1. Preliminaries

To avoid confusion, an initial terminological orientation is required (p. 23, n. 4). As I've indicated, the theory of agency I'm after is a theory of morally responsible agency, a notion rather more demanding than thinner senses of agency, where “agent” may reference any entity capable of purposive movement. This thinner sense distinguishes agents like puppies from objects like Pop-Tarts, and is common in psychology, as with the developmental literature. For example, human infants behave in ways indicating that they may categorize novel objects – in one study, an entity poetically titled an “agentive blob” – as agents, if the objects display performances indicative of perception, communication, and goal-directedness (Johnson Reference Johnson2003).

The properties sufficient for being an agentive blob are insufficient for being a morally responsible agent, even if they do mark important differences between puppies and Pop-Tarts. Lots of critters – honeybees, for example – exhibit perception, communication, and goal-directedness, without engendering serious temptation toward subjecting them to morally important responses; fine to be angry that you were stung by a bee (one damn thing after another!), but inapposite to be angry at the bee (which is, after all, only a bee). I've no interest in disparaging the sort of agency displayed by agentive blobs and honeybees – an important fact, if detecting this kind of agency is prominent in the human cognitive repertoire. It is not, however, sufficient for morally responsible agency; that notion, as I try to show, requires rather more. While I sometimes drop the modifier “morally responsible,” it is, unless otherwise noted, a morally responsible agency at issue whenever “agency” appears here (and in the book).

At this juncture, I should also provide methodological orientation, and say something about the status of the empirical material on which my remorselessly interdisciplinary approach relies. As the twenty-first century sputters toward its third decade, numerous replication failures attending iconic studies, particularly in experimental social psychology, have – as they should – occasioned much soul-searching by producers and consumers of the social sciences, and have – as they should – encouraged greater methodological scrupulosity in social scientific practice. The “RepliGate” controversy is not yet concluded (and involves controversy about whether there should even be a controversyFootnote ²), which means that Talking to Our Selves includes some of my own soul-searching about interpreting science in conditions of uncertainty (pp. 44–49; cf. Machery & Doris Reference Machery, Doris, Voyer and Tarantola2017).

RepliGate, and the inevitability of scientific controversy more generally, dictate taking one's science with a healthy dose of caution. But it should be stressed that empirical findings intimating the existence of incongruence are not “one off” curiosities, but are part of established trends, and the dual process perspective in which I situate the phenomena have been proposed for most everything psychologists study: attention (Schneider & Shiffrin Reference Schneider and Shiffrin1977), learning (Reber Reference Reber1993), memory (Roediger Reference Roediger1990), perception (Norman Reference Norman2002), reasoning (Evans & Over Reference Evans and Over1996), decision making (Kahneman & Frederick Reference Kahneman, Frederick, Gilovich, Griffin and Kahneman2002), person interpretation (Gilbert et al. Reference Gilbert, Pelham and Krull1988), delay of gratification (Metcalfe & Mischel Reference Metcalfe and Mischel1999), psychopathology (Beevers Reference Beevers2005), and moral judgment (Cushman et al. Reference Cushman, Young, Greene and Doris2010).

The theoretical division presupposed by dual process theory doubtless is, like all theoretical divisions, rather less crystalline in world than on page (p. 50). (It should especially be noted that the analytic/automatic distinction crosscuts the intelligent/unintelligent distinction: it's not smart to concoct intricate excuses for procrastination, and it's not dumb to reflexively bolt from danger.) Nevertheless, that different cognitive processes can proceed independently of, and sometimes oppositionally to, one another is a hypothesis supported by an impressive range of converging evidence (pp. 49, 52, 56). Indeed, even critics of dual process theory (e.g., Machery Reference Machery2009, pp. 121–150; Keren & Schul Reference Keren and Schul2009, pp. 141–42) acknowledge the existence of the cognitive conflict I'm calling incongruent parallel processing.

While individual studies ought to be critically examined on their merits (pp. 53–64), the skeptical worry does not depend on the fate of particular “criterial” experiments. (If the Ballot Order Effect turned out to be chimerical, for example, I'd have plenty of other illustrative effects to choose from.) Whatever controversy afflicts particular studies, labs, or research programs, the existence of incongruence (however exactly it should be theorized) isn't in question. It is this uncontroversial observation on which my argument relies.

2. Skepticism

I presume that the problem animating Talking to Our Selves is “intuitive”; I presume many folks will think that incongruence sits uneasily with their (pre-theoretical) self-understanding (pp. x, 80, 158). That is, I presume that for those people who believe that they're directing their own behavior and shaping their own life (as I presume many people believe), learning that they were prey to something like a Ballot Order Effect would give them serious pause.

Say you are an undecided voter, and as you enter the booth, you decide, finally, to help make your country great again, and vote for Candidate Creepy. Say that subsequently, I provide convincing evidence that you in fact voted for Creepy because his leading place on the ballot tipped your decision – if not for ballot order, you'd have voted for his slightly less objectionable rival. (For the moment, don't mind that the actual evidence is aggregate, and can't decide individual cases; more on that in a bit). How should you feel about your decision?

This pre-theoretical unease, I further contend, also afflicts many philosophical theories of agency and responsibility, such as the influential approaches associating agency with “reasons responsiveness” (Fischer & Ravizza Reference Fischer and Ravizza1998; Nelkin Reference Nelkin2011; Vargas Reference Vargas2013). Suppose you read research associating sedentary lifestyles with poor health outcomes, and commence dragging yourself to the gym; your behavior is responsive to the balance of reasons commending it, and you've thereby, with your exercising, pulled off an exercise of agency. But for your (imaginary) vote, you took yourself to be responding to Creepy-reasons, like reclaiming lost greatness, while your vote was actually decided by a rationally arbitrary “non-reason,” Creepy's place on the ballot. Here, your conduct was not appropriately responsive to reasons, and therefore not properly agential, according to reasons responsiveness theories. My point is not that these theories – or any other philosophical theory – cannot accommodate the disconcerting facts, but that they need to do so (p. 171). An anti-skeptical theory of human agency – one maintaining that exercises of morally responsible agency may be confidently identified in a good measure of human behavior – must account, in a systematic and detailed way, for the phenomena of incongruence, but philosophical efforts in this direction have hitherto been limited.

As I've said, the provocative cases of incongruence involve psychological causes of behavior that are not plausibly taken as reasons for that behavior, as when a “dumb” automatic process bypasses a “smart” analytic one. Theorizing reasons is the stuff of uncounted dissertations in philosophy and the decision sciences, but one doesn't need fancy theory to appreciate the difficulty. A simple, broadly subjective, account here serves: when someone treats a consideration as a reason they should be willing to treat that consideration as a justification for their judgment or behavior (pp. 43–44). The notion of justification is also a dissertation bogey, but once again, an elaborate theory isn't needed (nor need one be attributed to the reasoner). Many people commend and defend what they think and do, both to themselves and others, and when they engage in this activity, they are engaged in a reason-giving, justificatory practice (p. 44). The relevant divergence of reasons and causes obtains when people would decline to reference the psychological origins of their behavior – those origins that would figure in a reasonably complete and accurate psychological explanation of that behavior – as appropriate considerations in this practice.

I don't primarily intend familiar cases of what philosophers call akrasia, or weakness of will, where people succumb to desire, appetite, or temptation against their better judgment (pp. 52, 161–162): I know the sugar, grease, and gluten-filled donut is a nutritional apocalypse I'll surely regret later, but I nonetheless devour it like a starving lion. However much trouble they make for health and happiness, many such cases don't make special trouble for agency. For when I inhale the donut, I'm doing so on perfectly intelligible grounds: Winslow's salt-caramel donuts are absurdly tasty, I know I'll quite enjoy eating one, and I very much desire to do so. Nor is it the case that the desire is somehow puzzling, or alien to me; gustatory pleasure is something that may be appropriately treated as a reason, and I sometimes treat it that way. Much as I hate to admit it, I suffer an altogether unpuzzling affection for Winslow's salt-caramel donuts, and if this affection manifests in my body composition, it manifests as a result of my exercising agency, and is nobody's fault but mine. (Were I addicted to Winslow's salt-caramel donuts, there would be questions about my agency, but my interest here is the ordinary, not the psychopathological.)

In the most unsettling cases of incongruence, judgment is not overwhelmed, as in akrasia, but bypassed. And in many such cases there is little reason to think reason would be overmatched were the bypassing factor subject to scrutiny: probably not unduly difficult to disregard the influence of the Ballot Order Effect, when it's brought to one's attention. That's part of what makes the phenomena so unnerving. Hardly surprising that people can be moved by strong desires; more so, that they can be moved by factors that are scarcely intelligible as the objects of desire (p. 162).

A preliminary schema for skepticism about morally responsible agency is now visible (pp. 64–65). Cases of incongruence, where the pertinent psychological causes of her cognition or behavior would not be recognizedFootnote ³ by the actor as reasons favoring that cognition or behavior (were she aware of these causes at the time of performance), are cases where defeaters obtain. Where defeaters obtain, the exercise of agency does not obtain. If the presence of defeaters cannot be confidently ruled out, it is not justified to attribute the actor an exercise of agency. If there is general difficulty in ruling out defeaters, skepticism about agency ensues. (Again, akratic desirers will not always be defeaters, because such desires are frequently recognized as reasons [pp. 70–71].) In brief, the agency skeptic maintains that for any putative instance of agency, one cannot rule out the presence of a defeater (or defeaters) in the causal origins of that behavior, and one is therefore never justified in positing an instance of morally responsible agency.

Here, the standard for out-ruling is substantially ethical (p. 66). Responsibility attributions are associated with distribution of benefits (like reward and praise) and burdens (like punishment and blame), so the improper attribution of responsibility can result in the target of the attribution receiving unjust benefits or burdens. Thus, the requirement that the possibility of defeaters be “confidently” eliminated gets understood in terms that are more moral than purely epistemic. For one to be justified in an attribution of morally responsible agency, it should be the case that if one's initial attribution of agency is overturned by credible new evidence of defeaters, one is not guilty of wrongdoing for having apportioned benefit or burden according to the initial attribution. Viewed in this light, the skeptical thought is that the requisite moral confidence is never justified; one cannot be sufficiently confident that one is not guilty of wrongdoing when acting on attributions of morally responsible agency. So understood, the skeptic is not making an unreasonable demand for epistemic certainty (the familiar skeptical foible of “setting the bar too high”), but an eminently reasonable request that moral judgments be morally defensible.

It must be emphatically emphasized that the skeptical argument (like other skeptical arguments) does not require an empirical generalization about the frequency with which defeaters occur (p. 68). And a good thing: that argument is sure to be a stinker, whatever side one is on. For so far as I can see, nobody has any very exact idea what the frequencies are. It's obvious why: comparatively little behavior is closely observed, still less behavior is observed in controlled conditions, and for the behavior that is observed, there's often little assurance about the full range of explanatorily salient psychological antecedents. If the contest between skepticism and anti-skepticism had to be decided by speculation about frequencies, it wouldn't get decided.

Nevertheless, if the skeptical argument is to be more than a symptom of philosophical paranoia, the possibility of defeaters has to be something more than mere possibility. And so it is. The skeptical hypothesis – concerning the pervasive availability of behavioral explanations referencing defeaters – is not a loopy (and perhaps massively unlikely) proposition like philosophical fantasies of Evil Demons or Envatted Brains. Rather, it is a “live” hypothesis (Frances Reference Frances2005, pp. 560–61): it has been vetted by the relevant experts, and judged by a substantial number of them, on the basis of good evidence, to be about as likely as competing hypotheses (p. 66; cf. Davies Reference Davies2009, p. 169). Given the repeated appearance of incongruence in the context of scientific research, it is responsibly conjectured that defeaters occur with some regularity in everyday life; regularly enough, anyway, that the skeptical hypothesis is live, and has some non-trivial chance of being true, for any particular behavior. The matter can't be put more precisely than that. Nor need it be: the critical question concerns not how often defeaters should be thought to obtain, but how their presence can be ruled out for each putative exercise of morally responsible agency (p. 68).

I suspect anti-skeptical optimism is commonly funded by the imprecise suspicion that the envisaged defeaters are “pretty rare,” together with the sanguine assurance that pretty rare occurrences need not undermine the attribution of m orally responsible agency. Regarding the sanguine assurance, I'm not sure: depending on the gloss of rare or pretty rare, such events can certainly merit moral concern. (“Better to acquit 1,000 guilty persons than to execute a single innocent one.”) I suppose there is some gloss that warrants the assurance – 1 in 100?, 1,000?, 10,000? – but such a gloss is unavailable, faced with inevitable uncertainty about frequencies. And the imprecise suspicion fares no better than the sanguine assurance. Given how little is observed of – and the less that is known about – human psychology and behavior, it is obscure what could motivate justifiable confidence that defeaters are pretty rare. At least, that motivation had better not be (as I fear it often is) the unvarnished faith that the counterintuitive must be uncommon – a certitude inimical to the spirit of scientific discovery.

This framing allows us to better understand the relation of aggregate evidence, such as Ballot Order Effects, to particular instances, such as your vote for Creepy. What the aggregate indicates is that some individuals must be subject to the influence in question; otherwise, there would not be an effect (pp. 63–64). We cannot, for studies of this kind, be sure which individuals are affected. But this is not an objection to the argument, it is the argument. Once it be allowed that defeaters might be in play – and that's what must be allowed – we require considerations sufficient to warrant moral confidence they are not. That's the burden of ruling out. As with other skeptical arguments, the present skeptical argument is not undermined by uncertainty; it trades in uncertainty.

It should be noted that many of the discomfiting experimental findings may involve small effect sizes. One way to consider this question is in terms of the correlation coefficient (r), which quantifies the strength of association (covariation) between two variables: a coefficient of 1.0 indicates a perfect association, a coefficient of 0.00 indicates no association, and a coefficient of −1.0 indicates a perfect negative association. Effect size for correlation coefficients may be assessed with reference to a venerable proposal by Cohen (Reference Cohen1988, pp. 77–81) that for psychology (at least in “softer” branches like personality and social) a correlation of around 0.10 should be regarded as “small,” around 0.30 as “moderate,” and around 0.50 as “large.”

Alas, for much psychological research, correlation coefficients rating Cohen's large, or even moderate, occur relatively infrequently (Hemphill Reference Hemphill2003). When the famous “group effect” (Darley & Latané Reference Darley and Latané1968) is calculated as a correlation between the number of onlookers and bystander intervention, the value is a “moderate” −.38 (Funder & Ozer Reference Funder and Ozer1983, p. 110), and many celebrated findings in psychology, such as those relating implicit and explicit bias, may involve considerably smaller effects (Greenwald et al. Reference Greenwald, Banaji and Nosek2015). Are small effects “too small”? (Cohen [1988, p. 79] himself didn't think so.) For context, consider a rendering of effects in biomedical research as correlation coefficients (Meyer et al. Reference Meyer, Finn, Eyde, Kay, Moreland, Dies, Eisman, Kubiszyn and Reed2001, p. 130): regular aspirin consumption and reduced risk of heart attack, 0.02; chemotherapy and surviving breast cancer, 0.03; ever smoking and lung cancer within 25 years, 0.08. These relationships are not relationships, like the correlation between being human and being mortal, that are strong enough to be detected by “the naked eye” without the aid of statistical magnification (Jennings et al. Reference Jennings, Amabile, Ross, Tversky, Kahneman and Slovic1982, pp. 216–22). And obviously, they are clearly small by Cohen's lights.

However, such small effects may yet have practical import: take the aspirin, endure the chemo, and don't light up. The small relationship between aspirin use and reduced coronary risk might be thought of as pretty important, compared to something that (presumably) has no relationship, like owning a gray coat, or something that has a negative relationship, like obesity. Of course, if your habits or genes are bad enough, aspirin ain't gonna save you. Yet for many patients the costs associated with taking an aspirin every day or two are pretty minimal, so if you're at risk for a heart attack, why not? Furthermore, taking aspirin together with other interventions, like losing weight and exercising properly, might have a considerable cumulative effect.

If we can say that some lives are saved by small-effect interventions like aspirin (perhaps in concert with other interventions), why shouldn't we say that some exercises of agency are undermined by small-effect defeaters, like Ballot Order Effects (perhaps in concert with other defeaters)? To be sure, identifying statistically small aggregate effects does not allow confident conclusions about particular outcomes for particular individuals. But the small effects must be making a difference in some individual cases, or there would not be aggregate effects. Given the multitude of influences likely operative in any instance, one cannot confidently say where the difference was made, but this sort of uncertainty is actually part of the problem: what's making which difference for whom?

Once we allow that there are some of these rationally and ethically arbitrary influences on cognition and behavior, we are bound to admit there may be others. If stuff like that can make a difference, there could be many such influences in any particular instance. And while the impact of each such individual influence may be statistically small, the cumulative effect may be quite potent. For all one knows, any decision may be infested with any number of arbitrary influences. The claim is not that any one of the influences in question is momentous in the way illness, bereavement, and unemployment can be. Rather, the thought is that statistically small effects can sometimes be practically consequential – and an aggregation of such influences even more so.

I'm compelled to admit that I previously, while espousing skepticism about traditional conceptions of character, complained about small effect sizes in personality psychology, and it's arguable that some effects in personality are larger, perhaps considerably larger, than the kinds of effects I'm now celebrating. But for character skepticism, my concern was not that the personality effects are small, but that they are likely to be smaller than should be expected on familiar theories of character and personality (Doris Reference Doris2002, pp. 68, 71–75). Here, my position is the converse: the effects in question, small though they may sometimes be, are larger than should be expected on familiar theories of agency and responsibility – indeed, it often seems absurd that they could have any effect at all. This forces rethinking approaches to agency.

3. Reflectivism

Like much philosophy, Talking to Our Selves is structured agonistically. My primary agonist is reflectivism (pp. x, 17–23), a doctrine according to which the exercise of human agency consists in judgment and behavior ordered by self-conscious reflection about what to think and do. Typically, this doctrine is associated with a corollary: the exercise of human agency requires accurate reflection. In an exercise of agency, as construed by reflectivism, a person correctly divines the beliefs, desires, and other psychological states relevant to her decision, makes her decision in light of these states, and then acts accordingly. In short, reflectivism holds that the exercise of agency is characteristically reflective activity.

Reflectivism is another notion I guess is pretty darn “intuitive,” in the sense of enshrining a very familiar experience: you think about what you're hungry for, in perusing the menu, and I think about how much space I need, in perusing the real estate section. When we have these experiences of reflective agency, I confidently speculate, we don't usually think the outcome of our reflections was determined – to tweak my stock example – by the order of listings on the menu or in the adverts. (Let's rent the more expensive flat, it's listed first!)

I also insist that reflectivism is philosophically familiar: preoccupation with reflection is, arguably, the Western philosophical tradition's most distinctive feature, in both historical and contemporary guises, and is certainly a central theme in philosophical moral psychology (Doris Reference Doris2015b, p. 17; see also Arpaly Reference Arpaly2002, p. 20; Kornblith Reference Kornblith2010, p. 2; Reference Kornblith2012, p. 1). The most salient examples are the many Kantians in moral psychology and ethics, who often associate practical rationality with reflection (Korsgaard Reference Korsgaard1996, pp. 92–93; Reference Korsgaard2009, p. xi; cf. Moran Reference Moran2001, p. 127), resulting in a literature thick with references to the “reflective agent” and “reflective agency” (Velleman Reference Velleman1989, p. 5; 2000, pp. 12, 26–29, 124, 191–96; Wallace Reference Wallace and Watson2003, p. 437; Reference Wallace2006, pp. 150–51).

The accuracy corollary is equally recognizable (pp. 19–20). For reflectivists, agency requires that the actor detect practically salient facts about herself (Velleman Reference Velleman2000, p. 12); as Tiberius (Reference Tiberius2002, p. 13) puts it, a “person who does not know her deepest motivations because she is very psychologically complicated, or because she has formed impenetrable layers of self-deception, is missing something she would need to deliberate well.” If deliberation – which might be thought of as practical reflection – is predicated on erroneous self-understandings, agency is supposedly imperiled.

On this understanding, the self-ignorant don't deliberate effectively about what to think and do; in contrast, the practically effective “reflective agent” will enjoy some measure of accuracy in her deliberations. When the reflectivist's reflective agent cites reasons for her action, the reasons cited are supposed to accord with the causes of her action: if she believes she did something because she judged it the right thing to do, that judgment must figure in an accurate and appropriate causal explanation of why she did it.

Unfortunately, the empirical record intimates that reflective activity is probably neither so characteristic of, nor practically important to, human beings as reflectivists suppose it is (p. 22). Many human behaviors are thoughtless, and unconstrained by the deliverances of reflection; on those instances when people do reflect, there is little warrant for confidence that these reflections are informed by accurate self-awareness. If so, there's something seriously wrong with both reflectivism and its corollary (p. x), namely, that if reflectivism is the only philosophically viable account of agency, and the best empirical guess is that reflective agency is not an especially prominent form of human self-direction, there's going to be a shortage of morally responsible agency (p. 152).

The reflectivist might not be dismayed at this prospect; the exercise of agency is an achievement, and it might be that exercises of reflective agency are a scarce achievement (pp. 35–37). Or perhaps reflectivism trades in ideal agency rather than actual agency, and eschews empirical claims (pp. 22, 151). These are defensible positions, but perhaps of dubious appeal. My sensibilities in this respect are conservative (p. 158): I suppose that people (at least in the cultural context where this discussion resides) regularly attribute moral responsibility to one another (whether tacitly or explicitly), and I further suppose that enactments of this familiar practice are not pervasively mistaken. That is, I presume treating the intentional behaviors of normal healthy adults as exercises of morally responsible agency is a morally defensible “default” (pp. 33, 39). So I think the envisaged agency shortage would be the worse for reflectivism, not for the practice (p. 33). But deciding exactly how frequent exercises of agency are, on reflectivism or any other theory, is to enter a kind of discussion I've already said is doomed. Instead, the difficulty is how to rule out the existence of defeaters, and I don't think extant reflectivisms are well situated to do so.

Philosophical targets are famously elusive, and philosophers routinely disavow the views attributed them by critics (pp. 17, 108). (That's not my position, and anyway, your objection to it doesn't work.) It's therefore likely that some reflectivists will deny being committed to views compromised by the arguments I derive from the empirical literature. I disagree, and cite textual evidence for my reading (pp. 17–21) – which ascribes a familiar and intuitive view to my agonists – but this isn't the kind of disagreement that is likely to get amicably resolved.

Rather than playing too long at pin-the-tail-on-the philosopher, it is better to proffer a friendly challenge. Reflectivists typically do not consider in detail how to accommodate the empirical difficulties associated with incongruence (pp. 164, 171), and their theorizing about agency would benefit if they did so. The critical, “targeting” part of the book may be understood not as excoriation, but as invitation – an invitation to help advance the cause of empirically credible theorizing in moral psychology. The positive part of the book can be seen as an attempt to identify one path to such advance, via departing reflectivism. Because I endeavor to avoid unseemly triumphalism (p. 171), I don't deny that there may also be reflectivist means of egress. At this stage of debate, however, it is unclear to me how those means will be developed, or how successful they are likely to be.

4. Experience

Maybe the best “ruling-out” response to skepticism is plain experience (p. 78), in particular the experience of agency: the feeling of doing something (like deciding what to have for dinner) is manifestly distinct from the feeling of having something happen to you (like dinner setting off your digestion). I don't deny such experiences are commonplace, and I further grant that many “agent experiences” (my unlovely name for the feeling of doing) bear a reflectivist character, where it seems to one that one self-consciously and accurately inventories one's inclinations and circumstances, decides what to do on the basis of this inventory, and acts according to one's decision (p. 80).

What I do deny is that agent experiences have the stuff to block skepticism, because such experiences are surprisingly untrustworthy (pp. 96–97). (This is not to say that people are wrong about the experience of agency in the sense of recognizing when their movements are endogenously rather than exogenously generated [Gallagher Reference Gallagher2000]; again, at issue is the more demanding notion of morally responsible agency.) My denial is steeped in the clinical literature on confabulations, the extravagant fabrications produced by patients suffering neurological trauma or psychiatric illness (pp. 81–89). Given many striking demonstrations of failed self-awareness in healthy people, it's commonly suggested that the unafflicted confabulate, much as the afflicted do. The parallels are imperfect, but it is the case that healthy people suffer substantial self-ignorance, and that first-personal reports of self-awareness are of questionable reliability.

Likely implicated in such failures is the well-documented class of phenomena united under the heading “motivated cognition” (pp. 94–96). Although its prevalence is a matter of controversy, and the boundaries between motivational and cognitive processes are muddy, there seems little doubt that motivation powerfully influences cognition, so that reasoning and belief may be determined by extra-epistemological factors (Dawson et al. Reference Dawson, Gilovich and Regan2002, pp. 1379–81; Ditto & Lopez Reference Ditto and Lopez1992; Dunning et al. Reference Dunning, Kunda and Murray1999, p. 79; Gilovich Reference Gilovich1991, pp. 75–87, 84; Kruglanski Reference Kruglanski, Higgins and Kruglanski1996; Kunda Reference Kunda1990, p. 493).

For my purposes, perhaps the most salient variety of motivated cognition is self-enhancement (pp. 92–94): People are prone to inflated assessments of their attributes and performances (Alicke et al. Reference Alicke, Vredenburg, Hiatt and Govorun2001, p. 9; Dufner et al. Reference Dufner, Denissen, Zalk, Matthes, Meeus, van Aken and Sedikides2012, p. 538; Dunning Reference Dunning1999, pp. 5–6; Dunning Reference Dunning2006). For example, undergraduates may think themselves more popular than they really are (Zuckerman & Jost Reference Zuckerman and Jost2001), while their professors may think themselves better teachers than they really are (Blackburn & Clark Reference Blackburn and Clark1975, p. 249; cf. Cross Reference Cross1977).

I interpret the experience of agency as a kind of self-enhancement: people believe they control things they do not, a tendency called the “illusion of control” (pp. 134–36). For example, gamblers commonly seem to think they influence chance events (Davis et al. Reference Davis, Sundahl and Lesbo2000, pp. 1236–37; Toneatto et al. Reference Toneatto, Blitz-Miller, Calderwood, Dragonetti and Tsanos1997, p. 262; cf. Langer Reference Langer1975): in lotteries around the world, quick pick options, where a computer picks random numbers, account for only 10–20% of tickets sold (Simon Reference Simon1998, p. 247). As one player insisted, you shouldn't “trust the computer to pick your numbers. I trust myself more” (as quoted in Farrell et al. Reference Farrell, Cowley and Edwardson2005, p. 597).

It's not just that people exaggerate their control of their circumstances; they overestimate their control of their selves. Here's a patient with hemiplegic anosognosia “explaining” her inability to move her arm (p. 87): “I have never been very ambidextrous”; “I've got severe arthritis in my shoulder”; “Doctor, these medical students have been prodding me all day and I'm sick of it. I don't want to use my left arm” (as quoted in Ramachandran Reference Ramachandran1996, p. 125). The last is especially suggestive, as it replaces patiency with agency: For a person suddenly experiencing the helplessness of the sick role, the performance can be understood as a kind of self-enhancement, albeit a tragically ineffectual one.

I suppose that at least some such confabulations are sincere (p. 84); the anosognosic is not necessarily prevaricating. But whether or not this patient was experiencing agency, her words are – significantly – a self-presentation of agency. As we shall see in a moment, this kind of agential self-presentation is a central feature of human social life.

5. Values

Under the most generic description, my understanding of agency – ponderously titled an anti-reflectivist, valuational, and dialogic (or collaborativist) theory – represents a philosophically commonplace (though not universally endorsed) “compatibilism” (pp. 9–12): I think that moral responsibility is compatible with causal determinism, and people may sometimes be held morally responsible in circumstances where their behavior is determined by factors “external” to themselves. For compatibilists, the important question is not that a behavior was caused but how it was caused: for exercises of morally responsible agency, behavior must be caused in the right way.

Like many other compatibilists, I make a start on “in the right way” with P. F. Strawson's “reactive attitudes”: the motley of emotional and other interpersonal responses – indignation, anger, gratitude, admiration, and the like – with which human beings regulate their social lives (p. 23; Strawson Reference Strawson1962; Vargas Reference Vargas2004; Reference Vargas, Trakakis and Cohen2008; Reference Vargas2013; Watson Reference Watson, Fischer and Ravizza1993). For me, thinking about what (if any) reactive attitudes are apt helps establish whether someone is morally responsible (or not): once again, I shouldn't be angry at the bee who stings me, but it's perfectly appropriate for me to be angry at you if you poke me with a fork in a fit of dinner-hour pique. The reason for the differing reactions, my story goes, is that you are morally responsible for your behavior, and the bee not. The difference in responsibility, as my story goes on, is explained by the observation that you were exercising agency, and the bee not. The aptness of moral responsibility attributions and the associated reactive attitudes are, for behaviors of moral concern, characteristic symptoms of agency.

Next, I say a behavior is an exercise of morally responsible agency when the actor is self-directed while performing it, and further assert that behavior is self-directed when it expresses the actor's values (after Watson Reference Watson1975; Reference Watson1996; cf. Bratman Reference Bratman2007, p. 48; Smith Reference Smith2005; Sripada Reference Sripada2015a; Reference Sripada2015b). While this particular valuational theory is only one of numerous contending theories, and hardly the object of philosophical consensus, it does enable some pretty plausible observations: when the nicotine addict guiltily succumbs to craving and lights up, his behavior is not self-directed, but when he manages to resist a craving because he values his health, his behavior is self-directed.

For me then, morally responsible agency gets understood in terms of self-direction, and self-direction gets understood in terms of expressing values: behaviors are exercises of agency when they are expressions of the actor's values. In turn, values get understood in terms of desires (Bratman Reference Bratman2007, pp. 47–67; Harman Reference Harman2000, p. 135) – desires possessed of a reasonably reliable and substantial motivational force. Not just any such desires: the desires properly associated with value are those desires the actor accepts in a determinative role for her practical planning (Bratman Reference Bratman2007, pp. 64–66). For a desire to be associated with a value, it must also have a justificatory role: it must be something that the planner is amenable to employing in justification or defense of her plan.

The planner, to my way of thinking, need not be aware of her willingness to assign this justificatory role (pp. 27–28). People may have desires, values, and plans that they are quite unaware of, and their behavior may express their values without their knowing that it does so. (Consult locutions like, “I guess that was my plan all along,” and “I suppose this must be what I really wanted.”) This is a crucial feature of my account, because it makes room for agency in the absence of accurate reflection. If I speak out of cruelty while thinking I'm speaking out of honesty, you may well hold me responsible for hurting your feelings. And if pressed to defend your attribution, you might say something like, “Whatever he thought he was doing, he really wanted to be cruel.” Pretty good reason to attribute an exercise of agency, where someone acts on what matters to her. The less likely thing would be to say that someone did what mattered to her, because it mattered to her, but didn't do so as a morally responsible agent (pp. 160–61).

This provides occasion to make a bit clearer what the argument is not about (p. 69). It's commonly supposed that much behavior lacking “conscious control” makes trouble for agency (e.g., Levy Reference Levy2011, pp. 188–94; Wegner Reference Wegner2002, pp. 156–58, 170–71; Reference Wegner, Hassin, Uleman and Bargh2005, p. 28). But this supposition immediately encounters inconvenience, for it apparently excludes far too much from the offices of agency. Acts of kindness or callousness that are done habitually, with little in the way of conscious supervision, are customarily treated, and perfectly appropriately so, as exercises of morally responsible agency. The point generalizes, and widely. While there's controversy concerning the role of conscious control in skilled behavior (Christensen et al. Reference Christensen, Sutton and McIlwain2016; Montero Reference Montero2010; Noë Reference Noë2012, Ch. 6), surely many skilled behaviors are performed unreflectively (how else could they be done so quickly and effortlessly?), and just as surely, people are often credited – appropriately so – for these performances.

Adopting a valuational theory of agency like the one I adopt is not necessarily to reject reflectivism; a valuational theory might be fashioned as a reflectivist theory, as is, perhaps, Bratman's (Reference Bratman2007 p. 28, n. 5). Rather, the dispute concerns what psychological processes facilitate the exercise of agency. The reflectivist taps self-conscious, tolerably accurate reflection in this role, while I do not require accurate reflection for the exercise of agency. And this difference, I claim, leaves me better situated than the reflectivist to ameliorate agency skepticism.

6. Ignorance

Ignorance remains a paradigmatic excuse, which may block the attribution of responsibility (without Horton's big ears, the other critters couldn't have known that they were endangering the dust speck world of Whoville). But self-ignorance, such as self-enhancement and illusions of control, doesn't necessarily undermine morally responsible agency, and may often enable it (pp. 129, 136, 144, 158).

One important pathway by which self-ignorance may facilitate agency is motivational (pp. 136–37, 144). Turns out, writing this précis was nearly impossible for me, which seems a bit odd: why should it be hard to summarize a book you've already finished? A likely explanation notes that the book is muddled, and the author of modest ability. But this path leads to despair; if I take the elucidation seriously, I'll abandon my labors for the gym, a nap, or worse. On the other hand, if I attribute my troubles in summarizing to the profundity of my topic and the intricacy of my thought, I may be hopeful enough to fight, and write, another day. The point extends: plodding Professor Drudge might better be able to grind out long hours at his desk under the auspices of “talented” than “mediocre,” and given the contribution of perspiration to inspiration, this understanding can help Drudge produce work of sufficient quality to earn him the title “talented” – his mediocrity notwithstanding (p. 144). If Drudge values his professional status, his positivity may facilitate his agency.

Another example: unrealistic positivity has been implicated in improved health outcomes (Taylor et al. Reference Taylor, Lerner, Sherman, Sage and McDowell2003), and it appears that this self-enhancement extends to perceptions of control. In a study of cancer patients, perceptions of control were negatively associated with maladjustment; patients with higher perceptions of control were less likely to experience anxiety and depression (Thompson et al. Reference Thompson, Sobolew-Shubin, Galbraith, Schwankovsky and Cruzen1993, pp. 297–298). Moreover, there was a larger negative relationship between perceptions of control and maladjustment for those rated lower in physical functionality (Thompson et al. Reference Thompson, Sobolew-Shubin, Galbraith, Schwankovsky and Cruzen1993, pp. 299–300); perceived control may have been doing more good for those who had less control of their circumstances. Given the abundant evidence implicating psychological well-being in physical well-being, and psychological distress in physical distress (Brenner Reference Brenner1979; Diener & Chan Reference Diener and Chan2011; Steptoe et al. Reference Steptoe, Wardle and Marmot2005; Veenhoven Reference Veenhoven1988), it becomes very tempting to say that elevated perceptions of control, in so far as they're associated with better adjustment, are good for one's health.

Sick or well, valuing my health is presumably part of the reason I strive to promote it. Perhaps my beliefs about the efficacy of these efforts are motivated cognitions; I may think my labors more effective than they are, in part because I want them to succeed. But these motivated cognitions may simultaneously be motivating cognitions; my believing my efforts will work inspires me to undertake work that in fact work, even if they don't work so well as I believe (pp. 135–37). Here, self-ignorance, via motivational pathways, promotes the exercise of agency.

A similar story can be told for romantic relationships (136–37). According to one group of marital researchers (Fowers et al Reference Fowers, Lyons, Montel and Shaked2001, pp. 96–99, 102, 105), the presence of positive illusions is “nearly universal” among “satisfied spouses”; the mean estimate of divorce given by members of married couples was 10%, and the modal estimate was zero%, while scientific estimates of its likelihood are often in the range of 40–60% (Fowers et al. Reference Fowers, Lyons, Montel and Shaked2001, p. 105; for more on “marital optimism,” see Baker & Emery Reference Baker and Emery1993; Boyer-Pennington et al. Reference Boyer-Pennington, Pennington and Spink2001). Hand wringing over the “50% divorce rate” is, as any reader of Divorce Magazine (2004) can tell you, a prominent component of public discourse in the United States; apparently, romantic illusions may persist in the face a well-known body of undermining fact (difficulty in estimating divorce rates duly noted).

At the same time, relationship outcomes may be sensitive to effort: numerous studies indicate that couples counseling is effective (Bray & Jouriles Reference Bray and Jouriles1995; Hahlweg & Richtera Reference Hahlweg and Richtera2010; Sayers et al. Reference Sayers, Kohn and Heavey1998; Snyder et al. Reference Snyder, Castellani and Whisman2006). During relationship trouble, I'm willing to speculate, sturdy perceptions of control have motivational utility: if you don't think the quality of your relationship is responsive to effort, how do you get yourself to undertake the effort? But if you think couples counseling can help you walk it back from the brink, could be you, yours, and your therapist will put the needed work in.

By supporting value-conducive motives, illusions of control may facilitate behavior that helps realize values. Self-ignorance often functions to effect self-direction, and its absence can be an impediment to agency: a complete and accurate understanding of your career, health, or relationship prospects might prevent you from making them all you want them to be. If so, we've identified a pathway whereby self-ignorance supports, rather than impairs, agency. This pathway is often indirect (p. 127): falsely believing I can directly effect a valued outcome may support motivation eventuating in behaviors or circumstances that do, in fact, effect the outcome in question. The fact, we can say, is the child of the fiction (pp. 136–37).

7. Collaboration

To fill out this story, I understand self-ignorance in the context of sociality. In doing so, I identify a secondary agonist, individualism, which maintains that optimal human reasoning is exemplified by individual thinkers (pp. 103, 107–109). Against this, I pit collaborativism about rationality, where optimal human reasoning is held to be “socially embedded,” and then extend this collaborativism to agency, where many important exercises of individual agency are substantially social phenomena (pp. 103, 115, 122).

Start with simple examples (pp. 123–24): seatbelt laws have been found to increase seatbelt use (Gantz & Henkle Reference Gantz and Henkle2002; Shults et al. Reference Shults, Elder, Sleet, Thompson and Nichols2004), while public health campaigns have been found to decrease rates of smoking (Fiore et al. Reference Fiore, Bailey, Cohen, Dorfman, Goldstein, Gritz, Heyman, Jaen, Kottke, Lando and Mecklenburg2000). Assuming that people value their lives and health, these are cases where social processes enable people to better express their values in their conduct. But while there is little doubt that postponing death and disability is (for most people) a valued outcome, one might wonder if realizing these outcomes should count as an exercise of agency. The mechanism matters, and succumbing to media manipulation or yielding to government coercion may not seem appropriately agential.

Perhaps psychotherapy makes a more convincing illustration (pp. 124–25). People seek out psychotherapy in hopes of making their lives go better, so this active process has a more agential appearance than the succumbing to media campaigns or state regulations. Of course, people often enter therapy in response to something like duress: a stalled career, a strife-torn marriage, or stacks of unpaid bills. But these incentives are not obviously inimical to agency. On the contrary, people trying to change their lives for the better is an excellent place to look for agency, and psychotherapy can help effect this change.

Outcome studies, using both clinician assessment and client self-report, indicate that talk therapy works – it can ameliorate various adverse psychological conditions, such as depression and anxiety (Lambert & Ogles Reference Lambert, Ogles and Lambert2004; Luborsky et al. Reference Luborsky, McLellan, Woody, O'Brien and Auerbach1985, p. 609; Seligman Reference Seligman1993). We've now an appealing example of collaborativism about agency. First, the “talking cure” is very much a social treatment, where client and therapist work things through together more effectively than the client could do on their own. There's the collaborativism. Second, decreasing psychological discomfort and increasing personal efficacy are very likely values many clients in therapy hold, so the clinical process is reasonably thought to facilitate the expression of these values. There's the agency.

I contend that the success of this endeavor does not require accurate self-awareness on the part of the client. (Or, perhaps, accurate awareness of the client by the therapist; interestingly, the rubric under which therapy is conducted is not a critical determinant of clinical efficacy [Brown et al. Reference Brown, Dreis, Nace, Hubble, Duncan and Miller1999; Dawes Reference Dawes and Dawes1994, pp. 38–74; Luborsky & Singer Reference Luborsky and Singer1975; Wampold et al. Reference Wampold, Mondin, Moody, Stich, Benson and Ahn1997; Woolfolk Reference Woolfolk1998].) At the same time, a recurring theme in the clinical literature is that a “positive alliance” between therapist and client is associated with successful outcomes (Horvath & Symonds Reference Horvath and Symonds1991; Krupnick et al. Reference Krupnick, Sotsky, Simmens, Moyher, Elkin, Watkins and Pilkonis1996; Martin et al. Reference Martin, Garske and Davis2000; Orlinsky et al. Reference Orlinsky, Ronnestad, Willutzki and Lambert2004). For many consumers of psychotherapy, I'm guessing that comes as a relief: if you always end up talking to your therapist about fluff like television or sports instead of the deepest workings of your soul, you may yet be doing yourself good, so long as you're bonding with your therapist.Footnote ⁴ Supposing therapy can facilitate agency, as I've just suggested, we've here a case where agency is achieved without accurate self-awareness. In this instance, collaborativism and anti-reflectivism are complementary.

8. Rationalization

To better understand the synergistic contribution of self-ignorance and collaboration to morally responsible agency, consider Johansson, Hall, and colleagues' incredible studies of “choice blindness” (Hall et al. Reference Hall, Johansson, Tärning, Sikström and Deutgen2010; Johansson et al. Reference Johansson, Hall, Sikström, Tärning and Lind2006), where people fluently provide agential explanations for choices they didn't make (pp. 138–40). In Sweden, Hall et al. (Reference Hall, Johansson and Strandberg2012; cf. Hall et al. Reference Hall, Strandberg, Pärnamets, Lind, Tärning and Johansson2013) demonstrated choice blindness for moral and political attitudes. People strolling through a park were given a twelve-item survey with statements concerning either general moral principles or current moral issues, and asked to report their attitudes on a 9-point scale anchored at “completely agree” and “completely disagree.” After completing the survey, participants read aloud three of the statements they had responded to, and explained their positions. In manipulated trials, two of these statements were reversed: if someone originally agreed with “Even if an action might harm the innocent, it can still be morally permissible to perform it,” it now appeared that they had agreed to “If an action might harm the innocent, then it is not morally permissible to perform it” (emphasis added).

The reversal was noticed in only 47% of trails, and 69% of participants accepted at least one reversed statement. Although the politically active were more likely to correct reversals, people claiming to generally hold strong moral opinions weren't more likely to make corrections. Unsurprisingly, level of agreement was associated with correction: The more participants agreed or disagreed with a statement, the more likely they were to correct the reversal. But nearly a third (31.4%) of all manipulated trials with answers at the endpoints of the scale (1 or 9) were not corrected. And remarkably, when it came to explaining manipulated choices, 53% of the participants argued unequivocally for the reversal of their original position. People are able to quite assuredly justify and explain their choices – even when “their” choices are not choices they made!

While folks sometimes hold their nose and knowingly “go along to get along,” the social influence in choice blindness studies most probably proceeds subliminally. Otherwise it would be difficult to explain the fluidity with which participants explained their (non-) choices (p. 138; Johansson et al. Reference Johansson, Hall, Sikström, Tärning and Lind2006), as well as the apparently genuine surprise participants evince in post-experimental debriefing (Hall et al. Reference Hall, Johansson, Tärning, Sikström and Deutgen2010). But while the pseudo-explanations weren't conscious social niceties, they have something importantly in common with social niceties: such explanations are required by convention. It can be socially awkward to stand mute when questioned about the reasons for one's political or moral convictions; commonly, keeping silent or pleading ignorance won't do, and most any answer, or at least a wide range of answers, is better than not answering (Hirstein Reference Hirstein2005, pp. 4–5).

Not so much for garden-variety factual ignorance; fine to say that I don't remember the name of the 23rd U.S. president, or the first woman to summit Everest. But where reasons are required, not so much. I suspect that instances of “rational dumbfounding” (pp. 140–41) where people are unable to explain their behavior – particularly their intentional behavior – are pretty unusual. Even if rational dumbfounding is more common than I suspect, confessing rational dumbfounding still seems remarkable. People fluently produce socially serviceable explanations for what they do, even when their self-ignorance extends to the psychological origins of their behavior.

I call these performances rationalizations rather than confabulations, to distinguish them from clinical confabulation, and I deploy “rationalization” in a non-pejorative sense, absent any connotations of bad faith (pp. 141–43). Here, a rationalization is a (typically verbal) performance that presents judgment and behavior as rational. (Or, slightly less circularly, rationalizations make judgment and behavior make sense; cf. Gibbard Reference Gibbard1990, pp. 37–38, 156–59). If I'm right, a central form of rationalization presents a behavior as an exercise of agency: I chose to do so. I meant to do that. I had my reasons. Frequently, rationalizations may reflect illusions of control: people present themselves as achieving exercises of agency even when they have not.

My account favors approaches that construe agency as structured by narrative (Doris Reference Doris2015b, pp. 143–46; see also Dennett Reference Dennett and Dennett1991; Reference Dennett, Kessel, Cole and Johnson1992; Fischer Reference Fischer2006, pp. 106–23; Reference Fischer2009, pp. 145–77; Schechtman Reference Schechtman1996 Reference Schechtman and Gallagher2011; Velleman Reference Velleman2006; pace Strawson Reference Strawson2004). Who people are, and what they do, is shaped by the self-depictions, which I call biographies, they express to themselves and others. That one understands one has made a promise, and further understands oneself to be a person who honors her promises, may help ensure that the promise is kept. One's biography can secure behavior expressing one's values, even in the face of unfavorable circumstance or instable inclination. Thinking – and talking – of oneself as a dutiful promiser may be causally implicated in one keeping promises on those occasions that one does so, whether one is a dutiful promiser or not. Then to do its work biography needn't be accurate, so long as it is motivationally engaging for the teller.

Biographies may be private; I might live by a story I tell only to myself. But biographies are also presented socially, and serve as vehicles for the exchange of rationalizations. It's because I've had the life I've had – or seem to have had – that I'm justified in doing as I do. The trauma of experiencing a near-fatal automobile accident, for example, may justify my disinclination to see a movie prominently featuring car chases, and prompt you to propose seeing a less bombastic film. Human beings develop rationalizations collaboratively, and the central requirement for these rationalizations is not accuracy, but accord with one's interlocutors. People shape their lives, not as isolated reflectors, but as participants in an ongoing negotiation – a negotiation that simultaneously constrains and expresses who they are. In a slogan, agents are negotiations. If you like, call this notion of agency dialogic (p. 148). Here morally responsible agency requires not freedom from influence, but mutual influence, as individuals express their values in a collaborative process.

9. Skepticism (again)

I have an account of how morally responsible agency may be exercised, but ameliorating the skeptical problem also requires an account of how exercises of morally responsible agency are detected (pp. 159–64). The skeptic says we are never justified in attributing moral responsibility; in retort, the anti-skeptic must articulate conditions when it is justifiable to do so. (Reminder: the relevant notion of justification is substantially ethical.)

According to my valuational approach, archetypal exercises of morally responsible agency are expressions of the actor's values, so the problem is determining whether the actor's values are expressed in their conduct. To do so, on my understanding of the skeptical challenge, we must be justified in ruling out the presence of defeaters; when we have done so, we may have the requisite moral confidence that the relevant conduct expresses the actor's values, and may therefore be justified in attributing an exercise of morally responsible agency (p. 159).

In preparation for this work, we ought to realize that valuing has a temporal dimension: values are expressed over time, and often can only be identified over time (pp. 162–63). It will frequently be difficult to determine whether someone holds a value, in the absence of temporally extended trends in cognition, rationalization, and behavior. But with extended observation, a pattern of symptoms may emerge: as cognitions, rationalizations, and behaviors appropriate to a value tend to recur in a person's life, we, and they, may begin to have confidence that a person holds that value, and that particular behaviors, patterns of behavior, or life projects are expressing it (even where their behavior is less than consistent with respect to that value, as it very probably will be, if “situationism” about moral personality and behavior [Doris Reference Doris2002] is anything close to correct).

Conversely, if one focuses on isolated events, diagnosis may falter. (The same is true of medical diagnosis; that's why your doctor takes your history, or holds you overnight for observation.) It will frequently be obscure whether someone doing something is an expression of her values; the evaluative signal may be quite weak, against the background of situational noise.

In tracking the evaluative signal, sociality will be central; first-personal inquiry will frequently be augmented by second personal, collaborative, inquiry: a friend observes that I'm often downcast after a day at my “dream job,” or an old lover remarks that you seem much happier with your new partner (p. 163). Collaboration also occurs in institutional contexts: therapeutic and educational endeavors can help people figure out what matters to them. Sometimes, others have better access – or at least instructively different access – to a person's values than does the person herself. The extent of agency-impairing self-ignorance is considerable, but people are collectively possessed of epistemic assets fit to ameliorate it, assets deployed in the continuing social negotiation by which people order and make sense of their lives.

Attribution of agency and responsibility may be warranted when a pattern of cognition, rationalization, and behavior emerges, and that pattern is best explained as involving the expression of some value. Determining whether a particular action expresses a value, in the sense of being governed by a value relevant goal, as opposed to fortuitously conforming to that value (pp. 25–26), will very frequently make difficult work. But the emergence of trends across iterated cognitions and behaviors can underwrite confidence that the trend is to be accounted for by reference to a person's values, rather than a massively coincidental run of defeaters. Typically, the required evidence base must be both wide, covering multiple observations of behavior, and deep, licensing inference to the psychological states implicated in the behaviors. In such cases, the presence of defeaters need not be treated as a live possibility, which means that in such cases the skeptical challenge is defanged.Footnote ⁵

These welcome conditions may obtain less often than one might wish; given the vagaries of mind and world, defensibly attributing exercises of morally responsible agency, on any plausible theory, takes hard work. But if one focuses on isolated cases of reflective deliberation, as the reflectivist is wont to do, the work is much harder (p. 164). There, the possibility of defeaters – given all that is known about the potential for rationally arbitrary influences on judgment and decision – cannot usually be ruled out with confidence sufficient to warrant attribution of responsibility. Reflectivist paradigms of morally responsible agency, then, are epistemically challenged.

Reflectivists may insist they can accommodate my advice to depart emphasis on atomistic behaviors in favor of extended processes. Fair enough. But there's not only the problem of how agency is identified, there's also the problem of how agency is facilitated (p. 164). A compelling response to the skeptic will provide standards for attribution of agency, and an account of how agency may be realized in human lives – lives afflicted with surprisingly high levels of self-ignorance. Articulating standards will be cold comfort, without an account of how people may live up to those standards. A dialogic understanding of agency offers one such account – an explanation of how agency emerges in the face of limited self-awareness, through a process of collaborative negotiation. If accurate self-conscious reflection is required for morally responsible agency, the prospects for agency look rather worse.

Supposing I've now decent answers to the problems of how exercises of morally responsible agency are facilitated and identified, another question remains (pp. 164–66): does my approach to responsibility capture the “normative” character of responsibility related discourse and practice? It's widely accepted that normativity has something to do with the guidance of thought and behavior: as opposed to descriptive questions about how the world is, normative questions are prescriptive questions concerning what ought be done about it. Normative discourse, then, is oughty discourse. If so, a theory of responsibility should explain why attributions of responsibility (and their denial) carry imperatives about how the subjects of such attributions should be regarded and treated.

Accounts of responsibility such as mine, centered on reactive attitudes, capture something of normativity rather easily. The various reactive attitudes may make a motley assortment, but whatever else they are, they often involve emotions. Emotions are standardly thought to involve “action tendencies” (Nichols Reference Nichols2007, pp. 412–414), and even where they don't immediately move people to action, they prepare people for action: emotions structure the range of behavioral options (Prinz Reference Prinz2004, pp. 191–96).

To suffer an emotion is to be told what to do, or not do. But emotional imperatives don't carry all of the normativity that might be desired. For people can, and do, ask whether their emotions are appropriate, justified, or fitting (D'Arms & Jacobson Reference D'Arms and Jacobson2000; Reference D'Arms and Jacobson2006). In so doing, they're asking about what is helpfully called normative authority (Railton Reference Railton2003, p. 344): why should a command issued by emotion command my assent?

When responsibility is understood as I understand it, by way of reactive attitudes, the challenge is to identify compelling theoretical grounds for when and what reactive attitudes are appropriate. On my theory, responsibility is associated with the exercise of agency, and the exercise of agency with expressions of values, so the question becomes whether these expressions are appropriate targets for the reactive attitudes.

I think this question is readily answered: when someone's deeds manifest their values, it makes good sense to direct anger or admiration their way. I'm angry with the Wall Street Oligarch who orders a million-dollar renovation for his office ($1,400 wastebasket included) as his company fails and the economy falters, because I think he values status too much and humanity too little. I admire the man who donates 10% of his $1,000 monthly disability check to charity, because I think he's moved by the opposite complex of values. And were I pressed to justify my reactions, I could make a convincing case on the grounds of what matters, and fails to matter, to each man. The point might be put in the language of desert (Doris Reference Doris2015a; Vargas Reference Vargas2013, pp. 234–66): reference to each man's values explains why they deserve the attitudes I subject them to.

In associating responsibility attribution with emotionally infused reactive attitudes, I find something of the oughtiness associated with normativity. And by locating agency in the expression of values, I've located a perspicuous rationale for these reactive attitudes: there's pretty good reason for you to be angry with me for what I did, if what I did is a function of my mean-spirited matterings. This, it seems to me, is an account possessed of sufficient normative authority for the discourse and practice of responsibility attribution.

I don't insist my way is the only way. Because I am a pluralist about agency and responsibility, I allow that morally responsible agency may be exercised in other ways (pp. 171–77). But one of the ways people sometimes exercise agency, the way envisaged by the reflectivist, has been seriously overemphasized by philosophers, with the result that many philosophical theories are poorly situated to accommodate incongruence.

10. Selves

If one is overly ambitious, as I confess to being, one can extend valuational and dialogic perspectives on morally responsible agency to two notorious “problems of the self” (pp. 5–9, 179–97): continuity – what is required for a person to survive changes? – and identity, what distinguishes one person from another? (Philosophers use “identity” in both of these contexts; I've departed this dual usage by using “continuity” in the context of survival.) Actually, joining these two problems to the problem of agency is, however ambitious it may be, altogether necessary. For if persons are not entities persisting over time while possessed of relatively determinate identities, it is obscure on what responsibility is to affix.

To my thinking, as the problems are related, as is the solution: in completing my theory, I extend the collaborativist, valuational, dialogic approach to continuity and identity. I begin this extension by proposing that continuity is socially contingent (p. 182): personal continuity is predicated on psychological continuity, and psychological continuity is sustained by societal continuity, so if a perturbation in circumstances is substantial enough, as in cases of cultural devastation like those associated with the North American genocide (pp. 178–80), personal continuity may be compromised, physical survival notwithstanding.

My proposal seemingly conflicts with a standard dictum in the philosophical literature on personal identity: extrinsic factors don't count (p. 182). According to this dictum, whether a person at one time bears an “identity relationship” to a person at another depends only on facts about the two “person stages” and the relations between them (Noonan Reference Noonan1989, p. 152). Your survival has to do with you, and not your neighbors, friends, or family: a neighbor dying, or even a lot of neighbors dying, doesn't mean you've failed to survive.

Yet a moment's thought reveals that lots of stuff – within your skin and without – has to do with you and your survival. Given that persons are entwined in causal webs with strands trailing far beyond their body, it's hard to say what's intrinsic to a person and what's extrinsic. It's nevertheless obvious that cultural conditions are causally related – powerfully and pervasively so – with the psychology of those associated with them. One can therefore conclude that culture may affect personal continuity, and is not excluded from consideration by any plausible “no extrinsic factors” principle. Even if you're more confident than I in the existence of a tenable extrinsic/intrinsic divide, you shouldn't want to deny that identity-intrinsic factors may be causally impacted by social and cultural perturbations.

Here's one avenue for such impact: personal continuity varies with evaluative continuity (p. 183). When a person's values change, they become less like the person they were before; at the extreme, if a person's values are completely changed, I contend, they are no longer the same person. Culture is an important determinant of value, and cultures ebb and flow; so too does value. Typically, in cultural change, circumstances are mixed: some practices go on, and others falter; some values persist, and others fade away. But it remains the case that when sufficient cultural upheaval occurs, there may be disruptions of the practices required to support central values that are substantial enough to press questions about psychological continuity – and thereby personal survival. Answers to these questions will seldom be all or nothing; although change is a constant, significant evaluative continuity is likely the rule rather than the exception. Yet this point seems secure: factors well beyond the skin matter for continuity.

Explicit discussions of identity are less visible in mainstream “analytic” philosophy than discussions of continuity, but we can make a start with Taylor (Reference Taylor and Gutmann1994, p. 25), who describes identity as “something like a person's understanding of who they are, of their fundamental defining characteristics as a human being.” Terms like “fundamental” and “defining” are, unfortunately, not easily subdued. However, a hopeful opening gambit proceeds in terms of individuation (p. 187): one's defining characteristics are the constellation of attributes by virtue of which one is different from other people – even while one shares many attributes with other people (Appiah Reference Appiah2010, Ch. 1; Taylor Reference Taylor and Gutmann1994, p. 28).

I understand individuation, as I understand agency and continuity, by reference to value (p. 188). Identity may be a source of value: a ritual observance might be of value to me by virtue of my identity as a member of a certain religion, but valueless to those who do not share my faith. Identity may also be the object of value: someone might value her profession, ethnicity, or political affiliation. Most important here is that value may be a source of identity (the relation between identity and value is bi-directional). A person has the identity they do partly by virtue of having the values they do: Who I am has much to do with what matters to me.

If I'm right, a dialogic valuational theory can be crafted into a comprehensive theory of the self, because it accounts for notions central to philosophical thinking on selves (pp. 6–9): agency, continuity, and identity. The theory can also explain how agency, continuity, and identity emerge in actual human lives, by way of an ongoing process of collaborative rationalization. This process develops and sustains the self, despite the limits of reflection and the infirmities of self-awareness. It also provides material by which to answer skepticism about morally responsible agency, because the collaborative process facilitates identifying a person's values, and the role of these values in their behavior. That's a lot to get done! Hopefully, Talking to Our Selves makes a decent start on doing it.Footnote ⁶