualism about value is its ability to explain why we rely so heavily on our personal evaluative judgment, even when there are experts concerning the object of evaluation whose views we could adopt instead or we are part of a small minority. Taken together, and given the ultimate weakness of the lost disagreement objection, this constitutes a strong case for contextualism about value. 72 References Bjornsson, G. and Finlay, S. (2010) Metaethical Contextualism Defended. Ethics 121(1), 7-36. Broome, J. (1999) Goodness is Reducible to Betterness: The Evil of Death is the Value of Life. Ethics Out of Economics. Cambridge, UK: Cambridge University Press. Copp, D. (1995) Morality, Normativity, and Society. New York: Oxford University Press. Darley, J. and Goodwin, G. (2008) The Psychology of Meta-ethics: Exploring Objectivism. Cognition 106: 1339-1366. Cappelen, H., and Hawthorne, J. (2010) Relativism and Monadic Truth. Oxford: Oxford University Press. Dreier, J. (1990) Internalism and Speaker Relativism. Ethics 101(1), 6-26. _____. (1992) The Supervenience Argument against Moral Realism. Southern Journal of Philosophy 30, 13–38. Enoch, D. (2011) Taking Morality Seriously: A Defense of Robust Realism. New York: Oxford University Press. Finlay, S. (2014) Confusion of Tongues: A Theory of Normative Language. Oxford: Oxford University Press. Gibbard, A. (2003) Thinking How to Live. Cambridge, MA: Harvard University Press. Gowans, C. (2000) Introduction. In C. Gowans (ed.) Moral Disagreements: Classic and Contemporary Readings. London: Routledge, 1-43. Harman, G. (1975) Moral Relativism Defended. The Philosophical Review 84(1), 3–22. _____. (1978) What is Moral Relativism? in A. Goldman and J. Kim (eds.) Values and Morals. Boston: D. Reidel. _____. (1996) Moral Relativism. In G. Harman and J.J. Thompson (eds.) Moral Relativism and Moral Objectivity. Cambridge MA: Blackwell Publishers, 3–64. Kaplan, D. (1989) Demonstratives. In J. Almog, J. Perry, and H. Wittstein (eds.) Themes from Kaplan. Oxford: Oxford University Press. 481–563. 73 Khoo, J. (2015) Modal Disagreements. Inquiry 58(5), 511-34. Khoo, J., and Knobe, J. (ms) Moral Disagreement and Exclusion. Knobe, J., and Yalcin, S. (2014) Epistemic Modals and Context: Experimental Data. Semantics and Pragmatics 7(10), 1-21. Kolbel, M. (2004) Indexical Relativism versus Genuine Relativism. International Journal of Philosophical Studies 12, 297-313. Kraut, R. (2011) Against Absolute Goodness. Oxford: Oxford University Press. Lasersohn, P. (2005) Context Dependence, Disagreement, and Predicates of Personal Taste. Linguistics and Philosophy 28(6), 643-686. MacFarlane, J. (2007) Relativism and Disagreement. Philosophical Studies 132, 17-31. Mackie, J.L. (1977) Ethics: Inventing Right and Wrong. Harmondsworth: Penguin. Perry, J. (1997) Indexicals and Demonstratives. In B. Hale and C. Wright (eds.) A Companion to Philosophy of Language. Oxford: Blackwell Press, 586-612. Plunkett, D. and Sundell, T. (2013) Disagreement and the Semantics of Normative and Evaluative Terms. Philosopher’s Imprint 13, 1-37. Rawls, J. (2001) Justice as Fairness. E. Kelly (ed.), Cambridge, MA: Harvard University Press. Shafer-Landau, R. (2003) Moral Realism. New York: Oxford University Press. Silk, A. (ms) Truth Conditions and the Meanings of Ethical Terms. Smith, M. (1994) The Moral Problem. Oxford: Blackwell. Stocker, M. (1979) Desiring the Bad: An Essay in Moral Psychology. Journal of Philosophy 76, 738-53. Velleman, J.D. (2013) Foundations for Moral Relativism. Cambridge, UK: Open Book Publishers. Wong, D. (1984) Moral Relativity. Berkeley: University of California Press. _____. (2006) Natural Moralities: A Defense of Pluralistic Relativism. New York: Oxford University Press. _____. (2011) Relativist Explanations of Interpersonal and Group Disagreement. In S.D. Hales (ed.), A Companion to Relativism. Malden, MA: Wiley-Blackwell, 411–29. 74