Naturalizing Metaethics Jesse Prinz

Decades ago, it was suggested that epistemology could be naturalized, meaning, Author roughly, that it could be treated as an empirically-informed psychological inquiry. In more recent years, there has been a concerted effort to naturalize ethics, with a Jesse Prinz focus on questions in moral psychology, and occasional normative ethics. Less ef- [email protected] fort has been put into the naturalization of metaethics: the study of what, if any- thing, makes moral judgments true. The discussion presents a systematic overview City University of New York of core questions in metaethics, and argues that each of these can be illuminated New York, NY, U.S.A. by psychological research. These include questions about realism, expressivism, error theory, and relativism. Metaethics is beholden to moral psychology, and Commentator moral psychology can be studied empirically. The primary goal is to establish em- pirical tractability, but, in so doing, the paper also takes a provisional stance on Yann Wilhelm core questions, defending a view that is relativist, subjective, and emotionally [email protected] grounded. Johannes Gutenberg-Universität Mainz, Germany Keywords Error theory | Expressivism | Metaethics | Moral realism | Naturalism | Relativ- Editors ism | Sentimentalism Thomas Metzinger metzinger @uni-mainz.de Johannes Gutenberg-Universität Mainz, Germany

Jennifer M. Windt [email protected] Monash University Melbourne, Australia

1 Introduction will not present these results in detail here. My goal is programmatic: I seek to establish Moral philosophy has taken an empirical turn, the empirical tractability of metaethics. Some with experimental results being brought to of the experiments I describe are exploratory bear on core questions in moral psychology pilot studies, presented in an effort to motiv- (e.g., is altruism motivated by empathy?) and ate more research. Even with such preliminary normative ethics (e.g., how plausible are the results, we will see that some metaethical the- presuppositions of virtue theory?). Some of ories already enjoy greater empirical support the recent empirical work also bears on core than others. I will argue that the best-suppor- questions in metaethics. Metaethical questions ted theory at this stage of inquiry is a form of are varied, but they broadly concern the relativist sentimentalism. Defending this posi- foundations of moral judgments. What is the tion is subsidiary to my primary goal of ad- basis of such judgments? What, if anything, vertising the value of empirical methods in could render them true? Here I will argue that metaethical theorizing. There has already these questions can be empirically addressed, been an empirical turn in ethics, but and longstanding debates between leading metaethics has been less explicitly targeted by metaethical theories may ultimately be settled these new approaches. experimentally. I will describe empirical res- Talking about “an empirical turn” clearly ults that bear on core metaethical questions. I alludes to another turn in the recent history of

Prinz, J. (2015). Naturalizing Metaethics. In T. Metzinger & J. M. Windt (Eds). Open MIND: 30(T). Frankfurt am Main: MIND Group. doi: 10.15502/9783958570511 1 | 27 www.open-mind.net philosophy: the linguistic turn. When philosoph- W.V.O. Quine, who grew skeptical about philo- ers turned their attention to language, there sophizing through linguistic analysis, and em- was an effort to recast philosophical problems phasized the empricial revisability of philosoph- as linguistic in nature. A new set of technical ical claims (1969). Quine drew on the methods tools was brought into the field: formal se- of John Dewey, and insisted that “knowledge, mantics. Logic has been part of philosophy his- mind, and meaning […] are to be studied in the torically, but after the linguistic turn it was per- same empricial spirit that animates nautral sci- ceived to be an essential component of philo- ence” (1969, p. 26). More succinctly, methodolo- sophical training. Just as formal semantics in- gical naturalism can be defined as follows: creased philosophical precision with the lin- guistic turn, empirical methods have dramatic- Methodological naturalism =Df the view ally augmented our tool chest, and stubborn de- that we should study a domain using em- bates may begin to give way. The empirical turn pirical methods. is as momentous as the linguistic turn, and per- haps even more so. Formal semantics allowed us This is the kind of naturalism that has long to articulate differences between theories, and been advocated, but too rarely followed, in the empirical methods provide new opportunities domain of epistemology (Kornblith 1985). for theory confirmation. Neither turn rendered Neither metaphysical nor semantic naturalism traditional approaches to philosophy idle, but are equivalent to methodological naturalism. rather supplemented them. Within metaethics, Metaphysical naturalism is a view about what this supplementation may offer the best hope of exists, not about how to study it. Indeed, some settling which competing theories are true. non-naturalists in this metaphysical sense be- In calling for a naturalist metaethics, it is lieve that empirical methods can be used to important to avoid confusion with two other study non-physical or supernatural entities. Se- views. “Naturalism” is sometimes construed as a mantic naturalism is a view about how to state metaphysical thesis, and also sometimes as a se- theories (viz., in reductionist terms), but practi- mantic thesis. Metaphysically, “naturalism” tioners have rarely used empirical science in de- refers to the view that everything that exists fense of such theories (consider so-called natur- belongs to the natural world, as opposed to the alistic semantics). Methodological naturalism non-natural, or supernatural world. This is has been deployed in discussions of both first- sometimes presented as a synonym for physical- order ethics (e.g., Brandt 1959; Flanagan 1991; ism, which can be defined as the view that the Doris 1998; Greene 2007) and in metaethics world described by the physical sciences is com- (e.g., Railton 1993; Prinz 2007b). As Railton plete, in that any physical duplicate of this points out, a naturalist methodology could res- world would be a duplicate simpliciter. The ult in a reductionist theory of morality, but it causal closure of the physical world and the suc- need not (see also Boyd 1988). In principle, sci- cess of physical science are taken as evidence for ence could support traditional intuitionism, this metaphysical view. Semantic naturalism at- which is not naturalistic in either of these other tempts to reductively analyze concepts from one senses. domain in terms of another, which is considered more likely to be natural in a metaphysical 1.1 Methodological preamble sense. In philosophy of mind, this might involve defining psychological concepts in neural or Philosophy has always been methodologically causal terms, while in ethics it might involve de- pluralistic. Some use intuitions to arrive at ne- fining moral properties in terms of psycholo- cessary and sufficient conditions for the applica- gical, logical, or social terms (such as hedonic tion of concepts (e.g., Plato’s early dialogues). states, principles of reason, or social contracts). Some try to systematize and revise a large set Here I will be concerned with methodogical nat- of beliefs using reflective equilibrium (e.g., uralism, which has recent roots in the work of Rawls on justice). Some use transcendental ar-

Prinz, J. (2015). Naturalizing Metaethics. In T. Metzinger & J. M. Windt (Eds). Open MIND: 30(T). Frankfurt am Main: MIND Group. doi: 10.15502/9783958570511 2 | 27 www.open-mind.net guments to figure out preconditions for thought draws on the psychology of decision-making, and action (e.g., Kant). Some use aphorisms or and philosophy of mind, which has drawn on stories to reveal facts about ourselves or to envi- psychology, computer science, and artificial in- sion possible alternatives (e.g., Nietzsche and telligence. Over the last decade, empirical meth- the existentialist tradition). Some propose his- ods have also become widely used, and widely torical analyses of prevailing institutions and contested, in ethics. values (e.g., Hobbes, Rousseau, and Foucault). The resistance to empirical methods in Some disclose hidden social forces that buffer ethics is often chalked up to the fact that ethics prevailing categories (e.g., Marilyn Frye on is a normative domain, and empirical methods gender). Some analyze case studies (e.g., Kuhn), provide descriptive results. This can only be probe the structure of experience (e.g., Husserl), part of the story, however, as there has been or propose formalizations (e.g., Frege). These little uptake of empirical methods in and other methods suggest that philosophy is a metaethics. Metaethics is a descriptive domain; many-splendored thing, and among its many it does not tell us how to act morally, but forms one can also find the deployment of em- rather explores the semantic commitments and pirical results. Examples include Descartes and metaphysical foundations of such claims. I sus- James on the emotions, Merleau-Ponty on em- pect the reason for resistance is less interesting bodiment, and Wittgenstein on aspect percep- and more sociological. Psychology is a young tion. Empirical observations have often guided profession, which grew out of philosophy and philosophical inquiry. Locke was inspired by cor- physiology but then acquired its own institu- puscular physics, Marx took solace in Dar- tional standing in the academy, and it has had winian biology, and Carnap incorporated ideas to fiercely guard its status as a science by dis- from behaviorism. tancing itself from the humanities. Meanwhile, The term “empirical” is used in different philosophy underwent an analytic turn, which ways. In its broadest application, it refers to ob- led to a preoccupation with conceptual analysis, servational methods. Observation can include and an anxiety about psychologism. On this vis- an examination of the world, both inner and ion, the field began to model itself on logic or outer, with and without special instruments. mathematics, which were, in turn, taken to be a Even introspection can be regarded as a form of priori domains. I think this is a fundamental observation, as the etymology of the term sug- mistake. In many domains, the concepts that gests, and in this sense, the introspection of in- matter most are grounded in human usage, not tuitions is an observational method. Philosoph- in a transcendental realm like (allegedly) math- ers who use intuitions in theory-construction ematics. The arbiters of conceptual truth in- can be characterised as doing something empir- clude both the inferences we are inclined to ical in this broad sense. Linguistics has used draw and our linguistic behavior, both of which such intuitions to construct syntactic theories, can be studied empirically. I will not argue dir- and few would deny that syntax is an empirical ectly against a priori approaches, but will in- field. But the term “empirical” is also used stead make an empirical case, or better yet an more narrowly to refer to the use of scientific invitation, by attempting to illustrate how em- methods, which involve the design of repeatable pirical findings make contact with traditional observation procedures, and the quantification philosophical questions in metaethics. and mathematical analysis of data. The empir- One manifestation of the empirical turn ical turn in philosophy has been marked by a has been the rise of experimental philosophy. dramatic increase in the use of scientific results. This term most often refers to the work of Many philosophers have long held a posit- philosophers who conduct studies that probe ive attitude toward science, but the frequent use people’s intuitions about philosophical thought of scientific results (outside of the philosophies experiments. Strictly speaking, much of this of science) is a recent phenomenon. It became work is not experimental, since the term “exper- popular in naturalized epistemology, which iment” is often reserved in psychological re-

Prinz, J. (2015). Naturalizing Metaethics. In T. Metzinger & J. M. Windt (Eds). Open MIND: 30(T). Frankfurt am Main: MIND Group. doi: 10.15502/9783958570511 3 | 27 www.open-mind.net search for studies in which researchers attempt tions. As I use the term, it refers to the use of to manipulate the mental states of their parti- scientific results, whether obtained by a philo- cipants—experimental conditions are compared sopher or not, to address philosophical ques- against control conditions. Experimental philo- tions. The empirical turn should not be dis- sophy often explores standing intuitions, rather missed as philosophy-through-opinion-polls; it is than the factors that influence those intuitions a multi-pronged effort advance philosophical de- (e.g., Mikhail 2002). For example, some trolley bates by drawing upon observational methods studies simply poll opinions about the permiss- of any kind. ability of certain actions. That is a survey The motivations for the empirical turn are rather than an experiment. One can use survey varied, but the most general impetus is the be- methods to conduct experiments, however. For lief that some questions cannot be resolved by example, one could conduct a trolley study in more traditional philosophical methods. For ex- which some vignettes use evocative language in ample, philosophers interested in the physical an effort to manipulate participants’ emotions. basis of consciousness cannot rely on introspec- Few experimental philosophy studies do any- tion or on an analysis of the concept “con- thing like this. Most ask for opinions without scious.” And even those interested in analysis of manipulating psychological states. Thus, experi- concepts have worried about the limits of intro- mental philosophy characteristically examines spection. There are basically three different the- the content of people’s concepts and beliefs, ories of what concepts are: Platonic entities, rather than the underlying psychological pro- emergent features of linguistic practice, or psy- cesses. In this sense, experimental philosophy is chological states. None of these can be com- an extension of conceptual analysis. For those pletely investigated by introspection. Even psy- interestested in underlying processes, it can, to chological states can be difficult to introspect, this extent, be of limited interest. Some experi- because much mental activity is unconscious, mental philosophy has also been criticized for and because introspection may be prone to er- failing to meet standards of reliable behavioral ror and bias. Moreover, even if a philosopher reseach (Woolfolk 2013). That said, conceptual could perfectly introspect her own concepts, she questions are often important for philosophical would not know thereby that others shared the theorizing, and methodological problems with same concepts, and this would greatly limit the experimental philosophy can be addressed by scope of her theories. Some experimental philo- conducting better and better experiments. Of- sophers have argued that philosophers’ intu- ten the first efforts (including much of the work itions are not shared by laypeople. When philo- I will describe below) are best regarded as sophers and laypeople do agree on intuitions, analagous to pilot studies, in need of refinement there is still no guarantee that these accurately but successful enough to warrant more careful reflect reality. For example, most people (at investigation. In addition, many philosophers least in the West) find it intuitively plausible draw on (and increasingly conduct) studies that that human action derives from character traits, qualify as genuine experiments and meet the but some empirical philosophers (most notably standards of good social science. There is a long Owen Flanagan, John Doris, and Gilbert Har- tradition of philosophers using research pub- man) have drawn attention to psychological re- lished in social science journals to defend philo- search that challenges this assumption. sophical positions. For those who find paradigm Traditional and empirical approaches to cases of experimental philosophy too limiting philosophy are sometimes placed in opposition, (because they are based on conceptual intu- but they can also be regarded as interdependent. itions or fail to meet certain standards), there On one division of labor, traditional methods are are many other empirical results that can used to pose questions and to devise theories that provide illumination. The term “empirical philo- might answer those questions. Empirical methods sophy” can be used as a broader label to cover can then be used to test these theories. This is an both opinion polls and experimental manipula- over-simplification, of course, because observa-

Prinz, J. (2015). Naturalizing Metaethics. In T. Metzinger & J. M. Windt (Eds). Open MIND: 30(T). Frankfurt am Main: MIND Group. doi: 10.15502/9783958570511 4 | 27 www.open-mind.net

Figure 1: A Metaethics decision-tree tions can inspire theories, and traditional meth- we make moral judgements? Put differently, ods can sometimes refute theories (Gettier cases metaethics concerns the truthmakers of moral are a parade example in epistemology), but the judgements: what kinds of facts, if any, make proposed division of labor is a decent approxima- moral judgements true? That is a metaphysical tion. Traditional methods have limited testing question but it is normally approached se- power because theoretical posits are often difficult mantically by exploring what we are semantic- to directly observe, and empirical methods have ally committed to when we make moral judge- limited power in constructing theories, because ments. Metaethics differs from first-order ethics, theories outstrip evidence. In what follows I will which concerns the content, derivation, and ap- test theories derived from philosophical reflection plication of such judgements. against the tribunal of empirical evidence. There are many different metaethical the- ories, and a complete survey here would be im- 1.2 A roadmap to metaethics possible. I will focus on major theories that have emerged over the last two hundred and Let us turn now to the focus of discussion: fifty years, with emphasis on proposals that metaethics. Metaethical questions concern the dominated discussion in the twentieth century. nature of the moral domain. Metaethicists ask: To be clear from the outset, my goal is not to what kinds of things are we talking about when consider specific theories that have been ad-

Prinz, J. (2015). Naturalizing Metaethics. In T. Metzinger & J. M. Windt (Eds). Open MIND: 30(T). Frankfurt am Main: MIND Group. doi: 10.15502/9783958570511 5 | 27 www.open-mind.net vanced by currently active authors in The first question in the metaethics de- metaethics. Rather, I will survey broader classes cision tree is, I note, a question about emotions. of theories that have been around for some time More precisely, we can ask: are moral judgments (decades or centuries) in an effort to establish affect-laden? The term “affect” is used instead the relevance of empirical work. An adequate of “emotion” here, because it is broader. I in- examination of any recent theory would require tend the term to cover any conative state, such a narrower focus than I am after here, since as a preference, desire, or pro-attitude. For most each theory makes empirical commitments, if at of this discussion, I will focus on emotions all, in different places. rather than these other affective constructs, be- To facilitate discussion, I will map out cause emotions are implicated in the empirical the theories of interest using a decision tree work I will be considering. (Figure 1). The tree could easily be arranged The other key term in question 1 is “moral differently. Almost any branch could be the judgments.” By “moral judgments” I mean starting place, with other nodes occurring atomic judgments using thin moral concepts, higher or lower than they are in this rendi- such as wrong, bad, or immoral. The judgment tion. As we will see in a moment, I begin with expressed by “Shoplifting is wrong” would be an a question about “affect” or emotions. This example. There are many other judgments that may seem odd to some contemporary arise in moral contexts, including judgments metaethicists. Some contemporary metaethi- containing thick concepts, such as cruel or un- cists discuss emotions (such as Alan Gibbard just. One can also ask whether these are affect- 1990, and Simon Blackburn 1998), but others laden. On one analysis, thick concepts are hy- do not (for example, emotions are discussed brids that have both a descriptive and an evalu- less among moral realists). Historically, how- ative component, the latter of which may im- ever, emotions have been a central focus in plicate the emotions. For the sake of simplicitly metaethics. British moralists, who advanced I will ignore that debate here. many of the questions that continue to drive Notice that judgements are not sentences the subfield, often begin their analyses with a but rather the thoughts that sentences ex- discussion of moral sentiments. Indeed, the press. To propose that such thoughts are af- most famous controversy in metaethics before fect-laden is to say that each token instance the twentieth century is probably the dispute involves an emotion or other conative state. between British sentimentalism and Kantian There are different forms of “involvement” rationalism. Even in the twentieth century, that have been discussed in metaethics. On some of the most discussed debates concern some theories, moral judgments contain conat- emotions, such as the debate between emotiv- ive states as constituent parts. This was the ists and their opponents. Moreover, the recent view of Francis Hutcheson, David Hume, and empirical turn was triggered, in part, by dis- some other British moralists. One might coveries linking emotions to moral judgement. weaken this by saying that moral judgments So this starting point has considerable histor- do not contain emotions, but refer to them. In ical depth and great relevance to the method- this vein, John McDowell, David Wiggins, and ological sea-change that I am interested in Alan Gibbard suggest that moral judgments here. That said, I don’t intend this tree to be reflect the conviction or norm that it would be anything like a complete map of meta-ethics. warranted to feel certain emotions. Both of One could begin elsewhere and branch out in these approaches have gone under the heading further directions (I expand the tree leftward, “sentimentalism,” with the prefix “neo-” for but interesting questions also come up on the the views that say the link between moral right). Though incomplete, the nodes of this judgments and emotions is second-order. Here tree encompass much of what one might cover is a definition: in an introduction to metaethics in the Anglo- phone analytic tradition.

Prinz, J. (2015). Naturalizing Metaethics. In T. Metzinger & J. M. Windt (Eds). Open MIND: 30(T). Frankfurt am Main: MIND Group. doi: 10.15502/9783958570511 6 | 27 www.open-mind.net

Sentimentalism =Df Moral judgments es- apt if it is the kind of thing that can be evalu- sentially involve affective states, such as ated as true or false. Some affect-laden judg- emotions, in one of two ways: such states ments may turn out to have a merely expressive are constituent parts of moral judgments function. If I say, “Disco sucks!” I may not be (traditional sentimentalism); or moral attempting to represent a fact, but merely ex- judgments are judgments about the appro- pressing how I feel. Expressivists follow this priateness of such states (neo-sentimental- analogy: ism).

Expressivism =Df Moral assertions express Those who deny that moral judgments are af- mere feelings or non-assertoric attitudes, fect-laden fall into different categories, but two and do not purport to convey facts. of the most important metaethical theories of this kind are externalist moral realism and Charles Stevenson and A. J. Ayer are credited (some forms of) rationalism. Moral realists say with devising the emotivist theory of morality, that there are moral facts, which is to say that which is the simplest theory of this kind. A some states of affairs are truly right or wrong more sophisticated variant has been developed (cf. Sayre-McCord 1988). Externalist moral by Simon Blackburn, who proposes that moral realists add a further requirement, namely judgments aspire for quasi-truth, but not truth, mind-independence: and thus an ontologically neutral stand-in— which can explain why moral judgments have

Externalism moral realism =Df There are an assertoric form. Alan Gibbard says that moral facts and these obtain independ- moral judgments do not directly express feel- ently of our recognition of them. ings, as emotivists claim, but rather express the acceptance of norms according to which feelings If moral facts are mind-independent, it also fol- such as anger and guilt would be appropriate. lows that we can come to know them without All these theories have been broadly classified being moved by them. Like scientific facts, we as expressivist. can know that they obtain without feeling any Those who say that moral judgments are way towards them. Cornell realists, some intu- affect-laden and truth-apt need not deny that itionists, and many divine command theorists moral judgments are expressive, but they insist fall into this category. Moral rationalism is a that they more than express feelings; they as- view about how the epistemology or normative sert facts. If so, moral judgments can be true or status of moral truths: false. Subjectivism falls into this camp:

Moral rationalism =Df Moral truths can be Subjectivism =Df the truth of the judg- discovered and justified though a purely ment that something is morally good or rational decision procedure. bad depends on the feelings or other sub- jective states of someone who makes that Kant (1797) is traditionally read this way, judgment. though he also claimed that moral judgments involve moral feelings. For instance, one might propose that “killing is The remainder of my metaethics decision wrong” means “I disapprove of killing.” That tree concerns those who think that moral judg- judgment is true, if the speaker disapproves of ments are affect-laden. Among those who say killing, and false otherwise. As we will see, there that moral judgments essentially involve conat- are also more sophisticated forms of subjectiv- ive states, there is a divide between those who ism. Subjectivists are internalist moral realists: think that moral judgments are nevertheless they believe in moral facts, but they deny that truth-apt and those who deny this. This is the those facts obtain independently of our atti- second division of the tree. A judgment is truth- tudes.

Prinz, J. (2015). Naturalizing Metaethics. In T. Metzinger & J. M. Windt (Eds). Open MIND: 30(T). Frankfurt am Main: MIND Group. doi: 10.15502/9783958570511 7 | 27 www.open-mind.net

The term “cognitivism” has been used to play of the understanding, which results in aes- refer to any view on which moral judgments are thetic pleasure. A work may have purposeful truth-apt, which is to say they can be assessed purposelessness without our recognizing that for truth. Expressivists are non-cognitivists, and this is so, but when we recognize it, we feel a both subjectivists and external moral realists certain way. Within ethics, Francis Hutcheson are cognitivists. One could also have a cognitiv- may have held a view that was objectivist and ist theory and nevertheless insist that all moral subjectivist in just this way. He suggests that judgments are false. This would be an error the- moral facts are established by divine command, ory. but God has furnished us with a moral sense, and that sense works by means of the emotions;

Error theory =Df Moral judgments are when we see objectively bad actions, we feel dis- truth-apt, but they are never true. approbation. This has been called a moral sense theory, because it treats our moral passions as a The most famous error theory comes from J. L. kind of sensory capacity that picks up on real Mackie. Mackie argues that moral judgments moral facts. are incoherent. On the one hand, they presup- In contrast to this view, one might argue pose that moral facts are objective, which is to that moral facts are not objective, as Mackie say mind-independent. On the other hand, has maintained, but rather are dependent on moral judgments presuppose that moral facts our responses. This need not imply that moral are action guiding, and that suggests that they judgments are mere expressions of feeling; one directly motivate us. This suggests that moral might say instead that moral judgments refer to judgments must be affect-laden, or otherwise response-dependent properties. The idea of re- dependent on our subjective states. Since noth- sponse-dependent properties derives from John ing can be both objective and subjective, moral Locke’s notion of secondary qualities. Primary judgments can never be true. Opponents of the qualities, such as shape, for Locke, exist inde- error theory deny this and insist that some pendently of being perceived, whereas second- moral judgments are true. They are, in this ary qualities consist in the power that certain sense, moral realists. Moral realists who also things have to cause responses in us. Colors, for claim that moral judgments are affect-laden Locke, are not out there in the world, but con- must take Mackie’s challenge head on, showing sist in the fact that objects cause certain sensa- that truth is compatible with being action-guid- tions in us. The moral analogue of this view has ing. been called the sensibility theory, and its adher- Such sentimentalist realists face an imme- ents include John McDowell, David Wiggins, diate question. They can accept Mackie’s claim and David McNaughton. They resist the causal that moral judgments represent objective prop- language found in Locke’s theory of colors, but erties, and find some way to circumvent the in- say something close: coherence, or they could say that moral judg- ments refer to properties that are subjective, or Sensibility theory =Df moral properties are response-dependent. The first option might those that warrant moral emotions. seem untenable, since it accepts that moral judgments are both objective and subjective, an Strictly speaking, the sensibility theory is a apparent contradiction. But the contradiction form of subjectivism, since it says that moral can be mitigated by distinguishing between judgments refer to subjective properties (the sense and reference. One might say that moral property of warranting moral emotions), but the concepts have affect-laden senses—that is, we notion of warrant allows these theorists to avoid grasp them by means of feeling—and objective a pitfall or simple subjectivism. For a simple referents. Consider, for example, Kant’s aesthet- subjectivist it makes no sense to wonder ics, according to which beauty consists in a kind whether something that I disapprove of is really of purposeful purposelessness that causes a free- wrong, but for the sensibility theories I can en-

Prinz, J. (2015). Naturalizing Metaethics. In T. Metzinger & J. M. Windt (Eds). Open MIND: 30(T). Frankfurt am Main: MIND Group. doi: 10.15502/9783958570511 8 | 27 www.open-mind.net tertain such doubt because I can wonder truth-values if they are made by different whether an event really warrants what I happen observers. to feel. The notion of warrant here is not un- problematic, and it often goes unanalyzed. I will now try to show that each question on the There is one notable exception, however, and decision tree can be empirically illuminated. that is the ideal observer theory (Firth 1952; Some of the empirical results that I will present Brandt 1959): come from unpublished, exploratory studies. My goal here is not a detailed documentation of sci-

Ideal observer theory =Df The morally entific findings, but rather to establish, by good or bad is that which an observer means of example, ways in which empirical would regard as good or bad under ideal methods might be brought to bear on the fore- circumstances. going questions. The hope is that the studies described here might be taken up by others and Such circumstances might involve acquiring the improved upon. status of a moral sage (or consulting a moral sage), as on some virtue theoretic theories, or 2 Empirical resolutions to metaethical an ideal version of myself (Smith 1994). Ideal debates observer theorists are committed to response- dependence; they say that responses determine 2.1 Sentimentalism vs. rationalism and moral truth, and they further require that those externalism responses come from certain kinds of epistemic agents. Let’s begin with the first question on the Ideal observer theories offer a negative an- metaethics decision tree: are moral judgments swer to the final question in the metaethics de- affect-laden? No question in ethics has received cision tree. They specify conditions of ideal ob- more empirical attention than this. Dozens of servation in order to find an authoritative set of studies have attempted to determine whether responses among a diversity of moral opinions. emotions play a central role in morality, and the The hope is that one set of judgments will evidence has consistently shown that they do. emerge as epistemically superior to all others; Let me begin with an unpublished study of my on this view, all moral judges converge under own and then offer a brief review of the empir- ideal conditions. Here, moral truths work out to ical literature. be universal. This, of course, is a controversial To begin with, let’s consider folk intu- claim. Suppose we define ideal observers as itions. Do ordinary people use emotions as evid- those who are free from bias, aware of pertinent ence when attributing moral judgments? To test non-moral facts, and reasoning carefully. It this, I conducted a simple vignette study, which could turn out that, two such observers could pitted emotions against verbal testimony. A still disagree on moral matters. This prognosis group of college undergraduates taking an intro- leads toward the view that there is no way to ductory-level philosophy class responded to the arrive at moral consensus. Those who think following probe: that moral judgments are rendered true by a Fred belongs to a fraternity and his brothers judge’s response but deny consensus under op- in the fraternity sometimes smoke marijuana. timal epistemic conditions end up saying that Fred insists that he thinks it’s morally acceptable moral judgments are relative. This view can be to smoke marijuana. He says, “You guys are not stated as follows: doing anything wrong when you smoke.” But Fred also feels disgusted with his frat brothers when he

Metaethical relativism =Df Two judgments sees them smoking. One day, to prove that he expressed using tokens of the same word thinks smoking is okay, he smokes marijuana him- types, and grasped by tokens of the same self. Afterwards, he feels incredibly ashamed mental attitude types can have different about smoking the drug.

Prinz, J. (2015). Naturalizing Metaethics. In T. Metzinger & J. M. Windt (Eds). Open MIND: 30(T). Frankfurt am Main: MIND Group. doi: 10.15502/9783958570511 9 | 27 www.open-mind.net

Which of the following seems more likely: limitations we must move beyond experimental philosophy, and look for more direct evidence 1. Fred says he morally approves of marijuana that emotions actually are sufficient and neces- smoking, but in reality he thinks it is morally sary for moral judgments. But the study is still wrong. revealing, because it shows that emotions are 2. Fred feels badly about smoking marijuana, used as evidence in moral attribution. Most par- but in reality he thinks it is morally accept- ticipants make attributions that fall in line with able. sentimentalism. To show that emotions actually do con- In my small sample, 68.4% chose answer 1, sug- tribute to moral cognition, we can look at three gesting that the majority of them take emotions kinds of evidence: cognitive neuroscience, beha- as evidence for moral values, even when that vioral psychology, and pathology. In each do- directly contradicts self-report. This suggests main, sentimentalism finds support. There have that many people take emotions to be a suffi- now been dozens of neuroimaging studies on cient evidence for attribution moral attitudes. moral judgment tasks, and every one of them, An even more dramatic result was obtained to my knowledge, shows an increase in activa- when another twenty participants assessed this tion in brain structures associated with emo- scenario: tion, when moral decisions are compared to Frank belongs to a fraternity and his non-moral decisions. Key stuctures include the brothers in the fraternity sometimes smoke posterior cingulate, temporal pole, orbitofrontal marijuana. Frank insists that their actions are cortex, and ventromedial prefrontal cortex. morally unacceptable. He says, “You guys are There are only two groups of studies that even doing something wrong when you smoke.” But appear to depart from this pattern. Joshua Frank does not feel any anger or disgust when Greene et al. (2001) report that emotions play he sees his frat brothers smoking. One day, more of a role in deontological judgments than when they are not around, he smokes marijuana in consequentialist judgments, but their data himself. Afterwards, he doesn’t feel any shame show that, as compared to non-moral judg- about smoking the drug. ments, emotions are involved in both (see their Which of the following seems more likely: Figure 1). Moreover, Greene et al. use moral di- lemmas in which the common denominator is 1. Frank says he morally opposes marijuana saving lives—they manipulate the nature of the smoking, but in reality he thinks it is morally harm necessary in order to save five people in acceptable. danger. Thus, each moral judgment condition 2. Frank doesn’t feel badly about smoking presumably elicits the judgment that it would marijuana, but in reality he thinks it is mor- be good to help people in need. This positive ally wrong. moral judgment may be emotionally grounded, but the neuroimaging method subtracts away Here, 89.5% of participants chose response 1, in- this emotional information, because it is present dicating that they take emotions to be neces- in each scenario, and imaging results of this sary for the attribution of moral attitudes. Ab- kind report only contrasts between different sent the right feelings, verbal testimony is conditions. Thus, a major dimension of moral treated as an unreliable indicator of a person’s emotions may be systematically concealed by values. the method. The other study that fails to show This study has at least four serious limita- an increase in emotional responses during moral tions: people may not trust self-reports; the res- judgment is one condition in a series of imagin- ults were far from unanimous; it fails to distin- ing experiments performed by Jana Borg et al. guish evidence for moral attitudes and essence (2006). But, in that condition, a moral scenario of moral attitudes; and folk beliefs about moral is compared to a scenario about an encroaching judgments may be wrong. To get around these fire that threatens one’s property, and it is un-

Prinz, J. (2015). Naturalizing Metaethics. In T. Metzinger & J. M. Windt (Eds). Open MIND: 30(T). Frankfurt am Main: MIND Group. doi: 10.15502/9783958570511 10 | 27 www.open-mind.net surprising that moral judgments produce less of than when judging others. When another per- an emotional response than a case of personal son commits a crime against nature, we tend to loss. feel disgust, but when we perform an act Brain science resoundingly links moral deemed by others to be unnatural, the most judgment to emotion, but the method is correl- common response seems to be shame. Con- ational. Moral rationalists and externalists versely, when others commit crimes against per- could concede that moral judgments excite emo- sons, we feel angry, but guilt is the natural re- tional responses, while denying that these are sponse when we perform such acts ourselves. To the basis of moral judgment. Imagine the fol- test this hypothesis, I conducted a forced-choice lowing view: we use reason to arrive at moral study in which a group of college undergradu- judgments, but morality matters to us, so when ates had to pick guilt or shame in response to we arrive at those judgments emotions normally mildly “unnatural” acts (“Suppose your room- kick in. By analogy, reason might be used to de- mate catches you masturbating”), and mildly termine that certain life activities (smoking, harmful acts (“Suppose you take something high fat diets, sleep deprivation) are harmful, from someone and never return it”). 80% chose and, upon drawing that reason-based conclu- shame for the first case, and over 90% picked sion, we tend to experience corresponding emo- guilt for the second. tions, such as anxiety when contemplating light- Such findings demonstrate that different ing a cigarette. Neuroimaging results showing emotions play different roles. I mentioned responses to cigarettes might confirm this, three distinctions that are currently receiving showing emotion areas active when cigarettes empirical attention: the split between positive are seen, but that wouldn’t refute a rationalist and negative emotions (praise and blame), theory of how we arrive at the judgment that between two kinds of blame (crimes against cigarettes are dangerous. nature and crimes against persons), and To adjudicate between the thesis that between self- and other-directed blame. The emotions are constitutive of moral judgments, self/other distinction may be particularly im- as opposed to mere consequences, we need be- portant because it helps us see how moral havioral evidence. Numerous studies now estab- emotions differ from their non-moral ana- lish a causal link between emotion and moral logues. Anger (or at least irritation) and dis- judgment. When emotions are induced, they in- gust can both occur in non-moral contexts, fluence how good or bad things seem. Induction but they take on a moral cast, I submit, when methods have been widely varied: hypnosis, paired with dispositions to feel guilt and dirt, film clips, autobiographical recall, and shame, respectively. If I find eating insects smells. In one recent study, Kendal Eskine, Nat- physically revolting, I will experience disgust alie Kacinik, and I induced bitterness by giving when I see others eat insects, and disgust people a bitter beverage and found that moral when I inadvertently eat them myself. But if I judgments became more severe (Eskine et al. found insect eating immoral, it would not be 2011). In other recent studies Angelika Seidel disgust that I experience in the first-person and I use sound clips to induce specific emo- case, but shame. This feeling of shame would tions, and we have shown that different emo- motivate me to make amends for my actions, tions have different moral effects: anger induces or to conceal my wrongdoing from others, not more stringent wrongness judgments about simply to repel the unwanted food from my crimes against persons; disgust induces greater body. The self-directed emotions round out stringency on crimes against nature (such as the punitive cast of our moral attitudes. We cannibalism); and happiness induces stronger see morally bad acts as not just worth ag- judgments that it is both good and compulsory gressing against, but as worthy of apology. to help the needy (Seidel & Prinz 2013a, This need not be a second-order belief. Rather 2013b). There is also evidence that we feel dif- it is implicit in the fact that moralized behavi- ferent emotions when judging our own actions ors carry emotional dispositions toward self

Prinz, J. (2015). Naturalizing Metaethics. In T. Metzinger & J. M. Windt (Eds). Open MIND: 30(T). Frankfurt am Main: MIND Group. doi: 10.15502/9783958570511 11 | 27 www.open-mind.net and other that together promote a punitive tions completely. More powerful evidence comes attitude: a disposition to issue and submit to from the clinical populations who suffer from punishment. emotional deficits. For example, psychopaths, Putting this together, I propose that who suffer from deficit in guilt and other negat- standing moral values (the values that a given ive emotions, notoriously fail to appreciate what individual has for an extended period of time) is wrong with their actions (Hare 1993). Simil- consist in dispositions to feel the self- and arly, people with Huntington’s disease, which other-directed emotions that I have been dis- impairs disgust, show high incidence of cussing. Such an emotional disposition can be paraphelias, suggesting that they cease to see called a sentiment. On any given occasion when deviant sexual behavior as wrong (Schmidt & a standing value becomes active in thought— Bonelli 2008). Kramer (1993, p. 278) argues i.e., when a moral judgment is made—these dis- that anti-depressants can flatten affect in a way positions result, all else being equal, in an emo- that results in a “loss of moral sensibility.” tional state. The emotion that is felt depends There is also a positive relationship between on who is doing what to whom. For example, if alexithymia and Machiavellianism, suggesting I recall a situation in which I hurt someone’s that a reduction in emotional competence may feelings, I will have a feeling of guilt regarding act in ways that are more instrumental than that event, because a person was harmed and I moral (Wastell & Booth 2003). For better or was the culprit. This feeling of guilt toward an worse, there is no clinical condition in which all event constitutes my judgment that the action emotions are absent and behavioral function re- was wrong, and I gain introspective access to mains, but these findings suggest that selective this judgment by feeling guilt well up inside me. or global dampening of the emotions leads to If this is right, then emotions are not merely ef- corresponding deficits in moral judgment. That fects of moral judgments, but essential compon- is, people with diminished emotions seem to be ents of them. insensitive to corresponding parts of the moral Against this picture, one might object that domain, suggesting that they may not be form- emotions are merely a heuristic that can be ing moral judgments. used in certain circumstances, but not strictly The evidence summarized here suggests necessary for making moral judgment. Following that emotions arise when we make moral judg- the analogy mentioned before, anxiety might be ments, that emotions are consulted when re- used as a heuristic when deciding whether to porting such judgments, and that moral judg- smoke, but the judgment that smoking is dan- ments are impaired when emotions are unavail- gerous does not depend on fear, and was ini- able. Some of this evidence is preliminary, but, tially arrived at by the light of reason. for present purposes, let’s assume that the find- To establish that emotions are not merely ings hold up to further and more stringent test- helpful heuristics, one must see what happens ing. By inference to the best explanation, such when emotions are reduced or eliminated. To findings suggest that emotions are components look into this, Eskine (2011) gave people the of moral judgments. The idea is that, when bitter taste manipulation and then warned people say something is morally bad, the them not to let the feelings caused by that thought they are expressing on that occasion beverage interfere with the moral judgments. In consists of a negative emotion directed towards this condition, he found that moral judgments the thing judged bad. Emotions, on this view, were considerably less severe than a control con- function like predicates in thought. That is dition, suggesting that, when we ignore emo- what traditionally sentimentalists, such as tions, it is harder to see things as wrong. The Hume, seem to have maintained. Hume thought finding indicates, in other words, that moral ideas—the components of thoughts—were judgments subside when emotions are absent. stored copies of impressions, and the idea of The study cannot confirm this strong claim, moral badness consisted in a stored copy of the however, because people cannot suppress emo- impression of disapprobation.

Prinz, J. (2015). Naturalizing Metaethics. In T. Metzinger & J. M. Windt (Eds). Open MIND: 30(T). Frankfurt am Main: MIND Group. doi: 10.15502/9783958570511 12 | 27 www.open-mind.net

Traditional sentimentalism, which says ing it as morally bad. Reductive theories of that emotions (or sentiments) are actually com- value leave both questions open. If I form the ponents of moral judgments, differs conspicu- attitude that killing cannot be willed as a uni- ously from neo-sentimentalism. Neo-sentiment- versal law, I can still wonder both whether alists theories say that moral judgments are killing is bad and whether I am judging that it judgments about the appropriateness of emo- is bad. Sentimentalism leaves the first question tions. These theories do not straightforwardly open, but not the second. When experiencing predict that emotions come on line when we outrage at killing, it seems impossible to wonder make moral judgments, nor that a reduction in I am regarding killing as bad. I can of course emotions should interfere with our ability to wonder whether killing really is as bad as it moralize. Instead, they predict that people will seems. Such doubts can arise because I may not think about emotions when they make moral know the true source of the emotion I am feel- judgments. Correlatively, they also predict that ing. Perhaps my outrage comes from some ex- people with limited metacognitive abilities will traneous source (such as a bitter beverage), for lose their ability to make moral judgments; this example. But this open question does not is not the case (Nichols 2008). Thus, given the threaten the thesis that moral judgments are current state of evidence, traditional sentiment- constituted by sentiments. The only open ques- alism outperforms neo-sentimentalism empiric- tion that poses such a threat would be one ally. Traditional sentimentalism predicts a ro- about what my attitude is, not one about bust pattern of empirical findings. whether my attitude is true. The fact that some Rationalists and externalist moral realists sentiments are experienced as condemnatory ef- might baulk at this point and say that the em- fectively closes the question about whether pirical evidence lacks the adequate modal someone experiencing those sentiments is adopt- strength to support sentimentalism. The evid- ing a moral stance. By analogy, imagine tasting ence shows that emotions are often consulted a wine and wondering whether it really is deli- when making moral judgments, but this leaves cious, while experiencing gustatory pleasure. We open the possibility that we might also make can have this thought (a thought about truth), moral judgments dispassionately under circum- because we can’t be sure where the pleasure stances that have not yet been empirically ex- came from (was it the wine or the company?). plored. So stated, this objection is just an ex- But we can’t experience gustatory pleasure and pression of faith. It suffers from both conceptual wonder whether we are, at that moment, find- and empirical weaknesses. Conceptually, oppon- ing the experience delicious. Thus, gustatory ents of sentimentalism must say what moral pleasure is plausible a component of delicious- judgments are, such that they can be had dis- ness judgments. passionately. What thought is a dispassionate The foregoing may look like a conceptual person conveying, when she says, “Killing the argument for sentimentalism. But it can also be innocent is morally bad?” Any attempt to give construed as an empirical claim. The argument a reductive answer will be vulnerable to open- hangs on the premise that people experiencing question worries. No descriptive substitute for outrage take themselves to be making moral the phrase “morally bad” leaves us with a sen- judgments. This can be empirically tested. In- tence that is conceptually synonymous with the deed, all the evidence about people consulting third. their emotions when making moral judgments Arguably, the open-question argument stands as evidential support. Merely making does not threaten sentimentalism. Let’s distin- someone mad results in more negative moral at- guish two kinds of open questions. First, given a titudes. This can be interpreted as showing certain attitude towards killing, one can still that, when people are angry, there is no ques- wonder whether killing really is morally bad. tion for them about whether they are holding Second, given a certain attitude toward killing, something in negative moral regard. Conversely, one can wonder whether one is thereby regard- it would be easy to show that people do not ne-

Prinz, J. (2015). Naturalizing Metaethics. In T. Metzinger & J. M. Windt (Eds). Open MIND: 30(T). Frankfurt am Main: MIND Group. doi: 10.15502/9783958570511 13 | 27 www.open-mind.net cessarily draw this inference when they form the the theories of their opponents. At the moment, judgment that something cannot be willed as a there is no empirical evidence for such exotica. universal law. Opponents of sentimentalism owe More modestly, the empirically-minded us a positive account of evaluative thoughts sentimentalist might welcome an attempt to that avoids open-question worries as success- find evidence for opposing views. Little effort fully as sentimentalist accounts. has been put into this task, though empirical Opponents of sentimentalism might try to claims for emotion-free moralizing are occasion- bypass this demand by offering a non-reductive ally advanced. The most publicized example is account of moral judgments, treating thin moral Koenigs et al.’s (2007) study, which shows intact concepts as primitives. That possibility, which consequentialist judgments in patients who suf- was attractive to Moore, looks unmotivated fer from ventromedial prefrontal brain injuries, given the empirical evidence for an emotional which are thought to impair emotion. But this foundation. Every study suggests that emotions description is misleading. As the authors note, arise when we make moral judgments. All evid- ventromedial patients are highly emotional, and ence also suggests that when emotions are elim- their most notorious symptom is that they are inated, judgments subside as well. This does not insensitive to costs when seeking rewards. Pre- prove that we can make moral judgments sumably, reward- seeking is an affectively without emotions, but, by induction, it provides grounded behavior. The fact that these patients evidence. Some have argued that extant evid- make normal consequentialist judgments does ence is ambiguous about whether emotions are not entail that they rely on reason alone, but essental components of moral judgments or rather on their positive emotions. Since these mere accompaniments, but I have suggested emotions cannot be easily regulated by negative here that the former may provide a better ex- feedback in ventromedial patients, they tend to planation (and certainly better predictions) of be more consequentialist than healthy popula- the total pattern of data (Huebner et al. 2009; tions—that is, they are more willing to push a Waldmann et al. 2012). Until opponents of sen- heavy man in a trolley’s path in order to save timentalism can identify some clear cases of five. moral judgments without emotions, they will be Will better empirical evidence for rational- on the losing side of the debate. At the mo- ism or externalist moral realism be forthcom- ment, there is no empirical evidence that this ing? I doubt it. Rationalists hold that we can ever happens. arrive at moral judgments through reasoning. Notice too, that it would be relatively un- Unlike some sentimentalists, I think reasoning is interesting to show that, under as-yet-unidenti- important to morality. It is likely that we use fied and highly unusual conditions, people can reasoning to extrapolate from basic values to make what look like moral judgments in the ab- novel cases. But it is unlikely that we could use sence of emotions. The sentimentalist will reply reasoning to derive basic moral values. Philo- that the vast majority of ordinary moral judg- sophers have tried to do this for centuries with ments are emotionally based. If moral vocabu- no consensus behind any view. This might be lary is occasionally used dispassionately, senti- described as a strong empirical argument by in- mentalists can ask whether the thoughts ex- duction: thousands of smart, trained moral ex- pressed on such occasions are of the same kind perts have failed to identify a line of reasoning that we find, in study after study, in the usual that is widely recognized as providing adequate cases. Upon finding a class of dispassionate rational support for basic moral propositions. judgments, one might do best to posit an ambi- Moreover, when moral debates arise, there is guity in the category. The sentimentalist can little evidence that reasoning is efficacious on content herself with the project of providing a its own. Instead, societal transformations in val- metaethics for garden-variety moral judgments, ues seem to arrive with political upheavals, eco- while leaving open the possibility that there nomic revolutions, and generational change. At- may be psychological exotica, which conform to titudes towards slavery changed with the indus-

Prinz, J. (2015). Naturalizing Metaethics. In T. Metzinger & J. M. Windt (Eds). Open MIND: 30(T). Frankfurt am Main: MIND Group. doi: 10.15502/9783958570511 14 | 27 www.open-mind.net trial revolution, women’s suffrage came with a them. It is also open to a damaging empirical ob- world war, and increase in support for gay jection. Phenomenologically, it is true that moral rights correlates with the dissolution of tradi- intuitions often seem immediate and unbidden, tional social roles and economic transformations but this can be readily explained on a sentiment- that have made procreation more costly than alist account. Emotions are conditioned (by train- abstinence. I don’t mean to imply that there are ing or evolution) to arise automatically and often no rational arguments for these liberation move- intensely when certain actions, such as torturing ments. Rather, I am suggesting that those argu- babies, are considered. This gives an impression of ments take hold only when social conditions are immediacy without postulating any special con- right. It is noteworthy, for example, that sci- tact with moral reality. Moreover, these intuitions entific racism appeared very late in the history vary from group to group. For example, there is of slavery, suggesting that slavery was not empirical evidence that liberals and conservatives simply based on false beliefs about racial in- have divergent basic values (Graham et al. 2009). equality. In fact many societies have enslaved The presence of such foundational intuitions can their own people, and many proponents of sci- be explained demographically, and their lack of entific racism have been against slavery. Rather, convergence casts doubt on the existence of a advocacy of slavery seems to reflect a set of ba- moral faculty that reveals universal moral truths. sic moral values that changed in recent history: In other words, intuitionism is vulnerable to a de- values that say social standing can be determ- bunking argument. Social science coupled with ined by the lottery of birth. With industrializa- sentimentalism provides a good explanation of tion, models of labor based on the idea of self- deeply-held intuitions, so there is no need to sup- determination took hold, and the idea that pose that these intuitions reflect anything deeper. birth should determine social standing began to This point about moral variation, to which I wane. Of course, it hasn’t disappeared alto- will return, also counts against some forms of ex- gether, but it has been tempered by the the ternalist moral realism. Advocates of that posi- emergence of a new norm. Before industrializa- tion sometimes suggest that objective moral facts tion, the idea that human beings are born equal can be established by identifying the external and free might have seemed manifestly false, factors that best explain human moral behavior and thus it could have played no effective role or judgments. If moral behavior and judgments in any argument against slavery. With industri- vary from group to group, however, it is unlikely alization, this premise gained appeal, and be- that we will find an external common denomin- came the foundation of compelling arguments. ator underlying these practices. Such a search also Arguments are not inert, but they are only as seems unnecessary given that we already have good as the premises on which they are based, good explanations of moral behavior and judg- and the plausibility of those premises may de- ments in terms of socially-conditioned sentiments. pend on factors other than reasonging. It is pos- None of these arguments are the nail in the sible that reasons have little role in driving ba- coffin for externalist realist or rationalist theories. sic values. And if so, then the recent broadening They merely illustrate the relevance of empirical moral umbrella is not the result of a rational in- results. The findings mentioned here must be ex- ference to the conclusion that our basic values plained. It is my contention that sentimentalism cover more cases than we thought, but rather provides the best explanation of the findings I an irrational shift in basic values. have reviewed, but further arguments and evid- A realist might concede that such considera- ence could tip the balance in another direction. tions threaten rationalism, but vie instead for a kind of intuitionist perspective, according to 2.2 Cognitivism vs. non-cognitivism which basic moral truths are simply obvious. To me, this looks like a magical moral epistemology Let’s move on to the second question on the —one wonders what moral facts could be such metaethics decision tree: Are moral judgments that our moral sense could simply lock on to truth-apt? As positioned on the tree, this is a

Prinz, J. (2015). Naturalizing Metaethics. In T. Metzinger & J. M. Windt (Eds). Open MIND: 30(T). Frankfurt am Main: MIND Group. doi: 10.15502/9783958570511 15 | 27 www.open-mind.net question that arises for sentimentalists, raised that correlate with directive speech, such as those pressingly by the conclusion that moral judg- just mentioned, and he searched corpora of ments have a basis in the emotions. It is that spoken and written texts for co-variance between conclusion that seems to put truth-aptness in these elements and moral terms. He calls the dir- jeopardy, since emotions have not traditionally ective elements “suasion markers,” and the correl- been regarded as having accuracy conditions in ations between these and other linguistic items a the way regarded as allowing for truth. But, it “suasion score.” Non-cognitivism seems to predict should be noted that the question of truth-apt- a high suasion score, given the postulated direct- ness could also be raised independently of senti- ive function of moral judgments. This prediction mentalism. There are non-sentimentalist theor- fails. Not only is there no positive correlation ies that deny truth-aptness (for example, one between moral vocabulary and suasion markers, might say that moral judgments are imperative, there is actually a negative correlation, which ap- while denying that they need be passionate), proaches significance. This negative relationship and there are non-sentimentalist theories that was observed in seventeen out of nineteen differ- accept truth-aptness (the vast majority fall in ent categories of text that he examined. These this category). To keep things as neutral as pos- results are preliminary—a first foray into empir- sible, I will begin by asking whether there is ical ethics—but they provide compelling evidence any empirical evidence that moral judgments that moral discourse is not directive in nature. are non-cognitive, whether or not they are af- Non-cognitivism entails that moral discourse fect-laden. does not aim to refer to facts in the world. This The posing of this question is itself a de- carries another linguistic prediction that can be gree of philosophical progress, because non-cog- readily tested. Certainly adverbs are used to in- nitivists too rarely reflect on the predictions of dicate a focus on how things are in the world. their view. Indeed, the most obvious empirical These include “really,” “truly,” and “actually.” prediction fails resoundingly. If moral judgments These words have other uses (“really” can be a do not aim at truth, we might expect them to term of emphasis), but they often play a role in have a non-declarative syntactic. For example, emphasizing the factive nature of the modified we might expect them to take the form of im- phrase. Therefore, if non-cognitivism were true peratives or exclamations. But they do not. In one might expect these words to rarely be used as every language that I know of, moral judgments modifiers for moral terms. To test this, I used are expressed using declarative sentences, which Google search engine to search for and note the should stand as a profound embarrassment to frequency of three phrases: “really immoral,” the theory. Granted, non-cognitivists sometimes “truly immoral,” and “actually immoral.” To do propose elaborate logics to accommodate this this, I needed a baseline, and chose to compare fact, but it is surprising that they should have “immoral” to a word widely believed to designate to do so. One would expect the surface gram- a objective feature of the world. I chose “triangu- mar to reflect the non-cognitive form. lar,” a classic primary quality, on a Lockean To push things further, one might look for scheme. The results are as follows (as of March, more subtle linguistic evidence in favor of non- 2013): cognitivism. For example, some non-cognitivists assume that moral utterances have the illocution- “really triangular”: 6,500 hits ary force of directives, such as orders, requests, or “really immoral”: 10,600 hits demands. Directives often occur in speech con- “truly triangular”: 4,920 hits texts that contain words that play a role in per- “truly immoral”: 32,000 hits suasion, such as “come!”, “let’s”, or “we encour- “actually triangular”: 21,600 hits age you…” To empricially test this kind of non- “actually immoral”: 61,600 hits cognitivism, Olasov (2011) ingeniously used this technique for sociolinguistics, called corpus ana- Clearly, the adverbs that indicate a real-world lysis. He used a set of such linguistic elements focus are used more frequently for moral terms

Prinz, J. (2015). Naturalizing Metaethics. In T. Metzinger & J. M. Windt (Eds). Open MIND: 30(T). Frankfurt am Main: MIND Group. doi: 10.15502/9783958570511 16 | 27 www.open-mind.net than for terms designating objective physical against such theories, in favor of the view that features—over six times as common in the case emotions are bodily feelings (Prinz 2004), but of “truly.” I also tried the phrases “in truth,” contemporary feeling theorists still insist that “truthfully,” and “in actual fact”: emotions aim to refer. Feeling sad, for example, can be understood as a downtrodden bodily “truthfully triangular”: 6 hits state that represents loss. To say that the feel- “truthfully immoral”: 44 hits ing represents loss is to say that it has the func- “in truth triangular”: 46 hits tion of arising in response to losses, and hence “in truth immoral”: 1,350 hits carries the information that there has been a “in actual fact triangular”: 2 hits loss to a person who experiences it. In a like “in actual fact immoral”: 133 hits manner, pain may indicate tissue damage and fatigue may indicate energy depletion, even These truth-tracking phrases modify “immoral” though pain and fatigue are bodily feelings. between 7 and 166 times more frequently than None of these feelings are arbitrary. They pre- they modify “triangular.” Moreover, these differ- pare an organism to cope with specific condi- entials are misleadingly small because the baser- tions or events. Emotions qua feelings are in the ate for “immoral” is far lower than “triangu- business of keeping us abreast about how we are lar” (6,910,000 hits as compared to 11,600,000). faring. Each emotion has a different signific- This was just an exploratory study, but there is ance, and any one of them can misfire. I might a simple implication. Non-cognitivism makes be sad when there is no loss, or frightened when linguistic predictions, and when those are there is no threat. Such emotions would qualify tested, they do not seem to pan out. Non-cog- as erroneous. nitivists owe us evidence, or they must deny If emotions are in the business of repres- that their theory makes predictions, in which enting, then there is no difficulty supposing that case it would cease to be falsifiable. moral judgments are truth-apt. When we sin- In response, non-cognitivists might claim cerely assert that, “Killing innocents is bad,” we that there is one crucial line of evidence in favor express a negative feeling towards killing, and of the view, and it’s a line of evidence that we that feeling functions as a kind of visceral pre- have already seen. In the previous section, I sur- dicate. It attributes a property to killing (I will veyed studies suggesting that morality is affect- have more to say about this property below). In laden. At the start of this section, I said the this sense, moral discourse may be much like non-cognitivism is orthogonal to affect-ladeness, other forms of emotional discourse. If we say but some non-cognitivists would vehemently that some food is icky, we express a feeling, disagree with this. They would say that non- while also attributing a property. For example, cognitivism follows from affect-ladeness. Emo- the feeling of ickiness might represent the prop- tions are traditionally regarded as feelings, and erty of noxiousness, or perhaps something more feelings are not traditionally believed to be rep- subjective, such as the property of causing resentations of anything. If the thought that nausea in the speaker. Someone who calls some- killing innocents is wrong is really a bad feeling thing “icky” need not know what property that about killing, then why think this thought has feeling represents, but most language users any truth conditions? Does a feeling of indiges- probably recognize that in using this term we tion or irritation really refer? are attempting to say something about This move might have been compelling in whatever it is that elicits the feeling. By ana- the early part of the twentieth century, but the logy to “icky,” moral assertions can be under- last fifty years of emotion research have em- stood as both expressive and predicative. It is a phasized the intentionality of affect. Some philo- mistake, based on overly simplistic theories of sophers have adopted cognitive theories of the emotions, to assume that feelings cannot play a emotions, according to which emotions are semantic function. Once we see that feelings can identical to judgments. Elsewhere I have argued represent properties and function as predicates,

Prinz, J. (2015). Naturalizing Metaethics. In T. Metzinger & J. M. Windt (Eds). Open MIND: 30(T). Frankfurt am Main: MIND Group. doi: 10.15502/9783958570511 17 | 27 www.open-mind.net non-cognitivism no longer looks like a serious vantage. For example, if we believe that cheat- option. ing others is objectively bad and that belief is action-guiding, then we will hold others ac- 2.3 Realism vs. the error theory countable when they cheat, and we will resist cheating even when it might seem advantageous It is one thing to say that moral assertions aim to do so. This reduces the likelihood of free to represent and quite another to say that they riders and leads to an evolutionarily stable succeed in doing so. It is possible that when we strategy—one that can foster cooperation and say that an action is immoral, we aim to ascribe collective works. Evolutionary ethicists also typ- a property to it, but we do not succeed in doing ically endorse sentimentalism, suggesting that so. This is precisely what defenders of the error moral emotions have evolved to motivate such theory have claimed. So, even if the forgoing things as punishment and altruism. Mackie him- case for cognitivism succeeds, we must now des- self is not explicit about the role of emotions in cend the decision tree and ask whether moral his view, which makes it unclear what he means judgments are ever true. when he says that we perceive the discovery of The error theory, which states that moral alleged moral facts to be action-guiding. The judgments are truth-apt but always false, was link between judgments and emotions, emphas- first promulgated by J. L. Mackie (1977). ized by evolutionists, provides one answer. Mackie’s argument begins with the premise that The evolutionary addendum to Mackie’s moral predicates aim to represent properties argument may look like an empirical reason for with two important features. The first is ob- siding with the error theory. Natural selection is jectivity: moral properties are supposed to be a well-confirmed process, emotions have some the kinds of things that can obtain independent basis in evolution, and evolutionary models con- of our beliefs, desires, inclinations, and prefer- firm that emotionally-grounded moral instincts ences. The second is action-guidingness: moral would be adaptive. But there are empirical reas- properties are supposed to be the kinds of ons for doubting the evolutionary story, and for things that compel us to act when we recognize doubting the key premises in Mackie’s argu- them. Mackie’s second premise is that these two ment. Consequently, I think the case for the er- features are difficult to reconcile. Objective ror theory fails. properties are usually the kinds of things about The evidence for an evolved moral sense is which we can be indifferent. Mackie uses the underwhelming. A thorough critique cannot be term “queer” to describe properties that are undertaken here, but let me offer two broad both objective and action-guiding, and he also reasons for doubt (for more discussion, see suggests that such queer properties would re- Prinz 2007a). First, there is little evidence for a quire an odd epistemology. For these reasons, he moral sense in closely related species. Recall thinks we shouldn’t postulate objective action- that moral judgments are underwritten by emo- guiding properties. But, Mackie thinks that tions such as anger, disgust, guilt, and shame. moral concepts commit to the existence of such There is no evidence that the last three of these properties, and, thus, that moral judgments emotions exist in chimpanzees, and the anger posit properties that don’t exist. Therefore, they exhibit might better be described as react- moral judgments are systematically false. ive aggression, because there is little reason to In recent years, the error theory has be- believe chimps form robust tendencies to be come popular among evolutionary ethicists angry about third party offences when they are (Ruse 1991; Joyce 2006). Mackie’s theory leaves not directly involved. Evolutionists point out us with a puzzle. Why do people make moral that chimps engage in reciprocal altruism, and judgments if they are incoherent? Evolutionary other forms or prosocial behavior, but these be- ethicists purport to have an answer. They say haviors may not depend on any moral judg- that morality is an illusion that has been natur- ments. Indeed, psychopaths engage in reciprocal ally selected because it confers a survival ad- altruism (Widom 1976), and chimps often be-

Prinz, J. (2015). Naturalizing Metaethics. In T. Metzinger & J. M. Windt (Eds). Open MIND: 30(T). Frankfurt am Main: MIND Group. doi: 10.15502/9783958570511 18 | 27 www.open-mind.net have in ways that seem psychopathic; they can distinguish moral and conventional rules. Some be extremely violent (Wrangham 2004) and in- subsequent authors have argued that the learn- different to each others welfare (Silk et al. ing in question involves emotional conditionism 2005). (Blair 1995; Nichols 2004). Moral rules are Evolutionary ethicists might concede this aquired through the inculcation of emotions and argue that morality evolved in the human such as anger, guilt, and shame. There are species after we split from other primates. But strong negative feelings associated with hitting this position is vulnerable to a second objection: that don’t disappear when children imagine the there is good reason to think that morality in teacher saying it is okay to hit. Violating social humans is learned. Moral judgments derive from conventions may lead to other emotions, such as emotions that originate outside the moral do- embarrassment, but these are mitigated when main, such as disgust, which is first applied to we move from one social setting to another. For noxious agents and later expended to the social example, wearing a hat at the dinner table domain, through conditioning (Prinz 2007a). might be frowned on in some circumstances, but Even guilt and shame may be learned not when wearing a birthday hat at a birthday byproducts of non-moral emotions: shame is re- party. The idea that moral rules are learned by lated to embarrassment and guilt may be a emotional conditioning could also explain their blend of sadness and anxiety brought on by vi- motivational impact; emotions impel us to act, olating a social norm (Prinz 2005). These emo- so emotionally grounded rules seem to carry tions and their range of application depend on practical demands. This analysis would explain extensive conditioning in childhood. Moral vari- both features emphasized by Mackie—action- ation across cultures is considerable, as we will guidingness and objectivity—without assuming see, and shared moral values can be attributed that moral rules actually are objective. Thus, to widespread constraints on building a stable the error theory could get off the ground society (for example, stable societies must pro- without assuming that morality is a product of hibit wanton murder within the in-group). evolution. Moreover, there is no poverty-of-the-stimulus On closer scrutiny, however, this argument argument for morality; children receive ample is not strong enough to rescue the error theory. “negative data” in the form of punishment, and It conflates objectivity with authority independ- they directly imitate values in their communit- ence. It is true that children think hitting is ies. As I argue in greater detail elsewhere, argu- wrong even when it is permitted, but that does ments for innate moral norms have been uncon- not mean they think moral truths exist inde- vincing (Prinz 2007a). This suggests that moral- pendently of subjective responses. Many of our ity is learned, not evolved. subjective responses seem independent of what If morality is acquired through learning, authorities happen to say—our preferences for then one cannot bolster Mackie’s argument by food and music, for example. But we don’t ne- assuming that morality is the product of evolu- cessarily infer that these things are objective. So tion. This alone does not undermine the error it is a further empirical question whether ob- theory, however. Error theorists might abandon jectivity is an essential feature of how we under- the evolutionary approach and try to explain stand moral properties. systematic error by appeal to a learning story. This brings us to the heart of Mackie’s ar- There is some evidence that people tend to gument. Should we grant his first premise that treat certain rules as universally binding, re- moral assertions entail objectivity? Empirically, gardless of operative conventions. When asked the answer is a bit messy. When polled, many whether it would be okay to hit a classmate if people assume that morality is objective, but the teacher granted permission, children tend to many reject this assumption (Nichols 2004; say “no.” Turiel (1983, Ch. 7) who made this Goodwin & Darley 2008). In survey studies, discovery, denies that such objectivist leanings there is a nearly even split between objectivists are innate. Rather, he thinks children learn to and their opponents. Strikingly, belief in ob-

Prinz, J. (2015). Naturalizing Metaethics. In T. Metzinger & J. M. Windt (Eds). Open MIND: 30(T). Frankfurt am Main: MIND Group. doi: 10.15502/9783958570511 19 | 27 www.open-mind.net jectivity correlates with religiosity. Goodwin Suppose scientists discover that there are and Darley report that religious beliefs were the two kinds of things that people call im- strongest predictor of objectivity that they were moral. Would it be better to say: able to find. This suggests that beleif in ob- (a) The term “immoral” is misleading, and jectivity is not an essential part of moral com- it might be better to replace it with two petence, but is, rather, an explicitly learned terms corresponding to the two kinds of add-on that most often comes with religious cases. education. The authors also found that belief in Or objectivity goes down in cases of moral issues (b) The fact that there are different cases about which there is considerable public debate, is interesting, but doesn’t affect the word. such as abortion. This might be interpreted as The fact that we react the same way to showing, again, that objectivity is not a concep- these two things is sufficient for saying tual truth about the moral domain, but rather they are both members of the same cat- a negotiable add on, which can be abandoned in egory; they are both immoral. light of counter-evidence. Faith in objectivity goes up with certain relgious beliefs (e.g., divine When given these options, 75% chose option (b) command theory), and goes down when con- for immoral, resisting the first option which is fronted with the fact that decent, intelligent tantamount to an error theory. Exactly as many people have very different moral convictions. In chose option (b) for red, and a few more picked Quine’s terms, moral objectivism, when it is (b) for humorous (90%). In contrast, (a) was found, may be collateral information rather the dominant answer for the natural kind terms, than an analytic truth—a belief about morality tuberculosis and beetles (55% and 65% respect- that we are willing to revise. ively). This suggests that people do not treat To test this hypothesis, I conducted a sur- moral terms the way that they treat natural vey study in which I compared a moral predic- kind terms. Even if many people happen to ate (immoral) to two natural kind terms (beetle think that morality is objective (as the studies and tuberculosis), which paradigmatically aim by Nichols 2004, and Goodwin & Darley 2008, to designate objective properties, and to two suggest), they are willing to give up on this be- terms that are often said to represent secondary lief without abandoning their moral concepts. qualities (red and humorous). If natural kind They are willing to treat those concepts as re- terms have a presumption of objectivity, then sponse-dependent. any threat to that presumption should lead I think these results can be best inter- people to conclude that those terms don’t refer. preted as follows. Moral concepts are neutral Things are a little trickier with terms such as about moral objectivity. People can acquire red and humorous: many people believe that these concepts without any beliefs about what they designate objective properties, but are will- kinds of properties they designate. This neutral- ing to give up this assumption when presented ity begets a kind of resistance to error. If there with countervailing evidence. When told that are no objective moral properties, then it there is no unifying essence to humor, people do wouldn’t follow that moral judgments fail to not conclude that nothing is funny; they con- refer; it would mean only that they refer to re- clude that humorousness is a property that de- sponse-dependent properties. Thus, it is all but pends on our responses. In other words, ob- guaranteed that some moral judgments will jectivity is not analytically entailed by humor- come out true, and to this extent the evidence ous or red. It is collateral information. My study favors moral realism (defined as the view that was designed to see if immoral followed this there are truthmakers for some moral judg- same pattern. ments). Mackie mistakes a popular but dispens- A group of college undergraduates read able belief about morality for an analytic truth. the following vignette for the immoral case, His error theory rests on an error. In fact, his with comparable vignettes for the other terms: argument for the error theory may rest on two

Prinz, J. (2015). Naturalizing Metaethics. In T. Metzinger & J. M. Windt (Eds). Open MIND: 30(T). Frankfurt am Main: MIND Group. doi: 10.15502/9783958570511 20 | 27 www.open-mind.net mistakes, the second of which we will come to this question about moral terms. This question presently. Of course, this is just one study, and frames a historical debate between Francis other interpretations may be available, but it Hutcheson, who may have believed that our provides some evidence against Mackie’s con- moral sentiments track objective moral truths, ceptual claim and shows how empirical findings and David Hume, who suggests that morality might be used to explore whether moralizers depends on human responses. The claim that are, as he suggests, committed to objectivism. moral judgments track objective properties is Extant empirical evidence suggests otherwise. called the moral sense theory. It seems to have been defended by Francis Hutcheson in the 2.4 Sensibility vs. moral sense eighteenth century. It may even have been Kant’s considered view, since he had an object- The survey study just described suggests that ive procedure for arriving at moral truth, but one can possess moral concepts without know- also insisted that every moral judgment is asso- ing whether moral judgments refer to properties ciated with a moral feeling. The moral sense that are objective. The survey also brings out view finds an analogue in contemporary authors the possibility that people are willing to accept who combine external standards of moral truth the conclusion that moral truth depends on our with motivationally charged moral psychologies responses. But the survey does not settle (e.g., Campbell 2007; Copp 2001; see also whether a response-dependent theory is true. Railton 2009, who makes a modest move in that This is the next question on the decision tree. direction). The alternative view, which says As we have seen, Mackie thinks action-guiding- that moral judgments refer to response-depend- ness and objectivity are incompatible. This may ent properties, has been called the sensibility suggest that he sees no room for a theory that theory (McDowell 1985; Wiggins 1987). We can combines moral objectivity with the view that now ask whether there is any way to decide moral judgments have motivational pull. This, between these options empirically. however, is Mackie’s second mistake. The hypo- I think there is some reason to favor sens- thesis that morality has an emotional basis re- ibility over moral sense. For the moral sense veals a way out of Mackie’s argument for in- theory to be true, there would have to be a can- compatibility. Emotions are action-guiding in didate objective property to which our moral that they motivate us to act. But some emo- concepts could refer. Unfortunately, I cannot tions may also represent objective features of undertake a review of modern moral sense the- the world. Fear, for example, may represent ories here, but I will offer, instead, a more gen- danger, and danger may be an objective prop- eral line of empirically-informed resistance. erty. Emotions can represent objective proper- Moral rules are emotionally conditioned, and ties in a motivating way: they simultaneously communities condition people to avoid a wide pick up on information while compelling us to range of different behaviors. Within a given so- respond adaptively. The fact that fear is action- ciety, the range of things that we learn to con- guiding does not rule out the possibility that it demn is remarkably varied. Examples include is designed by evolution to track objective physical harm, theft, unfair distributions, neg- threats. Likewise, disgust is action-guiding but lect, disrespect, selfishness, self-destruction, in- it may register real sources of contamination. sults, harassment, privacy invasions, indecent This brings us back to “icky.” This emo- exposure, and sex with the wrong partners tionally-expressive term may refer to something (children, animals, relatives, people who are objective, like contamination, or to something married to other people). One might think that subjective, such as the tendency to cause feel- all of these wrongs have a common underlying ings of nausea. We can ask whether ickiness is essence. For example, one might propose that objective or subjective, even if we grant that the each involves a form of harm. But this is simply word “icky” is expressive. Expressive terms can not true. Empirical evidence shows that people have objective referents. Likewise, we can ask condemn actions that have no victims, such as

Prinz, J. (2015). Naturalizing Metaethics. In T. Metzinger & J. M. Windt (Eds). Open MIND: 30(T). Frankfurt am Main: MIND Group. doi: 10.15502/9783958570511 21 | 27 www.open-mind.net consensual sex between adult siblings and eat- example, Murphy et al. (2000) presented people ing the bodies of people who die in accidents with cases of incest and cannibalism where it (Murphy et al. 2000). Furthermore, harm itself was extremely salient that no one was harmed. is a subjective construct. It cannot be reduced They invited people to revise knee jerk moral to something like physical injury. Privacy viola- intuitions and rule that, on reflection, these vic- tions are regarded as a kind of harm, even timless actions are permissible. A piddling 20% though they don’t hurt or threaten health, revised accordingly, but 80% stuck to their ori- whereas manual labor is not considered a harm, ginal view. Moral sense theories seem to place but it threatens the body more than, say, theft. their bets on the 20%. The challenge is to ex- Similar problems arise if we try to define moral plain why the stubborn and considered opinions wrongs in terms of autonomy violations. Man- of the majority are performance errors of some datory education violates autonomy, but it is kind. considered good, and consensual incest is an ex- Given the diversity of things about which pression of autonomy, but is considered bad. people moralize, I think the sensibility theory is Realists would no doubt resist some of more promising than the moral sense theory. these claims, but theirs is an uphill battle. On Wrongness is projected, not perceived. The the face of it, morality lacks a common denom- property of being wrong is the property of caus- inator. Empirical surveys of human values sug- ing negative sentiments, not a response-inde- gest that moral rules are a potpourri, which can pendent property that those sentiments are de- be extended and contracted in any number of signed to detect. This conclusion follows from ways, with no fixed ingredients. Or rather, the an inference to the best explanation. Empiric- common denominator is not a property shared ally it looks as if there is no common essence to by the things we condemn, but rather by the the things that we find morally wrong—a find- condemning itself. Moral sense theorists liken ing that is difficult to explain on the moral morality to perception, and, in so doing, they sense model, but easy to explain on the assump- imply that there is an external feature of the tion that wrongness is response dependent. By world that our moral sentiments pick up on. analogy, imagine that we catalogue the things But there is little reason to believe this. Unlike that make people laugh, and find that they lack perception, there is massive variation in what a shared essence. This would imply that we moralize, and there is a perfectly good ex- laughter does not pick up on an objective prop- planation for this: the content of morality is de- erty. The things that we find funny are unified termined by social conditioning rather than by by the very fact that we are amused by them. the mind-independent world. Morality is not Likewise for the things we find immoral: disap- something we get by simply observing. probation carves the moral landscape. The foregoing is offered as an empirical challenge to moral sense theories, not a decisive 2.5 Relativism vs. ideal observers refutation. Too often philosophers stick with ex- amples of moral norms that clearly concern I have just been arguing that moral truth is re- harm or violations of autonomy. This inflates sponse-dependent. Moral judgments can be optimism about a unifying essence. If one uses true, but their truth depends on our sentiments. empirical methods to discover the full range of Something is immoral if it causes anger, disgust, things that people actually moralize (such as guilt, and shame in us. But now we can ask, victimless harms), the task of finding a unified who does “us” refer to here? Whose sentiments essence looks much harder. Moral sense theor- determine moral truth? This brings us to the fi- ists might reply that this diversity is illusory. nal question in the metaethics decision tree. They might say, for example, that people would Can divergent responses have equal claim to stop condemning victimless crimes on reflection. truth? That claim is amenable to empirical testing, Empirical evidence strongly suggests that and so far the tests provide little support. For moral sentiments vary, both within and across

Prinz, J. (2015). Naturalizing Metaethics. In T. Metzinger & J. M. Windt (Eds). Open MIND: 30(T). Frankfurt am Main: MIND Group. doi: 10.15502/9783958570511 22 | 27 www.open-mind.net cultures. Within a culture, the clearest divisions moral judgment can vary depending on whose are between political orientations. Liberals and values are being expressed. conservatives have interminable debates, even The ethical universalist can resist this con- when they are exposed to the same science and clusion by offering an antidote to moral vari- education. Research suggests that these debates ation. The most natural strategy would be to come down to fundamental differences in moral defend universality by developing an ideal ob- values. Conservatives are much more likely than server theory, and to argue that, under ideal liberals to emphasize purity, authority, and pre- epistemic conditions (which might include ex- servation of the in-group in justifying their ternal factors as well as being an epistemically moral norms (Haidt 2007). These things are ideal agent), judges would arrive at the same foundational for conservatives and largely irrel- set of moral values. This strikes me as woefully evant to liberals. unlikely. Once we grant that sentimentalism is Across cultures, differences are even true, and that our sentiments track response-de- greater. Everything that we condemn is accep- pendent properties, it’s not clear how to settle ted somewhere else (such as slavery and tor- on which observer is ideal. Two people who ture), and things that have been condemned by have the same factual knowledge may have dif- other cultures (such as women’s suffrage) have ferent sentiments as a result of differences in been embraced by us. There are cultures whose temperament (Lovett et al. 2012), reward sens- moral outlooks are dominated by considerations itivity (Moore et al. 2011), gender (Fumagalli et that we tend to downplay in the post-industrial al. 2010), class (Côté et al. 2013), and age West (sanctity and honor, for example), and (Truett 1993). Whose sentiments are right? ideals that are central to our moral outlook ap- Moreover, the standard traits associated with pear to be modern inventions (rights and the ideal observation may be problematic in the idea of human equality). moral domain. Should we consult someone who Descriptively, then, people do not seem to is disinterested when we know, empirically, that have the same moral values, within or across distance from a situation can lead to moral in- cultures. There is divergence in our sentiments. difference? Should we consult someone who has Some of this divergence might diminish if we not been conditioned by a particular culture filtered out cases where people were reasoning when we know that innate sentiments are un- badly or on poor evidence, but there is ample likely to deliver moral attitudes? Should we con- evidence that disagreements remain among sult someone who attends to every detail of a people who reason carefully and draw on the case, when we know that framing, vivid descrip- same factual knowledge. Indeed, if we filter for tion, and concreteness can alter moral judg- good reasoning, divergence might increase ments? These problems strike me as insuper- rather than decrease: consider professional able. There are no clear criteria for ideal obser- normative ethicists, who are experts at reason- vation and no reason to believe that careful ob- ing but nevertheless arrive at varied and novel servers would converge. moral perspectives that neither converge with In posing this challenge, I am inviting each other nor with the communities to which ideal observer theorists to look at empirical they belong. findings and propose epistemic standards that I think such descriptive moral relativism would overcome the sources of variation men- provides support for metaethical moral relativ- tioned here. Some ideal observer theories try to ism. This would be a terrible inference on its be empirically responsive in this way. For ex- own, as every metaethics textbook points out, ample, Smith (1994) advances the hypothesis but the inference gains plausibility if bolstered that ideal rational agents would converge, but by a premise I argued for above: moral truth is he also realizes that some readers might be re- dependent on our responses. If responses vary, luctant to share his optimistic outlook. To quell even under favorable epistemic conditions, and these doubts he makes three empirical observa- responses determine truth, then the truth of a tions (p. 188): there is considerable moral con-

Prinz, J. (2015). Naturalizing Metaethics. In T. Metzinger & J. M. Windt (Eds). Open MIND: 30(T). Frankfurt am Main: MIND Group. doi: 10.15502/9783958570511 23 | 27 www.open-mind.net vergence already (he cites the existence of thick cedures, they may actually promote variation, concepts as evidence: we all think brutality is leading to an endless proliferation of values and bad and honesty is good); there has been moral an ever widening gulf between those who cher- progress (he cites slavery, among other ex- ish diversity and those who reside in more tradi- amples); and entrenched disagreements often re- tional societies. From a social science perspect- flect faulty rationality, such as religious beliefs. ive, the prospects for a universal morality look Here, I think further empirical scrutiny would grim. weaken Smith’s case. Divergence is rampant, Once the case for relativism is established, and people disagree on the scope of thick con- the question arises: relative to what? Are moral cepts (is torture brutal? is espionage judgments relative to value systems? Are those dishonest?). Cases of (what we consider to be) systems individuated at the scale of cultures moral progress are, I’ve noted, often driven by and subcultures or do they vary across individu- economic upheavals and other irrational factors, als? Little empirical work has been done to ad- with reasoning playing a post-hoc role. Finally, dress this question, but let me end with a sug- disagreements remain after bad reasoning and gestion about how to proceed. When examining religiosity are controlled for; the examples men- the semantics of natural kind terms, philosoph- tioned, in formulating the challenge include ers have sometimes appealed to a linguistic divi- things such as temperament and framing effects. sion of labor (Putnam 1975). We defer to ex- I think empirical evidence provides little reason perts and thereby license them to adjudicate to expect that rational and informed observers the boundaries between natural kinds. Now we would deliver consistent verdicts. can ask, is there such a thing as moral expert- In light of such worries, universalists might ise? Do we appeal implicitly or explicitly to abandon the ideal observer theory and offer in- moral experts? Would we change our moral stead a procedural approach to consensus, ar- judgments if the designated members of our guing that people would and should converge if community told us we were morally mistaken? they arrived at their sentiments in the right We don’t know the answers to such questions, way. For example, many people might agree because moral expertise has not been intens- that it is good to arrive at decisions democratic- ively studied. I suspect there will be consider- ally, taking multiple sentiments into considera- able individual differences, with members of tion, and we might sentimentally endorse the more traditional societies showing more willing- outcome of democratically-resolved moral dis- ness to defer. But I also suspect that deference putes. Though I cannot make the case here, I in the moral domain will be less prevalent than suspect the problems with such a procedural for natural kinds; we are more inclined to take approach outweigh its prospects. Democratic ourselves as having morally authoritative in- decision-making does not result in moral con- sight. What is most clear, however, is whether sensus; it can even polarize. When such proced- the scope of the relativity depends ultimately ures increase consensus it is often through on how we use moral concepts and terms; and power and prestige rather than sentimental con- this is something that can be investigated em- vergence. Our faith in democratic procedures pirically. Naturalizing relativism will require the may also be an expression of moral relativism marriage of cultural anthropology and sociolin- rather than a solution. Democratic procedures guistics. From the armchair, it is tempting to are an historical anomaly, which emerged in the think there is a single true morality; introspect- modern period with the rise of capitalism, and ive reflection tends towards solipsism. they have often been used to oppress minorities and to impose the values of the many over the few. Perhaps such procedures are an improve- 3 Conclusion ment over totalitarian forms of decision-making, but they do not remedy relativism. Indeed, as Throughout this discussion, we have worked our societies move towards consensus-building pro- way down a metaethics decision tree. I have

Prinz, J. (2015). Naturalizing Metaethics. In T. Metzinger & J. M. Windt (Eds). Open MIND: 30(T). Frankfurt am Main: MIND Group. doi: 10.15502/9783958570511 24 | 27 www.open-mind.net made a case for a relativist cognitivist senti- fell out of fashion, however, and it has taken a mentalist sensibility theory. Admittedly, each of century to get back to this incipient moment. my arguments is only a first pass, and much With the linguistic turn, Anglophone philosoph- more could be said for and against these posi- ers became convinced that we should all learn tions. Many of the empirical findings that I logic because it would help us make progress. have described are preliminary. My main goal Logic did help, and it did not undermine philo- here is not to make a decisive case for any posi- sophy. Now, we can encourage all philosophers tion in metaethics. Rather, I am pleading for to learn about methods and results used in the the relevance of empirical methods in doing this relevant social and physical sciences. The payoff traditionally philosophical work. Moral philo- of this naturalistic turn may be vastly greater sophy is undergoing a process of naturalization. than the linguistic turn. Science, not formal lo- This has been felt most strongly in normative gic, is positioned to tell us whether morality is a theory (e.g., the debate about the status of human construction. character in virtue ethics) and moral psychology (e.g., questions about how deontological and Acknowledgments consequentialist judgments are made). I hope to have shown that the empirical work also bears This discussion has benefited immeasurably directly on metaethical questions—questions from the feedback of anonymous referees about what, if anything, is the source of moral and from Ying-Tung Lin, Jessica McCor- truth. mack, Thomas Metzinger, and Jennifer Windt. I Empirical work cannot replace philosoph- am grateful for their close reading and helpful ical toil. We need philosophy to pose questions suggestions. and identify possible theories. Experimental design is itself a kind of philosophical reasoning, and it takes considerable argumentation to move from data to theory. Naturalization is supplementation, not usurpation. But it is not just supplementation. The empirical arsenal may just be our best hope for adjudicating philosophical debates. Reflection can delineate the logical space, but we need observation to locate ourselves therein. Philosophers have al- ways relied on observation, in some sense, but scientific methods allow us to observe processes that are unconscious, inchoate, or distant in space and time. Empirical studies can test the content, prevalence, and malleability of intu- itions, and they can also tell us where our intu- itions come from—a question of central metaethical concern. We should embrace any tools that help us resolve the questions that we are employed to answer. A century ago, there was a linguistic turn, and philosophers began to treat traditional philosophical problems as amenable to semantic analysis. Around the same time, the boundary between philosophy and psychology was still blurred, and journals such as Mind published articles that we might now classify as psychological. Such crossovers

Prinz, J. (2015). Naturalizing Metaethics. In T. Metzinger & J. M. Windt (Eds). Open MIND: 30(T). Frankfurt am Main: MIND Group. doi: 10.15502/9783958570511 25 | 27 www.open-mind.net

References Goodwin, G. P. & Darley, J. M. (2008). The psychology of meta-ethics: Exploring objectivism. Cognition, 106 Blackburn, S. (1998). Ruling passions. Oxford, UK: Ox- (3), 1339-1366. 10.1016/j.cognition.2007.06.007 ford University Press. Graham, J., Haidt, J. & Nosek, B. A. (2009). Liberals Blair, R. J. R. (1995). A cognitive developmental ap- and conservatives rely on different sets of moral found- proach to morality: Investigating the psychopath. Cog- ations. Journal of Personality and Social Psychology, nition, 57 (1), 1-29. 96 (5), 1029-1046. Borg, J., Hynes, C., van Horn, J., Grafton, S. & Sinnott- Greene, J. D. (2007). The secret joke of Kant’s soul. In Armstrong, W. (2006). Consequences, action, and in- W. Sinnott-Armstrong (Ed.) Moral psychology, Vol. 3: tention as factors in moral judgments: An fMRI invest- The neuroscience of morality, emotion, disease, and igation. Journal of Cognitive Neuroscience, 18 (5), 803- development. Cambridge, MA: MIT Press. 817. Greene, J. D., Sommerville, R. B., Nystrom, L. E., Dar- Boyd, R. (1988). Essays on moral realism. Ithaca, NY: ley, J. M. & Cohen, J. D. (2001). An fMRI investiga- Cornell University Press. tion of emotional engagement in moral judgment. Sci- Brandt, R. (1959). Ethical theory: The problems of norm- ence, 293 (5537), 2105-2108. 10.1126/science.1062872 ative and critical ethics. Englewood Cliffs, NJ: Pren- Haidt, J. (2007). The new synthesis in moral psychology. tice-Hall. Science, 316 (5827), 998-1002. 10.1126/science.1137651 Campbell, R. (2007). What is moral judgment? Journal Hare, R. D. (1993). Without conscience: The disturbing of Philosophy, 104 (7), 321-349. world of the psychopaths among us. New York, NY: Copp, D. (2001). Realist-expressivism: A neglected option Pocket Books. for moral realism. Social Philosophy and Policy, 18 (2), Huebner, B., Dwyer, S. & Hauser, M. (2009). The role of 1-43. 10.1017/S0265052500002880 emotion in moral psychology. Trends in Cognitive Sci- Côté, S., Piff, P. K. & Willer, R. (2013). For whom do ences, 13 (1), 1-6. 10.1016/j.tics.2008.09.006 the ends justify the means? Social class and utilitarian Joyce, R. (2006). The evolution of morality. Cambridge, moral judgment. Journal Of Personality And Social MA: MIT Press. Psychology, 104 (3), 490-503. 10.1037/a0030931 Kant, I. (1797). The metaphysics of morals. M. J. Gregor Doris, J. M. (1998). Persons, situations, and virtue ethics. (Trans.). Cambridge, UK: Cambridge University Press. Noûs, 32 (4), 504-530. Koenigs, M., Young, L., Adolphs, R., Tranel, D., Cush- Eskine, K. J. (2011). From perceptual symbols to abstrac- man, F., Hauser, M. & Damasio, A. (2007). Damage tion and back again: The bitter truth about morality. to the prefrontal cortex increases utilitarian moral New York, NY: Doctoral dissertation, Department of judgments. Nature, 446, 908-911. Psychology, City University of New York. 10.1038/nature05631 Eskine, J. K., Kacinik, A. N. & Prinz, J. J. (2011). A bad Kornblith, H. (Ed.) (1985). Naturalizing epistemology. taste in the mouth: Gustatory disgust influences moral Cambridge, MA: MIT Press. judgment. Psychological Science, 22 (3), 295-299. Kramer, P. D. (1993). Listening to Prozac. New York, 10.1177/0956797611398497 NY: Viking. Firth, R. (1952). Ethical absolutism and the ideal ob- Lovett, B. J., Jordan, A. H. & Wiltermuth, S. S. (2012). server. Philosophy and Phenomenological Research, 12 Individual differences in the moralization of everyday (3), 317-345. life. Ethics & Behavior, 22 (4), 248-257. Flanagan, O. (1991). Varieties of moral personality: Eth- 10.1080/10508422.2012.659132 ics and psychological realism. Cambridge, MA: Harvard Mackie, J. L. (1977). Ethics: Inventing right and wrong. University Press. London, UK: Penguin. Fumagalli, M. M., Ferrucci, R. R., Mameli, F. F., Marce- McDowell, J. (1985). Morality and objectivity. London, glia, S. S., Mrakic-Sposta, S. S., Zago, S. S. & Priori, UK: Routledge & Kegan Paul. A. A. (2010). Gender-related differences in moral judg- Mikhail, J. (2002). Aspects of the theory of moral cogni- ments. Cognitive Processing, 11 (3), 219-226. tion: Investigating intuitive knowledge of the prohibi- 10.1007/s10339-009-0335-2 tion of intentional battery and the principle of double Gibbard, A. (1990). Wise choices, apt feelings. Cam- effect. Economics Research Paper, 762385. bridge, MA: Harvard University Press. http://ssrn.com/abstract=762385

Prinz, J. (2015). Naturalizing Metaethics. In T. Metzinger & J. M. Windt (Eds). Open MIND: 30(T). Frankfurt am Main: MIND Group. doi: 10.15502/9783958570511 26 | 27 www.open-mind.net

Moore, A. B., Stevens, J. & Conway, A. A. (2011). Indi- Schmidt, E. Z. & Bonelli, R. M. (2008). Sexuality in vidual differences in sensitivity to reward and punish- Huntington’s disease. Wiener Medizinische Wochens- ment predict moral judgment. Personality And Indi- chrift, 158 (3-4), 84-90. 10.1007/s10354-007-0477-8. vidual Differences, 50, 621-625. Seidel, A. & Prinz, J. J. (2013a). Sound morality: Irritat- 10.1016/j.paid.2010.12.006 ing and icky noises amplify divergent moral domains. Murphy, S., Haidt, J. & Björklund, F. (2000). Moral Cognition, 127 (1), 1-5. 10.1016/j.cognition.2012.11.004 dumbfounding: When intuition finds no reason. Unpub- (2013b). Mad and glad: Musically induced emo- lished manuscript, Department of philosophy, Univer- tions have divergent moral impact. Motivation and sity of Virginia. Emotion, 37 (3), 629-637. 10.1007/s11031-012-9320-7 Nichols, S. (2004). Sentimental rules: On the natural Silk, J. B., Brosnan, S. F., Vonk, J., Henrich, J., Pov- foundations of moral judgment. New York, NY: Oxford inelli, D. J., Richardson, A. F., Lambeth, S. P., Mas- University Press. caro, J. & Schapiro, S. J. (2005). Chimpanzees are in- (2008). Sentimentalism naturalized. In W. Sin- different to the welfare of other group members. nott-Armstrong (Ed.) Moral psychology: The cognitive Nature, 435, 1357-1359. 10.1038/nature04243 science of morality: Intuition and diversity (pp. 255- Smith, M. (1994). The moral problem. Oxford, UK: 274). Cambridge, MA: MIT Press. Blackwell. Olasov, I. (2011). Register variation and the moral cog- Truett, K. R. (1993). Age differences in conservatism. nitivism debate. Unpublished manuscript, City Univer- Personality and Individual Differences, 14 (3), 405-411. sity of New York, Graduate Cente. 10.1016/0191-8869(93)90309-Q Prinz, J. J. (2004). Gut reactions: A perceptual theory of Turiel, E. (1983). The development of social knowledge: emotion. New York, NY: Oxford University Press. Morality and convention. Cambridge, UK: Cambridge (2005). Imitation and moral development. In S. University Press. Hurley & N. Chater (Eds.) Perspectives on imitation: Waldmann, M. R., Nagel, J. & Wiegmann, A. (2012). From cognitive neuroscience to social science. Cam- Moral judgment. In K. J. Holyoak & R. G. Morrison bridge, MA: MIT Press. (Eds.) The Oxford handbook of thinking and reasoning. (2007a). Is morality innate? In W. Sinnott-Arm- Oxford, UK: Oxford University Press. strong (Ed.) Moral psychology, vol 1: Evolution of mor- Wastell, C. & Booth, A. (2003). Machiavellianism: An als (pp. 367-406). Cambridge, MA: MIT Press. alexithymic perspective. Journal of Social and Clinical (2007b). The emotional construction of morals. Psychology, 22 (6), 730-744. Oxford, UK: Oxford University Press. 10.1521/jscp.22.6.730.22931 Putnam, H. (1975). The meaning of “meaning”. Mind, Widom, C. S. (1976). Interpersonal conflict and coopera- Language and Reality: Philosophical Papers, Volume 2 tion in psychopaths. Journal of Abnormal Psychology, (pp. 215-271). Cambridge, UK: Cambridge University 85 (3), 330-334. 10.1037/0021-843X.85.3.330 Press. Wiggins, D. (1987). A sensible subjectivism. In D. Wig- Quine, W. V.O. (1969). Ontological relativity and other gins (Ed.) Needs, values, truth: Essays in the philo- essays. New York, NY: Columbia University Press. sophy of value (pp. 185-214). Oxford, UK: Blackwell. Railton, P. (1993). What the non-cognitivist helps us to Woolfolk, R. L. (2013). Experimental philosophy: A see the naturalist must help us to explain. In J. methodological critique. Metaphilosophy, 44 (1-2), 79- Haldane and C. Wright (Eds.), Reality, Representation 87. 10.1111/meta.12016 and Projection (pp. 279-297). Oxford, UK: Oxford Uni- Wrangham, R. (2004). Killer species. Daedalus, 133, 25- versity Press. 35. (2009). Internalism for externalists. Philosophical Issues, 19 (1), 166-181. 10.1111/j.1533-6077.2009.00165.x Ruse, M. (1991). A companion to ethics. In P. Singer (Ed.) A companion to ethics (pp. 500-510). Oxford, UK: Blackwell. Sayre-McCord, G. (Ed.) (1988). Essays on moral realism. Ithaca, NY: Cornell University Press.

Prinz, J. (2015). Naturalizing Metaethics. In T. Metzinger & J. M. Windt (Eds). Open MIND: 30(T). Frankfurt am Main: MIND Group. doi: 10.15502/9783958570511 27 | 27