Blame: It’s nature, targets, and functions


Blame: Its nature, targets, and functions.


This paper asks three questions: (1) What is holding-responsible, particularly blaming, as an expressive practice? (2) Who is an appropriate target of blame? And (3) What is the function of blame? I answer that (i) Blaming is an expressive practice, in which (somewhat surprisingly perhaps) blame can be expressed unconsciously by the blamer; (ii) the appropriate target of the reactive attitudes (resentment, disapprobation, indignation) is the wrongdoer, but the appropriate target of the conative force of blame – its demand for a response – is a suitably sensitive audience, not necessarily the wrongdoer; and (iii) one of blame’s primary functions is to enhance relational equality, because relational equality is a precondition for a fair and rational blaming practice.

Key words: responsibility; blame; reactive attitudes; implicit bias; eliminitivism

  1. Introduction

In this paper, I ask three questions: (1) What is holding-responsible, particularly blaming, as an expressive practice? (2) Who is an appropriate target of blame? And (3) What is the function of blame? (What is it for? Why do we do it? Why should we do it?).

I come to the following conclusions.

(1) Responsibility is an expression of the reactive attitudes, which, on various proposals, have emotional contents (emotionalism), cognitive contents (cognitivism), or both. I show that, because emotions and (cognitive) judgments can, in some cases, be unconscious, blame qua emotional/cognitive response can be unconscious. This is not necessarily a refutation of extant theories, but an under-appreciated implication of neuroscientific data on the nature of emotions, judgments, and implicit states implicated in emotional and cognitive activations. Although there is an abundance of scholarship on responsibility for unconscious states, very little has been published on unconscious blame, even though this type of blame is extremely consequential.

(2) Although Strawson says that moral incompetents are not apt targets of the reactive attitudes, I argue that moral incompetents can be apt targets of the emotional and cognitive contents of blame – resentment, disapprobation, negative valuation – but not apt targets of the conative orientation of, or demand implicit in, blame. Instead, the proper target of blame’s demands and solicitations is a suitably sensitive audience. Thus, blame’s emotional and cognitive contents, on the one hand, and blame’s conative orientation, on the other, come apart, and moral incompetents are only partially outside the scope of the reactive attitudes. This view vindicates contemporary theories on which blame can be fittingly addressed to the community (e.g., Bell 2014).

(3) Blame is most likely multi-functional, because versatile (multi-directional) blaming practices have more traction, potentially motivating more people to heed its demands. That said, surely one of the most urgent and immediate functions of our blaming practice is to enhance relational equality, or equality of moral and epistemic esteem and standing (Anderson 2015). The reason is that relational equality is a prerequisite to other posited aims – e.g., registering one’s protest against wrongdoing (Smith 2013), resisting wrongdoing (Bell 2013) – since our perception of wrongdoing is distorted by implicit biases rooted in pervasive cultural stereotypes. So, relational equality must be one of blame’s utmost normative functions – the function that it should, and must, serve, if it is to function fairly and rationally on a large-scale interpersonal level.

  1. Responsibility as an expressive practice: What is ‘an expression’?

One of the most influential accounts of moral responsibility in philosophy is Strawson’s view (1963), on which responsibility is an interpersonal practice in which we express the reactive attitudes, such as blame, praise, approbation, disapprobation, resentment, and so on. For this paper, I will focus on the negative reactive attitudes, particularly blame.

Many contemporary theorists construe Strawsonian holding-responsible as an expressive or communicative exchange, in which an agent expresses a reactive attitude toward another agent in response to the target’s quality of will (or other morally salient quality). McKenna (2014), for example, describes blame as part of a conversational practice with three paradigmatic stages: (1) agent A performs a morally problematic action, (2) agent B expresses a negative reactive attitude to A, and (3) A responds to B. In an ideal conversational exchange, A’s expression and B’s response will be fitting, meaning, in McKenna’s terms, intelligible to the recipient and appropriate to the context. If the wrongdoer is morally incompetent, she is not an apt target for blame.

Other theorists have different ideas about the expressive function of blame. Angela Smith argues that blame is an expressive practice that registers one’s protest against someone’s morally problematic treatment of oneself or others (2013). Macalister Bell holds that blame expresses a hostile attitude toward others, as a means of resisting wrongdoing (2013). She then identifies a number of sub-functions that subserve the resistance aim: registering the damage done to our relationships, educating its target, motivating offenders to avoid recidivism, educating members of the moral community, and expressing or affirming one’s own values. Christopher Bennett describes blame as a way of expressing disapproval through symbolic gestures that register the offender’s damage to important relationships (2013). Victoria McGeer describes blame as an expressive practice that produces prosocial behaviours by drawing attention to past offenses (2013: 171). These accounts differ in their construal of the function that expressions of blame paradigmatically serve, but they agree with the proposition that blame is an expressive practice.

Let’s assume that the expressivist view is right. At this juncture, we can ask: what does it mean to express blame?

This is a germane question because there is disagreement about what cognitive states are implicated in holding-responsible, and thus what cognitive states blame can express. Many expressivists think that emotions are a “canonical,” if not necessary, part of our responsibility practice (McGeer 2013: 168). McGeer argues that creatures like us, with our evolutionary history, cannot entirely expunge emotions, including ‘uncivil’ emotions, from blame. She sees blame as characteristically emotionally-charged, as well as “rapid and involuntary” (2013: 172), though (possibly) reflectively mediated and thus ‘domesticatable.’ Many people would urge that holding-responsible also implicates judgments or cognitive contents, which mediate (and potentially ‘domesticate’) our knee-jerk, affective responses (e.g., Smith 2013; Sher 2010; McKenna 2013). If we grant that reactive attitudes typically implicate emotional and/or cognitive contents, then (in either case) we might be tempted to assume that they are typically conscious, or at least consciously available. We know when we are blaming someone.  

(Many theorists hold that the reactive attitudes also have a ‘conative orientation,’ or an intention to produce some type of change or response in another person. I will hold off on discussing the conative view until the next section, where it will be more relevant).

In the first place, one might assume, as Williams James did, that an ‘unconscious emotion’ is a contradiction in terms (1884) – that there is a tangible ‘feeling-that’ quality to emotional states. On this assumption, unconscious blame is impossible. One might assume, too, that judgments are reflective in nature – that they involve a conscious representation of their object. If these assumptions are right, then perhaps we can be temporarily unaware of our blaming attitudes due to inattention or a transient distraction, but these states are nonetheless available to consciousness.

Here are some typical examples of blame that seem to affirm this picture. Someone loses confidence in his friend Joe after learning that Joe ridiculed him behind his back, and he modifies the relationship accordingly (Bennett 2012, citing Scanlon 2008). Isa blames her friend Marie for accepting a disempowering relationship with an abusive man (Bell 2014). A college students posts a ‘NO TOUCHY’ post-it on her pizza box to deter her friend from eating the leftovers, exemplifying ‘anticipatory blame’ (MacNamara 2014: 160). In all of these examples, unless the blamer is acting in a somnambulant state, we must imagine the person consciously blaming the target, i.e., expressing consciously-felt emotions and (most likely) consciously-represented judgments. For instance, Isa, if questioned, would probably say that she reproaches Marie, and disapproves of Marie’s decision to stay with her partner. She has conscious access to her blame-constitutive emotional and cognitive states.

Emotions and judgments, however, are not necessarily available to consciousness. (I’ll discuss emotions first and judgments shortly thereafter). Mounting neuroscientific evidence supports the existence of unconscious/implicit emotions, or states with the neurocognitive profile of an emotion that are not available to conscious awareness. Smith and Lane offer as evidence “emotionally relevant stimuli presented so briefly that perceptual awareness is not possible, [which nevertheless] reliably influence preferences (Zajonc, 1980), consummatory behavior (Winkielman and Berridge, 2004), and… other emotion-related physiological/ behavioral reactions (Tamietto and de Gelder, 2010)” (2016: 17). A growing literature, they add, “has also established that unconscious or implicit attitudes and beliefs have a profound effect on social behavior (McConnell and Leibold, 2001)” (2016: 217) – they are causally efficacious. Joseph LeDoux, a neuroscientist and writer, has argued that the activation of emotional circuits can trigger overt behaviour without giving rise to any conscious feeling, if those activations do not interact with neural systems implicated in conscious processing (2012). These activations are called ‘implicit emotions.’ To give some concrete examples, research subjects exhibit facial expressions, physiological arousal, and afferent feedback constitutive of fear, without subjectively representing fear (Khilstrom et al. 2000); in dissociation patients, conscious awareness of an emotion can be severed from conscious processes, rendering the emotion subjectively opaque (Bucci 2016); in patients with chronic health problems, activation in emotional circuits often presents as somatic symptoms with no conscious neural correlates – no consciousness of the emotion (Kroenke 2003; Konnopka et al., 2012; Sharpe and Carson, 2001).

These may not be canonical cases of emotions, but they are still valid cases, and these unconscious emotional activations can motivate overt behaviour. If there can be motivationally efficacious unconscious emotions, there can, on an affective view, be unconscious blame. For instance, if I implicitly fear, resent, or experience somatic distress in response to someone’s salient attributes, I may express those unconscious neurocognitive states toward their object (in negative ways fitting of the emotion) without realizing it. That is, I may (affectively) unconsciously blame the person. This doesn’t mean that unconscious blame is the norm, but it can happen.

Judgments can also be unconscious. Many theorists now see them as observable dispositions or functional states, which can be unconscious neurocognitive states. On the dispositionalist view, if I consciously believe that I value my friendship with Greg, but I have a disposition to avoid interacting with Greg, I may dispositionally judge Greg to be an inadequate friend, though I consciously represent him as a good friend (depending on relevant background factors). On the functionalist view, a judgment plays a reliable causal role in a certain pattern of behaviour. There is evidence that unconscious states can play a reliable role in producing goal-oriented behaviours. For example, unobtrusive priming of cooperation (using words like ‘dependable,’ ‘helpful,’ and ‘supportive’) caused subjects assigned to fisher roles in a resource-management game to cooperate to maintain the critical limit of 70 fish in the lake (Bragh et al. 2001: 1018). Notably, there was no difference between a group given unobtrusive primes, and a group given explicit instructions, suggesting that non-consciously activated goals can play just as reliable a role in the production of cooperative goal-oriented activity as the conscious intention to cooperate.

Similarly, unobtrusive priming of high-performance behaviour (using words like ‘win,’ ‘compete,’ and ‘success’) caused subjects to resume an intellectually demanding task (Scrabble) after a disruption, even when they were given the chance to play a less-demanding, ‘fun’ task instead (Bragh et al. 2001). This suggests that non-consciously activated goals are fairly robust: they persist over time. Indeed, non-consciously activated goals “show the same quality of persistence over time toward the desired end state, and of overcoming obstacles in the way, as… consciously set goals” (Bragh et al. 2001: 1015). That is, they have the quality of a disposition, and they can be just as causally efficacious at producing goal-oriented behaviour as conscious intentions.

On a dispositionalist picture, then, we can interpret the goal-pursuing behaviour provoked by non-conciously activated goals in each task as a judgment that cooperation/high performance (depending on the task) is worth pursuing, even if the judgment was not reflectively available to the actor at the time of action. Ditto for the functionalist interpretation. Granted, the value of the primed goal may have been consciously available to the subjects—they may have honestly affirmed that they did, e.g., consciously value cooperation in the resource-management task. Still, if one of them were to deny this, we would have reason to doubt their conscious appraisal, and favour the functionalist/dispostionalist interpretation on which they (implicitly) did value the goal.

On the cognitivist view, then, blame may sometimes manifest as an unconscious negative judgment. If I persistently fail to cooperate with certain types of people due to non-consciously activated states, perhaps I (functionally) negatively appraise, and thus blame, those people.

Another reason to doubt that judgments are necessarily conscious is that most people take judgments to be “intertwined” with emotions, both conscious and unconscious, meaning that judgments sometimes contain unconscious affective states (Chekov 2009).

Another compelling example of unconscious blame is blame motivated by implicit bias, where the normative significance of the action is first-personally opaque. Although implicit bias is a contested notion, the general consensus is that implicit biases are implicit associations between concepts, manifested in overt behaviour (but see Levy 2015 & Mandelbaum 2013 for exceptions, to be discussed shortly). Implicit racial bias, for example, is tracked by the Harvard Implicit Association Test by measuring subjects’ reaction times (‘response latencies’) to positive and negative words (‘smile,’ ‘honest,’ ‘disaster,’ ‘agony’) after priming the subject with a black face or a white face. Most white people are faster to pair black faces with negative words and white faces with positive words than vice versa. This means that they score positively for implicit racial bias.

It is debatable whether IAT scores translate into real-world discrimination against Black people (i.e., overt manifestations), but the reason many people think they do is that anti-black discrimination is a pervasive part of the cultural landscape, even though most people would (presumably sincerely) deny being racist. To give just a few examples of this phenomenon: Employers who had advertised an employment equity policy nonetheless favoured resumes with European American names over identical resumes with African American names by a ratio of 2:1 (Bertrand & Mullainathan 2004); managers discriminated against job applicants who used African American Vernacular English rather than Standard English in phone interviews, rating those applicants as less intelligent, less ambitious, and less qualified (Henderson 2001); adults view black girls as less innocent and more adult-like than white girls, resulting in more frequent and severe punishments in schools and the juvenile justice system (Epstein et al. 2015); African American men face harsher sentences than white men for the same federal crimes, controlling for criminal history, age, education, and citizenship (USSC 2017).

These studies show that America’s expressed attitude (i.e., overt disposition) toward black people as a group ranges from avoidant to punitive. We avoid renting to them, hiring them, and giving them equal educational opportunities; we punish them, even as young children, by barring them from equal participation in our most valued social institutions (schools, employment, housing, civilian society). Should we say that, if we engage in these avoidant, punitive, and generally exclusionary practices, implicitly rather than explicitly, they are not expressions of blame? If we are not committed to the requirement of consciousness, this constraint is unmotivated.

Neil Levy (2015) disputes the implicit-association model (as well as the implicit-belief model) and proposes that implicit biases are sui generis states, with some belief-like and some association-like properties. Like beliefs (but unlike implicit associations), implicit biases have some propositional structure, on his view, based on studies showing that previous exposure to counter-stereotypical images (women in a kitchen, men in an office) can produce counter-stereotypical responses on implicit association tests: subjects who had been exposed to stereotypical images were faster to associate competency/agency words with female faces than male faces (De Lemus et al. 2013). This suggests that implicit biases interact with the propositional structure of beliefs, albeit not inferentially – these states are still unconscious processes. Levy, in fact, holds that implicit biases, while not associations, are both unconscious, and motivationally efficacious (2015: 809). This theory, then, still supports the view that blame can be implicit, inasmuch as it can involve implicit biases that render its normative significance first-personally opaque. When we express or enact implicit biases, we do not realize that we are doing so, meaning that the evaluative significance of our behaviour is unavailable to reflective retrieval; we are not aware that are treating others in (often) aversive, avoidant, and punitive ways.

Notably, our emotions and judgments can be informed by implicit bias. Subjects exposed to a bad smell exhibited harsher moral judgments than controls, showing that implicit disgust can mediate explicit moral judgments (Schnall et al. 2008). Inducing disgust in subjects also produces intuitive disapproval of gay people, even in subjects who deny having any anti-gay bias (Inbar et al. 2018). In these cases, the subject is aware of harbouring aversive emotions and judgments, but not of the causal antecedents of those states. These are cases of ‘moral dumbfouning’ (Haidt 2001): the subject is aware of her emotion/judgment, but not of why she has it. On the other hand, if someone displays aversive behaviour in response to a prime without knowing it, the person’s dispositional emotion/judgment is first-personally opaque. This is a paradigmatic case of ‘unconscious blame,’ but moral-dumbfounding cases of blame are also possible.

We can define ‘unconscious blame’ as a negative overt response to another person, motivated by unconscious neurocognitive states, which are not directly available to consciousness, nor immediately reflectively retrievable. These negative responses are typically patterned because the implicit emotional, cognitive, and conceptual contents of blame, as we have seen, tend to be reliably activated by certain stimuli (unobtrusive primes, salient cultural stereotypes), and to persist in spite of disruptions. However, un-patterned responses should not necessarily be discounted as candidates for blame, even if they are not the norm. Expressions of unconscious blame are ‘negative’ or ‘aversive’ in that they express a ‘negative stance’ toward their target, as expressed in avoidant, punitive, or otherwise antisocial, patterns or instances of behaviour.

This account departs from the Strawsonian view that blame is a response to an agent’s quality of will, since blame in these cases responds to salient triggers and subjective states, not the target’s qualities. But this view allows that blame is typically a response to an agent’s perceived quality of will, mediated by the perceiver’s implicit and explicit emotions, judgments, and attitudes, and situational triggers. When blame is well-placed or ‘fitting,’ it accurately responds to the wrongdoer’s objective quality of will. In practice, blame frequently misses its mark due to implicit states. Unbiased blame is non-discriminatory, while biased blame is bigoted in various ways.

The idea that holding-responsible (including blaming) can be unconscious is not exactly revisionary. Expressivists in particular grant that our responsibility practice regularly involves manifestations of unconscious attitudes, though they focus on unconscious states at the ‘moral contribution’ stage (‘stage 1’ above). McKenna, for example, says that committing an unintentional offense is blame-able (2014); Smith holds that forgetting about a close friend’s birthday warrants moral approbation (2005); and Sher says that forgetting a family pet in the backseat of a hot car is blameworthy (2010). These are ‘omissions cases,’ in which an omission (to do what one ought, take proper precautions, exercise due moral vigilance, or what have you) reflects a negative quality of will, opening the person to blame. Many people now take this ‘anti-reflecitivist’ stance on responsible agency, according to which people can be responsible for unconscious transgressions (Doris 2016). That said, few theorists have discussed the role of unconscious states in stages two and three of our responsibility practice: the stage at which we express the reactive attitudes (‘moral address’), and the stage at which we respond to these attitudes (‘moral accounting’). This is not to say that theorists deny that we can express and respond to these attitudes unconsciously, but if they believe that we can, they don’t explicitly say so.

This points to an asymmetry in the literature, in which an abundance of work has been done on whether one can be blameworthy for an unconscious violation, whereas relatively little has been said about whether one can express and respond to the reactive attitudes unconsciously. If I am right, then we can, at least, unconsciously express these attitudes. When we unconsciously express a reactive attitude, we are unaware of some normatively salient feature of our behaviour.

Here is an example (taken from an earlier paper [citation withheld for anonymity]). Physicians prescribe fewer pain-killers to Black patients than white patients when both exhibit the same symptoms (Silverstein 2013; Cleeland et al. 1994). This may be related to the fact that white observers show less physiological arousal in response to Black people’s pain than white people’s pain, with their level of arousal correlating with their level of implicit racial bias (Forgiarini et al. 2011). That is, physicians may prescribe fewer pain-killers to Black patients due to implicit racial bias. If so, then they are, on scrutiny, treating Black patients in a punitive way, denying them the medical treatment they need or deserve on the physician’s evaluation of equivalent suffering in white patients. They may also distrust their Black patients’ ability to take pain-killers responsibly compared to their white patients.[1] If this analysis is right, then physicians who manifest racial prescription bias are unconsciously blaming (i.e., punishing, distrusting, implicitly judging as undeserving or incompetent) their Black patients. They are not consciously aware of doing this, but their overt behaviour nonetheless expresses blame to Black patients.

Again, I am not saying that theorists would not count this as an instance of expressive blame; I am saying that there is little discussion of such cases – cases that are extremely consequential in their impact on the life prospects of historically disadvantaged groups.

The view that implicitly biased prescription practices can count as blame requires a move away from an intuitive perspective on blame: the first-person standpoint. Physicians motivated by implicit racial bias are not aware that they are expressing these biases in their prescriptive habits, much less that they are unconsciously blaming their Black patients, so we cannot rely on the first-person standpoint to deliver accurate evaluative descriptions of the subject’s own expressive habits. (This is a rejection of what Sher [2010] calls ‘the searchlight view,’ applied here to blaming rather than to blameworthiness). We cannot necessarily rely on the average observer’s appraisal either, since most people are not in a good position to evaluate whether a perceived pattern of behaviour is an instance of implicit bias. To evaluate whether a person P’s pattern of behaviour counts as blame, we need to assess P’s stated intentions and overt behaviour, and compare these data to relevant psychological and sociological research or trends. We should, in a sense, adopt a ‘heterophenomological perspective’ to blame, in Dennett’s terms (1991) – a perspective that compares a piece of phenomenological data against a range of intersubjective and empirical data. A physician’s prescriptive practices can be evaluated for implicit bias by observing the person’s treatment of different patients, in comparison with national prescription trends. This allows us (if there is sufficient data) to determine if unconscious blame is explanatorily potent in a particular case. (Identifying unconscious blame is easier on a population level than an individual level – that is, it is easier to determine whether a certain social group unconsciously blames another social group, than whether a certain individual unconsciously blames another individual in response to the person’s identity or group affiliation. Nonetheless, the population trends are manifestations of individual patterns of behaviour).

This is a suggested methodology for identifying unconscious blame. It does not bear on the question of whether unconscious blame is a valid construct. This question has already been settled by the discussion about the nature of emotions and judgments, which can be implicit.

  1. Fittingness constraints: Communication to whom?

Assuming that responsibility is an expressive practice, another pertinent question is: What is its appropriate target?

Strawson (1963) argued that the reactive attitudes respond to an agent’s quality of will, unless there are salient extenuating conditions. Excusing conditions mollify the reactive attitudes, rendering them less severe, whereas exempting conditions trigger a complete suspension of the reactive attitudes, and a switch to ‘the objective attitude,’ i.e., an emotionally detached, typically avoidant, stance. Adopting this stance involves treating the target as, “perhaps, an object of social policy; as a subject for what, in a wide range of sense, might be called treatment; as something certainly to be taken account, perhaps precautionary account, of; to be managed or handled or cured or trained; perhaps simply to be avoided” (1963: 66). We refuse to engage emotionally with the target in the objective stance, and we aim to exclude the person from ‘the moral community,’ the group of morally-responsive agents. This is a fitting attitude toward moral incompetents, those who are deeply or constitutionally incapable of responding to moral claims.

If the objective attitude is an emotionally detached stance, then it excludes blame, and other reactive attitudes, from its scope. The ordinary defense for this position is that blame toward moral incompetents has no point. The point of blame, many assume, is to elicit a fitting emotional reaction in the target, in response to a salient moral demand. This is its ‘conative orientation’ (or one of its dominant conative orientations, if there are several): blame solicits a fitting response from the target. This conative aspect of blame can be retrospective or prospective, or both. Sher, for instance, describes the conative orientation of blame as a desire that the wrongdoer “have responded, or that he be disposed to respond, to what we consider a compelling moral reason” (2006: 105). There is a retrospective desire that the agent have behaved otherwise, a prospective desire that he (be disposed to) respond appropriately in the future. Others see blame as not merely desiring, but demanding, a response. We can call this a ‘strongly conative’ orientation: the blamer doesn’t just wish the wrongdoer had acted otherwise, she calls for action. This strong conative orientation is required, or presupposed, by many influential theories.

McKenna, for instance, construes blame as, in effect, a request for a ‘moral account’ from the transgressor (2014). Similarly, McNamara says that blame solicits a response from the wrongdoer, and that “successful” cases of blame are those that “receive a response” (2012: 159).

Others see this constraint as only partial, or defeasible. Smith’s protest account interprets blame as “implicitly seeking a response” from the object of protest (2013: 40), consistent with Strawson and McKenna. Absent this eliciting function, blame could be “a one-sided adjustment of attitudes,” and, thus, “deeply non-relational” (41) – an unacceptable conclusion. But Smith thinks that blame can also serve another function: to register a complaint against someone’s attitude – though this might be an ‘imperfect’ or ‘incomplete’ case. McGeer similarly argues that blame can serve two functions: regulating people’s behavior, and appraising an action as wrong. Blame, then, does not necessarily solicit a response from the wrongdoer – it can serve a purely signaling function. Bell (2013), too, thinks that blame can serve goals other than appealing to the wrongdoer: it can function to motivate and educate the community and to signal one’s values to the group.

These last three views (Smith’s, McGeer’s, and Bell’s), which we can call ‘multi-functional accounts of blame’ (because they see blame as serving more than one aim), mark a departure from the Strawsonian ‘fittingness’ constraint, on which blame is fitting only if the proper target – namely, the wrongdoer – is suitably sensitive to the demand. The justification for this departure is that blame’s proper ‘target’ is not necessarily, or exclusively, the wrongdoer; rather, the target of the conative orientation of blame, which seeks a response, must be some other individual or group capable of responding appropriately (with sympathy, vicarious indignation, political activism). The ‘conative target,’ in other words, is a sensitive moral audience, those to whom the expressive act is communicated, and from whom uptake is demanded, sought, or anticipated. This condition satisfies blame’s expressive ‘point’ – it communicates with a fit respondent. That said, the wrongdoer is surely an appropriate target of blame’s emotional/cognitive contents (resentment, distrust, disapproval), for, whom else could be? If we cannot extirpate the emotional contents of blame from human psychology, they must have some target, and this must be the wrongdoer. Yet the morally incompetent wrongdoer is not an apt target for blame’s conative orientation.

This multi-directional view of blame’s ‘point’ respects the relational element of Strawson’s theory, but it expands the scope of the reactive attitudes beyond the relationship between the complainant and the accused; it recognizes that blame can be expressed to the community, in relationship with the community. The incompetent wrongdoer, however, is not completely outside of blame’s scope; this person is an apt target of blame-constitutive emotions and judgments, though not of blame’s demand for a suitable response. That is, on my interpretation, the emotional/cognitive contents of blame, on the one hand, and the conative force of blame, on the other, come apart. In separating these two elements, we can see how the conative element of blame can be fittingly addressed to an audience, in keeping with the multi-functional view.

This interpretation implies that moral incompetents are not exempt from blame in the way that Strawson envisioned; they may be exempt from blame’s conative aspect, but they are still apt targets of blame’s emotional and cognitive contents. That is, when we adopt the objective attitude toward someone, we only partially suspend the reactive attitudes; we suspend a given attitude’s conative force, but we do not suspend the attitude’s emotional and cognitive contents. We are permitted to be resentful, reproachful, and censorious toward morally incompetent wrongdoers.

This is, I think, a more realistic view of blame’s psychological profile. Do the families of psychopathic (non-responsive on many views) murder victims suspend their resentment toward the psychopath? No, but they might turn to the community and the justice system for an appropriate response to their understandable emotions. (An appropriate response would be sympathetic resentment, sequestration, and other actions that ‘take the side of’ the victims and protect the community from harm. Note that there are additional fittingness constraints on blame and punishment, such as proportionality and intelligibility, which I do not have time to address here).

The multi-functional view also calls into question McKenna’s model. If the audience can be the target of blame’s demands, then the ‘moral accounting’ stage must be expanded to include a range of fitting responses from variously situated respondents. This is because the audience is not in a position to give a ‘moral account’ (apology, explanation, restitution) on behalf of the transgressor – only the transgressor himself can do this. Thus, the ‘moral accounting’ stage must be expanded to include other types of response – vicarious resentment, collective action, and so on.

This means that we can tweak Strawson and McKenna to make them compatible with a multi-functional view, by separating the emotional/cognitive and conative elements of blame. But why should we think, all things considered, that blame can be addressed to an audience?

There are normative and descriptive reasons on offer. The descriptive reason is that (1) this is how blame actually functions, and (2) we should endorse a descriptively accurate theory. McGeer provides an evolutionary (descriptive) account of blame, on which the reactive attitude evolved as rapid and involuntary responses to perceived transgressions. (However, as ‘mentalizing creatures,’ we are capable of reflecting on and revising these responses, within limits set by evolution). These evolved responses “play a critical role in regulating behaviour by way of making salient the demands that shared norms place on our actions and attitudes” (2014: 183). They promote pro-social behaviour by ‘coding’ certain practices as wrong, and ‘signaling’ disapproval of those practices. They can serve these ‘coding’ and ‘signaling’ functions in a variety of ways: not simply by demanding a response from a transgressor, but also by demanding recognition from the community. We are constrained in how far we can suppress and alter these responses, but we can implement plans and policies to help ‘guide’ them in pro-social ways. In sum, emotions, while somewhat plaint, are an inextricable part of our blaming psychology.

Notably, even if blame does not function well in modern society compared to our early evolutionary history (more on which shortly), it remains true that emotion-laden blame is an ineradicable part of our moral psychology, though it may ‘misfire’ due to differences between modern and ancient ecologies. The direction of blame and the existence of blame are different things: while we can perhaps ‘refine’ blame to minimize its harmful effects, we cannot eliminate it.

The normative reason in favour of multi-functionalism is that blame should serve a multitude of functions, as these functions help to regulate human behaviour in positive ways. Bell points out that blame can serve a variety of (local) functions (perhaps in the service of a general, over-arching function), aside from eliciting a reaction from the transgressor; it can also educate, motivate, and register a moral complaint with, the moral community. Indeed, if we think of holding-responsible as a functional practice that serves the moral ends of the community, it is, I think, arbitrary to cut off any local aims that subserve this goal. To be effective, expressive practices must be versatile, appealing to and ‘connecting with’ as many people as possible. Any number of specific aims, then, may be compatible with the general aim of “responsibilizing” people (Pettit 2007).

Bell advances this argument as a way of defeating the view that blame is ‘positional,’ or fitting only toward those over whom we have moral authority, rooted in our relationship with the blamee (viz., Darwall 2006, Cohen 2006). Against this view, Bell claims that we can blame strangers, co-conspirators, the deceased, and (as I have urged) moral incompetents. The rationale for this departure from the ‘positional’ view is that blame is multi-functional, and not all of its functions are indexed to authority relations. We have responsibilities, not only as members of specific relationships, but “as critics” and “third parties” (2012: 265). Indeed, we witness wrongdoing every day; the fact that we may not be in a position to solicit a response from the wrongdoer does not entail that we cannot, in a meaningful sense, blame the person. Arguably, when we ignore the transgressions of moral incompetents, we make ourselves complicit in their actions.

I agree with Bell’s objection to the ‘positionality’ constraint, insofar as I agree that morally insensitive agents can be resented, distrusted, and censured, but I do not think that we thereby need to abandon the intuitive idea that the conative orientation of blame must be directed to a responsive target. While the emotional and cognitive contents of blame are fitting for the transgressor, the proper recipient for the demand for a response may very well be someone else.

If this is right, then the answer to the question, ‘who is an apt target of blame?’ is broader and more nuanced than Strawson envisioned. Moral incompetents are apt targets of blame-constitutive emotions and judgments, but not of moral demands; these demands should be addressed to a suitably sensitive audience.

  1. Functionalism: Communication to what end?

We have seen a variety of functionalist interpretations. McGeer, Bell, and Smith agree that blame’s functions include (1) registering someone’s behaviour or quality of will as morally problematic, and (2) soliciting a response from the wrongdoer or the community. I have called these views ‘multifunctional’ simply because they allow that blame can function not merely to demand a response from the wrongdoer, but to do other things as well. These theorists, however, posit distinct higher-order aims that blame is supposed to serve, under which other (subordinate) aims can be subsumed. For Bell, blame’s main function is to “resist wrongdoing” (266); for Smith, it is to protest (or “register one’s protest of”) someone’s treatment of people (27); and for McGeer, it is to enhance the moral fitness of (or ‘responsibilize’) the group. These aims overlap: protest and resistance are similar – perhaps even coextensive – concepts, and the aim of responsibilizing the group surely encompasses resisting and protesting violations, whatever else it may entail.

Rather than arbitrating which of these views is superior, in this section I will argue that each is too thin to serve as a normatively adequate theory of blame, and that any adequate theory of blame must make relational equality a priority. The reason is that our perception of wrongdoing will be distorted as long as relational inequality is the norm. If our blaming responses register, protest, or resist ‘wrongdoing’ by our lights (as they must), they will systematically miss their mark, as ‘our lights’ are coloured by the conditions of social injustice in which we live (viz. Fricker 2007, Medina 2013). Repairing relational inequality, then, must be one of our explicit aims and priorities as members of the moral community. If we do not effectively prioritize this aim, we risk blaming people in unjust and irrational ways – ways that harm marginalized groups.

Bell and Smith posit similar aims for blame (protesting and resisting wrongdoing), while McGeer thinks that blame registers offenses and responsibilizes people. I don’t doubt that blame should do these things, but in our society, it does not advance these aims effectively. This is because our perception of wrongdoing is distorted by implicit biases informed by cultural stereotypes, collectively speaking. (Individuals are biased to a greater or lesser degree depending on their social position, learning history, neurocognitive profile, etc. What follows here is a population analysis, abstracting away, for the most part, from individual perception. This analysis is compatible with the idea that some people are moral-epistemic saints, impervious to implicit bias; but, on a standpoint epistemological picture, most people’s perception will be affected by implicit biases rooted in cultural stereotypes and scripts, since individual perception reflects background epistemic conditions, or what Jose Medina calls ‘the social imagination’ [2012]).

As we saw in section 2, physicians systematically distrust and punish Black patients; educators systematically punish and sanction young Black girls; employers systematically distrust and avoid Black job applicants; the justice system systematically gives Black men harsher prison sentences. These are examples of misplaced blame – specifically, misplaced unconscious blame – in which our (collective) blaming reactions subvert their putative function(s), viz., protesting and resisting wrongdoing, and responsibilizing people. These attitudes, in effect, malfunction in current social conditions, making society worse off and less responsible. We are less responsible qua blamers, blamees, and third-party critics, as we are more mired in distorting cultural stereotypes and scripts that prevent us from discerning who objectively deserves blame (or protest or avoidance), and who does not. The more our blaming practices miss the mark, the harder it is for us to accurately perceive wrongdoing, and to differentiate morally salient qualities from morally neutral demographic attributes that happen to be stigmatized by patriarchal-colonial-cissexist culture. The more likely we are, in other words, to illicitly blame and punish members of culturally disadvantaged groups, and to illicitly praise and reward members of socially privileged groups –  to get blame wrong. Meanwhile, the underprivileged are deprived of equal access to responsibilizing institutions, such as quality education, housing, and lucrative jobs, which is an affront to their dignity as persons. This not to say that these individuals are particularly epistemically flawed, but rather, it is to emphasize that they lack access to resources that enhance responsibility – for example, safe housing in which to responsibly raise children, occupations with which to responsibly pay off one’s student loans, etc.

Because our blaming practice is systematically biased against historically disenfranchised groups, some theorists have adopted an eliminativist position, arguing that we ought to eliminate blame because it does more harm than good (Waller 2016, Levy 2012). In other words, on a descriptively accurate account, taking proper account of modern social conditions, it is accurate to say that blame serves the function of unfairly punishing and oppressing minorities. Prima facie, this is the opposite of McGeer’s evolutionary account, on which blame serves to enhance moral-group fitness. But the two views are compatible if we see McGeer as talking about blame’s function in the Paleolithic era, in which societies were more egalitarian (Dyble et al. 2015), and Waller as talking about blame in modern times, in which inequality is systemic – particularly in America, the most economically unequal developed country (Allianz 2016). Perhaps blame’s evolved function has been co-opted by wealthy colonial-patriarchal-cissexistto advance the material and political interests of the elite. If so, then blame no longer serves its evolved function; it serves a man-made function: to promote and reinforce modern asymmetries of power.

I think that something like this story is right, but I agree with McGeer that blame is a canonical feature of human moral psychology, not something that we can eliminate, though we can ‘domesticate it’ with social engineering and careful planning. So, eliminativism is not a viable option, pace Waller. I also think that partially eliminating blame would be a mistake, as it would prevent us from responding sensitively to the suffering of the oppressed – something that we cannot and should not do. Since, on the most optimistic projection, we will not eliminate global inequality in our lifetime (indeed, income inequality in America is steadily growing (Saez 2016)) – it would be wrong, and probably psychologically impossibly, to completely suspend the reactive attitudes. It would be wrong because the detachment involved in a full suspension of the reactive attitudes would constitute callous indifference to the suffering of the oppressed; and achieving this state is most likely psychologically impossible, in any case, because we are not, as McGeer drives home, the kind of creatures who can sustain indifference in the face of undeserved suffering (with the exception of psychopaths). When we see a starving child, we feel both sympathy for the child and outrage against the person or people responsible for the child’s plight. This is how we are built.

Eliminativism also faces an aggregation problem, since not everyone can be expected to endorse it as a policy, and if only a few conscientious eliminativists suspended their blame (assuming this were possible), this would take the pressure of public condemnation off the worst offenders.

Rather than eliminating blame, I agree with McGeer that we should try to domesticate it, and I submit that this domesticating project involves promoting ‘relational equality,’ or equality of “authority, esteem, and standing” (Anderson 2015: 65). The reason is that equality is a prerequisite to holding people responsible in a fair and rational manner, given that inequality creates stereotypes and pernicious social scripts that impair our ability to recognize wrongdoing. Specifically, inequality creates implicit biases that prevent us, as a society, from distinguishing ‘wrong’ from ‘stigmatized,’ ‘blameworthy’ from ‘socially marginalized.’ Thus, relational equality must be an explicit aim, or end, of our blaming practice. This does not preclude the aims of protesting and resisting wrongdoing and registering wrongdoing, but it is temporally prior to the effective (perfect) attainment of these ends: if we do not eliminate inequality-based biases, we cannot accurately perceive people’s morally salient qualities and accurately determine who warrants blame. We do not need to completely eliminate inequality prior to blaming, but we must reduce the influence of inequality on our blaming practice if blame is to hit its mark every time.

I say ‘relational equality’ because I have been describing blame as a relational practice (between blamer and blamee, or blamer and society, or some combination of these), and relational equality, as envisioned by Elizabeth Anderson, is also a relational notion: it seeks to promote fair and equitable interactions across a variety of humanly-valuable dimensions. In particular, relational equality entails not only equality of resources, but also equality of epistemic and moral standing, esteem, and regard. To blame people fairly, we must blame them in a way befitting their (objective) moral and epistemic standing – a way that enhances relational equality. When we blame people on the basis of implicit bias, we unfairly morally disdain them; we treat them as morally and epistemically ‘beneath us.’ When we avoid, ostracize, or exclude people for no good reason, we treat them as morally unworthy of our interest, attention, and cooperation; we take ‘the objective attitude’ toward them without warrant, twisting it into a tool of objectification and marginalization. These misguided (but common) blaming practices are based on, and partly constituted by, relationships of unequal epistemic and moral standing. Notably, Anderson’s view is a response to classic egalitarian theories, which focus on distributing resources fairly, but which tend to treat the recipients with an attitude of “contemptuous pity” (Anderson 2000: 6), characterizing them as irresponsible, lazy, and poorly endowed. Anderson argues that the point of egalitarianism should be to allocate goods in the spirit of respect for human dignity, not to transfer funds from the ‘responsible’ to the ‘irresponsible,’ ‘lazy,’ and ‘pitiable’ – that is, she rejects the premise of classic egalitarian reasoning. I am suggesting that this should also be the guiding aim of blame: to distribute moral and epistemic regard fairly, so as to respect the objective moral and epistemic qualities of members of the moral community. Enhancing relational equality and enhancing the rationality of our blaming practice, on this view, go hand in hand.

Although I have been focusing on population-level blaming practices, the goal of enhancing relational equality can, and should, be implemented on both a personal and an institutional level. Individuals can take steps to try to mitigate their implicit biases, and institutions and governments can implement policies to reduce the adverse effects of implicit bias on the common good. As I have discussed these approaches elsewhere, I will not elaborate here, except to note that I believe that individuals should try to cultivate epistemic virtues (Fricker 2007) and vigilance (Murray 2017) to the greatest extent possible, while institutions and government should implement epistemically responsible protocols (Longing 2008). Even if the institutional approach is more efficacious (as McGeer [2013] suggests), the first method is indispensable, given that fully responsible social institutions are not on the horizon. While we are waiting for effective social engineering policies to come into effect, we should work on our own moral character.

  1. Conclusions

I have argued that blame functions to express emotional and cognitive states to the wrongdoer, but its demands are properly addressed to a suitably sensitive audience. Blame can be expressed unconsciously, as evidenced by neurocognitive, psychological, and sociological data. Because expressive blame as a dispositional response is influenced by implicit biases and other morally-irrelevant situational factors, we should strive to blame people in a way that enhances relational equality, or equality of standing, authority, and esteem. Thus, a prerequisite for rational blame, and one of its proper aims, must be relational equality. This might sound circular, but the same can be said of the heart: pumping oxygenated blood throughout the body is a prerequisite for a fit heart, and this is also the heart’s function. Similarly, relational equality is a prerequisite for a fit blaming practice, and it is also the function (or one of the utmost functions) of this practice. A fit moral community, in which people hold each other responsible on fair and rational grounds, aims to promote relational equality, and the more they do this, the closer they get to this aim.

[1] Distrust is a reactive attitude on McGeer’s view (2008).


Responsibility as a non-ideal conversational practice in non-ideal social conditions: 3 mediating variables


  1. Introduction: Responsibility as a non-ideal conversational practice 

On one influential school of thought, moral responsibility is an interpersonal practice in which someone deploys the reactive attitudes of blame, resentment, approbation, forgiveness, and so on, to another person (P. F. Strawson 1963). Michael McKenna (2013) describes this practice more precisely as a conversational exchange with three paradigmatic stages: (1) a moral contribution, in which someone performs a morally-relevant action, (2) a moral address, in which someone deploys the reactive attitudes, and (3) a moral accounting, in which the addressee responds to the addressor, by, e.g., taking responsibility or denying responsibility.

I think that we can hold people responsible outside of standard conversational contexts, as I have argued elsewhere, but I think the conversational model captures a paradigmatic and very familiar mode of holding-responsible, so I will use it as a basis for the present inquiry.

There are debates about whether the reactive attitudes are purely cognitive (Smart 1961), affective (Wallace 1994), or conative (Sher 2006). These debates concern the internal contents of these states, i.e., those experienced or invoked in the addressor and addressee. One of the objections to non-affective accounts is that they are ‘too sanitized,’ and do not present a realistic picture of moral psychology: even if negative affect is not an essential feature of blame, it is a characteristic reaction, triggered in normal human beings in response to perceived moral violations (McGeer 2013). Non-affective accounts are often driven by normative concerns, particularly the concern that we should construct a ‘civilized’ account of blame that reduces strong emotions (ibid). Whatever the benefits of this approach, it is reasonable to worry that this conception of blame is not psychologically realistic.

Moreover, eve if blame is ideally unemotional, it is, in reality, typically influenced (if not constituted) by affective and automatic processes, which can distort ideal judgments of blame. Thus, it is worthwhile to consider the role of these cognitive mediators. I will propose that affective and automatic states play an integral role in blame and praise.

What is even more neglected in the literature is the fact that responsibility is not simply a set of internal agential states, but rather a complex social practice, influenced by the visible attributes, body language, and lexical preferences of the interlocutors. If the reactive attitudes are part of a conversational practice, as McKenna says, this practice has lexical contents and modes of expression chosen by, or characteristic of, the speakers. Some of these are internal contents of speech (lexical), while others are expressive features of speech (vocal intonation, body language); but all of these variables, whether ‘internal’ or ‘external,’ influence how we perceive and respond to speakers in our conversational encounters. Because we respond to one another quickly and relatively automatically in quotidian conversational exchanges, we are liable to respond to the vocal traits, lexical mannerisms, and visible attributes of our conversational partners.

This is already an important departure from the idealized picture of blame as a rational and impartial normative judgement, but it still misses an important mediating variable: social conditions. Our lexical preferences, body language, and visible appearance are not value-neutral features of our selves, accorded equal weight in our day-to-day conversational exchanges; rather, these traits are valued more or less depending on how closely they approximate to the patriarchal, colonialist, Eurocentric, heteronormative, cisgender cultural norm. The “ideal speaker” – the speaker whose practices and attributes are most respected and valued – is a social construct rooted in a historical value system that favoured, and still favors, cisgender white males, and silences or marginalizes other groups in most contexts. (There are exceptions – for example, men are seen as less credible care-takers because care-taking is coded as feminine). To this day, the speech of historically disenfranchised groups is not given the same credit as the speech of cisgender white men on average (meaning across a broad range of contexts) (Fricker 2007). These biases are the root of “mansplaining” (Sonit 2012), “whitsplaining,” and other communicative practices in which the same speech content is more valued when expressed by the “ideal speaker” according to cultural schemas or stereotypes (Valian 1999). If responsibility is a conversational practice, then it is a practice informed by dominant perspectives on the value of gendered, sexed, and raced speakers.

Once we see the reactive attitudes as part of a complex social practice, embedded in and informed by conditions of social injustice, the scope of analysis expands vastly beyond mere psychological investigation. Blame and praise are subject to, not only normative and psychological, but also anthropological, sociological, linguistic, and historical analysis. Responsibility, understood as a non-ideal conversational practice embedded in non-ideal social and epistemic conditions, stands at the intersection of myriad fields of inquiry.

In this post, I will explore three (I believe) somewhat under-theorized variables that inform “our responsibility practice,” by which I mean the practice in which we express and respond to praise and blame: (1) the role of empathy, (2) the lexical preferences of speakers, and (3) the physical attributes of speakers. I will discuss these variables at the level of social groups, eliding individual differences amongst groups, for the purpose of depicting general trends, in the spirit of social anthropology or feminist sociology. Since I have discussed the role of implicit bias in blaming and praising elsewhere, I will not emphasize this variable here, except when it plays a mediating role in variables (1)-(3).

2. Caveats & clarifications

Some caveats are, of course, in order. The ‘contextualizing’ or ‘situating’ of responsibility – by assessing it as a socially embedded conversational practice – is essentially a project in social anthropology, sociology, or standpoint epistemology. This means that I will be making generalizations about how empathy and group-level communicative habits affect social groups. For example, I will be saying that empathy is disproportionally experienced and enacted by women, and patterns of empathic concern disproportionally benefit cisgender white men, somewhat indemnifying them against blame and punitive sanctions. This doesn’t mean that every member of each social group fits the operant generalization, but these generalizations point to significant social trends – trends that tell us something about how responsibility is distributed across and amongst social groups.

Second, I will be discussing very broad groups, and will not be giving every social group equal attention, simply due to lack of space. I also will not talk much about how converging intersections of oppression compound disadvantages (particularly, susceptibility to blame), while converging intersections of privilege ramify advantages (especially indemnity against blame and penalties). But these intersections are often implied if not explicit.

Third, by “our responsibility practice” I mean to refer to a conversational practice involving the reactive attitudes, where the reactive attitudes are essentially communicative acts, conveying attitudes of “resentment, gratitude, forgiveness, anger,” love, indignation, approbation (Strawson 1963: 66), antipathy (Watson 1982), as well as trust and distrust (McGeer 2008, Helm 2014). These responses mediate our interpersonal relationships in positive and negative ways, depending on how they are enacted. While these reactions likely involve emotional contents in the typical case, I am not committed to saying that conscious affect is a necessary component. As others have argued, we can communicate resentment, distrust, anger, etc., in a dispassionate way, e.g., by unemotionally “unfriending” a contact on FaceBook (viz., Smith 2013: 32). More importantly, however, we can, and often do, communicate resentment, distrust, and hostility toward others in our overt behavior, without consciously experiencing the concomitant emotions, when the resentful/distrustful/hostile (etc.) behavioural response is provoked by unconscious cognitive processes (e.g., implicit bias), without accompanying conscious affect or belief. For example, if a physician deems an African American patient insufficiently responsible to use prescription pain-killers as directed, the physician may be communicating distrustantipathy, or hostility to the patient, without consciously feeling these emotions or holding relevant beliefs about African Americans. These “attitudes” are nonetheless conveyed in the physician’s prescriptive decisions. This interpretation of the reactive attitudes accommodates many of the criticisms of both “emotional” accounts and “sanction” accounts, and resembles Smith’s communicative approach (2013), which is a good fit for the conversational model. Conversation is, after all, an essentially communicative exchange between speakers. (I am not, however, committed to Smith’s protest account, as will become clear in section 6).

Fourth, my analysis of social variables that distort our responsibility practice extends to other, non-conversational practices of holding-responsible, such as praising someone in absentia and blaming the deceased. But for this particular project, I am focusing on the conversational account as delineated by McKenna, as this is an important locus of analysis.

 3. Empathy

It is reasonable to think that empathy plays a role in our responsibility practice. (I use the term empathy broadly to encompass a range of emotional responses to what others are feeling, including sympathy and vicarious joy and anger, following Heidi Maibom 2017).  In Watson’s influential article on the reactive attitudes (1982), he surveys possible explanations for Strawson’s most controversial excuse, viz., having had peculiarly unfortunate formative circumstances. Watson says that empathy can play a role in softening our sense of blame toward a victim of childhood trauma, but this is not a rational basis for clemency: it is, rather, an unreflective emotional response with no normative weight. Still, he suggests that empathy plays a role in ordinary moral cognition.

Erin Kelly (2017) similarly contends that compassion is naturally elicited in response to excusing conditions, such as childhood abuse, and this emotion mollifies ordinary feelings of blame. But Kelly believes that clemency is justified because the perception of an excuse provides rational grounds for suspending or modifying or ordinary sense of antipathy. This view sees compassion as a natural, as well as normatively significant, aspect of blame. Like compassion, empathy may justify a mollification of ordinary blame.

More recently, Jesse Prinz (2011) has argued that empathy is not necessary for moral judgment, but there is an observed correlation between empathy and moral competence, or facility with making moral judgments. Antti Kauppinen (2017) argues, similarly, that even if empathy is not necessary for moral judgment, it is typically implicated in this process. Indeed, “people who lack the ability to put themselves in the place of others and feel for them… appear to have trouble with moral insight and appreciating the grounds of pro-social moral principles, even if their rational powers are largely intact” (2017: 20). If we see the reactive attitudes as a species of moral judgment – which seems nature – then there are good empirical grounds for seeing the reactive attitudes as mediated by empathy.

Notably, empathy is not elicited directly in response to the feelings of others without mediation by other factors. Rather, its role in moral cognition is influenced by cultural stereotypes. This is shown by research on empathy and judgments of guilt or innocence.

For example, perspective taking, which is a component in empathy, predicts a respondents’ tendency to assign guilt to a defendant in a sexual harassment case, and to rate the defendant’s behaviour high in severity, pervasiveness, and unwelcomeness (Zimmerman & Myers 2013). This partly explains why women (higher in perspective-taking) make harsher sexual harassment judgments than men (ibid.). Rape-myth acceptance (or acceptance of stereotypes about rape victims) and respondent gender are also known to influence judgments of rape cases (Sussenbach et al. 2016), though gender might be a heuristic for empathy. This suggests that empathy and rape-myth acceptance interact, probably implicitly (ibid)., to influence judgments of guilt in sexual harassment cases.

Similarly, researchers on racial bias find that “adults view black girls as less innocent and more adult-like than their white peers, especially in the age range of 5-14” (Epstein et al. 2017). This helps to explain why Black girls are five times more likely than white girls, and twice as likely as Black boys, to be suspended at school, three times as likely as white girls to be referred to juvenile detention, 20% more likely to be charged with a crime, 20% more likely to be detained, and so on. Black boys, too, are seen as less innocent than white boys (Goff et a;. 2014), but Black girls are still twice as vilified. This is probably because Black girls stand at the intersection of two types of bias: gender and racial. Both femininity and Blackness are stigmatized (albeit in different ways) in our culture, due to the pervasiveness of stereotypes rooted in European colonial patriarchy. There is also evidence that empathy mediates judgments of guilt and innocence in general; for example, when people are given more personal information about a victim (e.g., the victim’s name), they exhibit more empathy for the victim, and blame the victim less, while blaming the perpetrator more (Phyllis & Costa 2004). This does not capture all of the dynamics of blame, of course, but it suggests that empathy, mediated by cultural stereotypes, may play a role in the cultural vilification and collective blaming of Black girls.

 Supposing that empathy mediates our attributions of guilt and innocence in a way that, on balance, vilifies women, Black people, and (especially) Black women (when relevant stereotypes are salient), what can we infer about the characteristic role of empathy in our responsibility practice? Prima facie, the interplay between empathy and cultural stereotypes produces a system of relations that, on balance, disproportionally punishes historically disenfranchised groups, and disproportionally rewards privileged groups (i.e., cisgender white males) – at least in contexts in which implicit biases are salient. (For example, women are seen as less innocent when gender bias is primed). While individuals may be more or less susceptible to empathy-mediating cultural stereotypes, the data show that responsibility as a social practice disproportionally favors the privileged. This means that they are relatively indemnified against blame and punitive sanctions.

We can trace still more general trends by evaluating the role of empathy in broad social structures. Who bears the burden of empathizing with others in our culture? And who receives the benefits of that empathic investment in the common good? This will tell us something about social roles, and how those roles influence our blaming and praising habits.

There is evidence that white people are less empathetic to Black people than other white people. For example, research indicates that white observers show less physiological arousal in response to Black people’s pain than white people’s pain, and their lack of arousal correlates with their level of implicit racial bias (Forgiarini et al. 2011). This supports the theory that there is a “racial empathy gap” that favours white people on balance (ibid). This theory, in turn, helps to explain why Black people have less access to analgesics for a range of painful heath problems, and have trouble accessing needed prescriptions (Silverstein 2013). In another experiment, subjects reported greater empathy for a white defendant than a Black defendant in a larceny scenario, as well as making attributions that were more situational (and less characterological), and assigning more lenient punishments to the white defendant (Dovido & Reed 2002). This may be why Black people are more often subjected to police searches, police violence, drug-related arrests, mandatory minimum sentences, longer sentences, felony disenfranchisement, and so on (Quigley 2016). These practices highlight the problem of systemic racism in judicial, sentencing, and enforcement norms in the U.S. (ACLU 2014). The retributive attitudes that underlie these judicial norms are part of our broader responsibility practice. It is a reasonable conjecture, then, that the racial empathy gap distorts our responsibility practice, resulting in disproportionate blame toward Black people.

Next, who is expressing empathy in our culture, and who is caring for the vulnerable? On balance, women. Longitudinal research spanning almost 40 years shows that women demonstrate higher empathic response scores than men of the same age, and the difference increases with age (Mestre et al. 2009). Empathy is expressed in the activation of mirror neutrons, and women show more activation in the inferior frontal cortex, which involves mirror neurons, when asked to focus on their own feelings or the feelings of others, compared to men (Schult-Ruther et al. 2008). Women also demonstrate more contagious yawning, which implicates mirror neurons (Norscia et al. 2016). Is it any wonder, then, that daughters provide more care to their elder parents than sons, and invest more time and resources in caretaking (Grigoryeva 2017)? Or that women still do a majority of the domestic labour, even when they earn more than their male partner in a heterosexual relationship (Layonette & Crompton 2014)? The balance of evidence shows that women are the primary empaths, and primary caretakers, in the U.S.

Who are the beneficiaries of women’s disproportionate investment of empathy and care? Predominantly cisgender white men (CWM, for short), and other white women – but, notably, CWMs are contributing far less than their fair share to the “empathy economy.” In fact, when we consider white men’s socioeconomic status (SES), it becomes apparent that they are contributing far less than they owe, and collecting far more than they deserve, generating a massive imbalance in the distribution of empathic goods and services.

White men are the wealthiest social demographic; for every dollar earned by a CWM, white women earn 82 cents, Black men earn 73 cents, and Black women earn 67 cents (Nelson 2017, Patten 2016). This is significant because there is evidence that class bias, in addition to racial bias, plays a significant role in the justice system and other social institutions. The Sentencing Project (2013) reports that poverty exacerbates racial bias: poor black defendants received sentences on average 13% longer than other defendants. Obviously rich Americans have better healthcare and better access to medicine than poor Americans, even with the Affordable Care Act. Furthermore, one study purports to show that differential empathic responses to pain in Black and white people are due to class-bias, not racial bias per se (Trawalter et al. 2016). It is reasonable to think that empathy is mediated by class bias, resulting in a system of relations that favours the wealthy, i.e., predominantly CWMs. Thus, CWMs benefit indirectly from financial privilege.

High-SES individuals benefit more, but contribute less than other groups to the empathy economy. There is abundant evidence that wealthy people are empathically impaired. For example, the wealthy are less attuned to others’ “motivational relevance”; higher-SES people pay less attention to other people in everyday contexts, and become less attentive when primed with a human face, compared to less wealthy people (Dietze & Knowles 2016). Wealthy people report less compassion, and show less heart-rate deceleration (a common empathic response to others), in response to videos depicting suffering in others (Stellar et al. 2012); they rate low in scores of empathic accuracy, judge the emotions of others less accurately, and make less accurate inferences about emotions from images of facial movements (Kraus et al. 2016); they’re more likely to cut off other drivers at intersections and pedestrians at cross walks, even after making eye contact with the person (Piff et al. 2010); and so on. This suggests that wealthy people are not contributing fairly to the empathy economy, and they may even lack the reflective capacity to do so. (These studies measure relatively automatic, unconscious responses to social cues, which are not under the agent’s immediate reflective control). The critical point is that wealthy people – predominantly CWMs – are less empathic on average than other groups.

The resulting picture is an empathic economy (so to speak) in which women do a disproportionate amount of the empathic labour, CWMs receive a disproportionate amount of the empathic rewards, and Black people are relatively empathically disenfranchised. This “secondary” economy mirrors the financial economy in that CWMs are at the top. It also, by the same token, reinforces the financial economy, encouraging women to perform low-paying, undervalued empathic (caretaking) labour, and situating wealthy white men as the normal and nature recipients of this labour, as persons whose pain matters, who deserve better healthcare, and who are presumptively innocent and credible.

If empathy mediates our blaming attitudes and practices, as the evidence suggests, this is very troubling indeed, since these imbalances will tip the scales in favor of white men. They will result, that is, in an emotional economy that unfairly blames and sanctions Black people, while offering compassion, forgiveness, and indemnity against sanctions to CWMs.

The key implication for responsibility as a social practice is that the characteristic role of empathy in our system of social relationship is to encourage unfair and disproportionate blame toward Black people, and unfair tolerance and clemency toward CWMs, on balance. Women, meanwhile, are being (in effect) emotionally extorted, but their undervalued empathic labour, while beneficial in many context (such as when elder care is needed), reinforces this racist hierarchy, at least when racial stereotypes are salient. Women should not necessarily cease empathizing, but should carefully redirect their empathy.

Some would say that we should eliminate empathy from our moral lives, including our moral responsibility practice. Perhaps most notably, Paul Bloom, the author of “Against Empathy: The Case for Rational Compassion” (2016), argues that we should excise empathy from moral decision-making, but he allows that empathy may play an important role in other aspects of social life (Robinson 2017). Whether eliminating empathy is a realistic psychological proposal, it is clear that we must, at the very least, recalibrate empathy to effectuate a fairer distribution between privileged groups and historically disenfranchised groups, and we should encourage men to bear more of the empathic burden. This recalibration/redistribution approach may be combined with a program of limiting the role of empathy in moral cognition to whatever extent possible. Yet I worry that we cannot excise empathy from moral judgment without thereby banishing it from social life altogether, since the two are deeply imbricated; moral judgments are principally judgments about our relationships to other people: what we owe others, how we should treat them, etc. So if we excise empathy, perhaps we will lose the motivation to care for others. Still, the cultivation of “rational compassion” could perhaps mediate empathy in positive ways, in which case rational compassion and empathic recalibration would be complementary approaches. Perhaps we should temper empathy with rational compassion.

4. Communicative habits

If responsibility is indeed a conversational practice involving two or more speakers, it must involve communicative habits, such as lexical preferences and vocal register, used by the speakers involved in the exchange. This is clear enough. But these practices are not equally demographically distributed or value-neutral. On the contrary, they are unevenly distributed across demographics, and unequally valued based on their distribution patterns, and how closely they approximate to the idealized cisgender male norm. On scrutiny, these value imbalances generate inequalities in our responsibility practice.

In “Language and Woman’s Place” (1973), Robin Lakeoff argued that women have distinct (average) lexical habits, acquired through sex role socialization, and these habits are generally socially disvalued. Examples include: “weaker expletives (oh dear versus damn); trivializing adjectives (divine versus great); tag questions used to express speakers’ opinions (The way prices are rising is horrendous, isn’t it?); rising intonation in declaratives (as seen in the second part of the sequence, What’s for dinner? Roast beef?); and mitigated requests (Would you please close the door? Versus Close the door) (Tannen 1975: 10-18, cited in Kendall & Tannen 2015: 549). Later, Zimmerman and West found that men interrupt women more often than the reverse in ordinary conversations (1975, cited in Kendall & Tannen 2015: 549). Recent sociolinguistic research confirms that women’s communicative habits enhance cooperation whereas men’s enhance social dominance in general (Leaper, 1991; Mulac, Bradac, & Gibbons, 2001), and that, “on average, women use more expressive, tentative, and polite language than men…, especially in situations of conflict” (Basow & Rubenfield, 2003, cited in Merchant 2012). Furthermore, women are still interrupted more than men (Hancock & Rubin 2014), including in judicial arguments, even when controlling for seniority (Jacobi & Schweers 2017).

 Women also use distinct, acculturated vocal registers, which are also socially disvalued. These registers include creaky voice (“vocal fry”), breathy voice, whisper, and certain stylized intonations (Sicoli 2017). High instances of creaky voice have been observed amongst Chicanos as well as “upwardly mobile urban American women,” and this register is often criticized as being “annoying, irritating, or a fashion fad” (ibid., cf. Yuasa 2010, Wolk, Abdelli-Beruh & Slavin 2012). Research finds that people hold negative attitudes toward voices with vocal fry and positive attitudes to those without it (Abbey & Alison 2014). In addition, breathy voice is perceived as feminine (Borsel et al. 2009), and this perception can trigger implicit gender bias in contexts in which femininity is stigmatized, e.g., leadership positions (Hoyt & Burnett 2013) and job interviews (Latu et al. 2015). There is also evidence that people find speakers with lower-pitched voices (i.e., predominantly male voices) more attractive, competent, and trustworthy on average (Garber 2012).

In addition to lexical and vocal patterns, speakers can express emotions in their voice. It is well known that when women express anger, they are seen as less competent, whereas when men express anger, they are seen as more competent (Brescoll & Uhlmann 2008; Brescoll & Uhlmann, 2008; Tiedens, 2001). In one study, women who expressed a dissenting opinion with anger were perceived as less convincing, whereas men who expressed the same opinion with anger were perceived as more convincing (Salerno et al. 2016).

Women are not the only group to face discrimination as a result of their acculturated vocal habits. African American Vernacular English (AAVE), spoken by many Black people, is distinguished from so-called “Standard English” (spoken primarily by white people), and is given second-class status in the popular imagination. Standard English (SE) is “valued by the general public as being more logical, more precise, and even more beautiful than other varieties,” whereas “other, nonstandard varieties are stigmatized as corrupted forms of the standard and [in most jurisdictions] kept out of the classroom” (Siegel 1999: 701). (There are some exceptions, such as the Oakland Ebonics Resolution of 1996, which mandated instruction in AAVE for native vernacular speakers; but this policy was extremely controversial and met strong resistance [CNN 1997]). Many researchers believe that preference for SE in the classroom partly explains the racial gap in average reading achievement scores (Gill 2013); AAVE speakers essentially face discrimination because their native language is not accepted as legitimate in institutional contexts, and they are forced to conform to the white vernacular standard. White children, by contrast, are not expected to do all schooling in a non-vernacular second language.

Indeed, prejudice against AAVE extends to virtually all American social institutions. Research on housing discrimination finds that applicants face discrimination even when they never meet the rental agent in person, when the applicants use AAVE or have a feminine voice on the phone; Black women were found to face the greatest discrimination (Massey & Lundy 2001). Another study finds that both Black and white managers discriminate against job applicants who use AAVE, or SE with some elements of AAVE, in phone interviews; they rated the Black applicants as being less intelligent, less ambitious, and less qualified (Henderson 2001). It is likely that Black and white managers showed similar levels of implicit racial bias (IRB) because IRB is acculturated in everyone, though white people demonstrate higher IRB on implicit association tests (Project Implicit), particularly those who live in southern and eastern states, since IRB is geographically concentrated (Mooney 2014). Bias against AAVE may also, according to linguist John Rickford, give rise to systematically racist judicial norms, as evidenced in the acquittal of George Zimmerman in the fatal shooting of Trayvon Martin (Rigogliosi 2014).

 Trans people also face discrimination on the basis of their vocal quality. Trans folks may may or may not have gender-normative voices, but in either case, they face high rates of discrimination on the basis of vocal style. If they have non-normative vocal qualities (e.g., speaking frequency, resonance, pitch, breathiness), they face characteristic transphobic discrimination, such as refusal of medical care, housing, goods and services, workplace discrimination, and high incidences of violence (Whittle et al. 2007). If they have normative feminine voices, they face the kinds of discrimination characteristically experienced by feminine women. Those with normative masculine voices are most likely to escape voice-based discrimination, but this still leaves a majority of the trans population open to unfair treatment on the basis of their communicative habits.

Because communicative habits are largely gendered, sexed, and raced in our culture, and the characteristic communicative practices of CWMs are perceived as the normative ideal in most contexts, all other social groups face relative discrimination on the basis of their characteristic speech habits. Speakers of African American Vernacular English receive lower academic scores, are perceived as less intelligent, ambitious, and qualified for jobs, and are seen as less worthy of housing, amongst other indignities; women are perceived as less attractive, competent, and trustworthy than male speakers, and are interrupted and silenced more often, indicating that their speech is less valued and less believed; trans people face discrimination in housing, healthcare, employment, and so on. These group-level inequalities suggest that we see CWMs, on balance, as more responsible than other groups across a variety of key dimensions: as renters, homeowner, medical patients, employees, students, academics, coworkers, and knowers, or communicators of valuable information (see Fricker 2007). We see this group as presumptively responsible in these respects, and are reluctant to blame, distrust, or sanction them.

Hence, our responsibility practice is tipped in favour of speakers whose lexical habits, vocal register, and diction is stereotypically cisgender, white, and male.

5. Non-verbal communication & the body

Conversation can take place on the phone or a voice chat app, but it often occurs in person. This makes room for habits of non-verbal communication and physical appearance to play a role in speaker perception.

Feminist philosophers have analyzed how women’s physical embodiment is inscribed by patriarchy, in such a way that women’s bodies, on average, are smaller than men’s, and take up less space (Beauvoir 1964); women’s gender-normative clothing, including dresses, fitted clothing, long hair, and high-heel shoes, are more constraining, and women’s normative body language is confined, tentative, and uncertain (Young 1990: 145-147). Women do not “manspread,” or extend their bodies in space, to the same extend as cisgender men. Fat women are perceived as “unruly” and “problematic,” because they do not conform to the ideal of femininity: small, delicate, and “disciplined” (Gay 2017). Weakness, childishness, and vulnerability are sexualized in women but not in men (Wade 2013).

Women who violate the norms of feminine embodiment – who take up physical space – are punished. Weight-based discrimination affects people’s employment prospects, educational experiences, romantic relationships, health care accessibility, and mental health treatment, but it disproportionally harms women (Fikkan & Rothblum 2012). For example, men report significantly less desire to work with a fat woman, but show no similar bias against working with a fat man (Jasper & Klassen 1990). Anecdotally, it appears that women who “manspread” attract more stares and glares, whereas men who do the same are seen as more attractive (Petter 2017). Women who wear high heels (Gueguen 2016) and have long hair (Mesco & Beresczkei 2013) are more sexualized by straight men.

This suggests that women who violate feminine norms of non-verbal communication and physical embodiment are perceived as being less responsible across several key domains, including; as homeowners, tenants, employees, and coworkers. They also are seen as less eligible sexual partners, meaning that they are punished or sanctioned as sexual agents. While this population may not be explicitly criticized for defying binary gender norms, they are, in effect, treated with distrust and antipathy in relevant domains.

In addition, people with gender-nonconforming mannerisms – such as men with feminine body language and bodily comportment, or women with masculine features – are liable to similar kinds of discrimination, viz., in housing, employment, and relationship opportunities. They face similar distrust and antipathy in central aspects of their lives.

6. Relational equality

I have argued that the role of empathy, lexical preferences, and physical embodiment in our responsibility practice – defined as an interrelated network of conversational exchanges – biases this practice in favour of CWMs and against other social groups, at least, in contexts in which relevant stereotypes are salient. (For example, women are disfavored in leadership roles, but not in caretaking roles, since women are perceived as “natural caretakers”). These inequalities create biased responsibility attributions across key institutional domains (medicine, the law, education), as well as our interpersonal relationships.

How can we rectify these distorting biases? Clearly, we cannot treat our responsibility practice as independent of our broader context of institutional and social interactions, since it is a constitutive part of this human “ecosystem,” and cannot, in practice, be dissociated from it, or analyzed in isolation without remainder. The biases that affect broader social systems necessarily affect responsibility as a part of that network. Attempts to isolate constitutive features of responsibility, without taking into account this broader picture, may be illuminating in their own right, but they are necessarily incomplete.

There are various proposals for how to address inequalities and biases, some of which I have discussed elsewhere, but addressing these questions is beyond the scope of the present analysis. Instead, I will make some closing statements about the relation of responsibility to equality.

While philosophers have devoted much time and attention to analyzing the nature of excusing and exempting conditions and the psychology of blame and praise, they only recently revived the idea that responsibility is for something, i.e., has a particular function in our shared social life. Functionalist proposals include that blame enhances moral agency (Vargas 2008), protests moral violations (Smith 2012), expresses certain moral values (Franklin 2012), and shields moral communities from harm (Bell 2012). These proposals are all prima facie compelling, but we might seek a unifying thread that ties them together. If I am right that inequality systematically distorts our perceptions and judgments of responsibility, then one of the central aims promoted by any moral conversationalist should be equality, or the reduction of bias. Hence, at least one of the functions of responsibility – perhaps the main function – should be to cultivate equality, in order for the practice of holding-responsible to be fair, equal, and mutually respectful. If this practice is not fair, then disenfranchised groups have no good reason to want to participate in it, and the “moral community” that Strawson envisioned is impossible. As things stand, many people are morally disenfranchised and have very little reason to trust others or expect responsibility attributions to be rational. Hence, enhancing equality is a primary goal for anyone interests in having a legitimate responsibility practice, as opposed to a weak facsimile in which blame and praise reinforce historical oppressions.

Moreover, enhancing equality helps to enhance other, higher-order goals, including: enhancing the moral agency of the group, protesting moral violations (insofar as they pertain to fairness, equality, and the dignity of persons), expressing a commitment to important moral values, and shielding the moral community from systemic injustice. So, enhancing equality enhances the aims touted by other functionalists. Indeed, this aim may be logically prior to the others, insofar as having a community in which trust and cooperation are expected and exchanged is prerequisite to realizing these other moral goals.

The importance of equality outside of financial transactions is defended by Elizabeth Anderson (2015), who promotes the value of relational equality, or equality of authority, esteem, and standing. Relational equality ensures justice on an institutional and interpersonal level. This view is distinguished from classic distributive theories of justice (e.g., Ralws), which focus narrowly on the fair distribution of financial resources, ignoring non-monetary values such as dignity and respect. These distributive theories, on scrutiny, are condescending and disrespectful to the socially disenfranchised and marginalized, who are viewed as, in effect, charity cases who need handouts from generous benefactors. Instead of seeing the vulnerable as pitiable and wretched, justice in the true sense requires that we see the least well-off as equal citizens, deserving of equal respect and standing with other citizens. This perspective on what we owe to others treats the historically disenfranchised with the dignity that they deserve as persons.

Notably, one of the social practices within the purview of relational equality is our responsibility practice, and this practice fails to treat historically disenfranchised groups with the respect that they deserve, due to the influence of hegemonic stereotypes. To achieve the ideal of relational equality, then, we need to rectify inequalities within our responsibility practice. But these two goals are co-implicated in a positive feedback loop, since achieving relational equality requires purging our responsibility practice of harmful biases, and eliminating biases from our responsibility practice requires relational equality. We must pursue these ends, then, at the same time. The thing to do is to be vigilant about cultural stereotypes and myths, and debunk them whenever possible, and as effectively as possible, as this will advance both ends simultaneously. In other words, to advance relational equality within our responsibility practice and elsewhere, we need to look beyond any individual context, to the role of cultural stereotypes in broader cultural narratives and public discourses. The project, then, is an extremely copious and interdisciplinary one, which requires cooperation and constant effort.

7. Concluding remarks


Responsibilities (moral, epistemic, practical), and why they matter (relational equality).


There are various kinds of responsibility identified in the philosophical literature. These include moral responsibility (e.g., Strawson, Wolf, Fischer), epistemic responsibility (e.g., Fricker, Medina), and responsibility as a kind of self-efficacy (e.g., Waller). It may not be obvious how these dimensions of responsibility intersect, but they are all tied to personhood, and to evaluative attitudes that respond to features of personhood and relevant background conditions. Without these capacities, an agent is lacking in some critical feature of personhood, something that rational humans value—either moral, epistemic, or practical agency; and people who lack these features without an excuse or extenuating circumstance are amenable to negative evaluation, whether moral, epistemic, or practical. People who excel in these capacities, particularly in the face of adversity, are praiseworthy, epistemically virtuous, or self-efficacious. They deserve laudatory treatment. These exceptional individuals, too, are capable of having functional relationships and achieving worthy goals, and for this reason, they are likely to enjoy higher wellbeing than those who lack these dimensions of human agency. These are people we want to invest in because they are reliable, versatile, and responsive to facts.

These three capacities are interrelated in that they all function to bring about a positive achievement – a positive goal or outcome – and a deficit in any one facet could undermine the attainment of this goal or outcome, as well as the cultivation of the other facets. For example, if Jeff the Jerk is antisocial, he may also be sexist, because in a patriarchal society it is easier to be selectively antisocial to vulnerable groups like women, and to harass and discriminate against precisely those groups. If Jeff is epistemically insensitive to women’s credibility, he is not only epistemically flawed, but also morally flawed (sexist, misogynistic). If Jeff is a CEO who wants to run his company effectively, but he discounts feedback from women due to epistemic insensitivity (an epistemic flaw) or sexism (a moral-epistemic flaw), he is going to discount valuable perspectives in corporate decision-making, undermining his own pragmatic goals as CEO (see Sandra Harding 2015 on the collective effects of ignorance). While a person can be good but epistemically flawed, like Huck Finn (see Arpaly 2016), epistemic sensitivity makes moral sensitivity more likely and more robust across circumstances and time. If Huck Finn had rebuked slavery (instead of thinking it was justified), he would have been disposed to act appropriately in response to all African Americans, not just his friend Jim. He helped Jim, which was virtuous on Nomy Arpaly’s view, but how would he have responded to other enslaved persons, with whom he had no prior acquaintance? Epistemic sensitivity seems to reinforce moral virtue, and vice versa. People who care about morality are more likely to care about how their epistemic profile affects oppressed groups, and how their epistemic deficits could potentially be remediated (e.g., by education, exposure to countersterotypical exemplars, the adoption of evidence-based policies, etc.). And epistemically flawed people are likely to treat marginalized groups in immoral ways because they don’t care to cultivate epistemic virtues. Furthermore, morally and epistemically irresponsible people will be poor at achieving their pragmatic goals just in case they discount evidence or distrust experts and knowledgeable people on prejudiced grounds. Clearly, those with volitional deficits (e.g., low self-efficacy) will be poor at initiating and executing morally and epistemically responsible plans, just because they are poor at executing any plans. In this way, the three salient dimensions of responsibility are deeply intertwined.

The upshot is that people who are strong in one facet of responsibility are likely to be strong in all facets, and a deficit in one facet is likely to impair the others. This is something akin to Socrates’ ‘unity of the virtues’ thesis, but applied to dimensions of responsibility. Yet it is weaker than Socrates’ thesis, because it only claims that each dimension makes the others more robust, or more resilient across different circumstances, not that each dimension is a necessary prerequisite for the others. A person could be morally responsible in one domain without epistemic or pragmatic responsibility, but in an unfamiliar situation, epistemic sensitivity to the demands of the situation and self-efficacy could serve to enhance moral responsibility. For example, I might be morally upstanding in my day-to-day life, but if I were to move to a different country with radically different cultural norms, I would have to learn the ropes pretty quickly to avoid committing unintentional norm violations. Jeff might have grown up in a culture infused with toxic masculinity, but he had better pay attention to changing cultural standards if he wants to avoid committing workplace harassment, and he should apologize for past transgressions. In both cases, the agent has to update his epistemic profile to respond sensitivity to accessible moral norms. In this way, heightened sensitivity across each dimension of agency enhances the robustness of the other aspects.

That said, none of these dimensions of agency is reducible or eliminable; each aspect can be individuated on the basis of its object, or the thing it tracks (moral, epistemic, or practical facts). While these capacities are ontologically distinct, they are, in effect, implicated in a positive feedback loop in which each dimension positively reinforces the others.

Responsibility across all three dimensions is also vulnerable to the same undermining or defeating factors. These factors can be congenital, but more often than not they are environmental, and environmental factors always mediate the expression of overt behaviour. To give an example: there is mounting evidence that adverse childhood experiences (ACEs), such as emotional, physical, and sexual abuse, intimate partner violence, and poverty, can impair responsibility across all three dimensions. People high in ACEs (i) are more likely to commit violent criminal offenses like rape and assault as adults (Craparo 2017); (ii) are less capable of participating in epistemically valuable trust relationships (Ijzendoorn et al. 2011), and (iii) are more susceptible to depressive disorders (Chapman et al. 2004), alcoholism (Rothman et al. 2008), attempted suicide (Dube et al. 2001), and other behaviours that impair self-efficacy and practical achievement. These dysfunctions, rooted in ACES, undermine the achievement of moral, epistemic, and pragmatic goals, and in this sense they can be seen as deficits in responsibility. Identifying these factors helps us predict and diagnose responsibility-relevant deficits in populations with responsibility-umpiring causal histories. These populations are also prone to adverse health outcomes like ischemic heart disease, cancer, and chronic lung disease (Felitti 1998). There are positive correlations, in other words, between ACEs and responsibility deficits, and between ACEs and poor health outcomes.

Responsibility may be a mediating psychometric factor between childhood conditions and certain life outcomes, just as self-efficacy is a mediating psychometric factor between situational adversity and avolition on social cognition theory (Bandura 2006). Unsurprisingly, people low in responsibility due to adverse experiences tend to be less healthy and less satisfied than people high in responsibility. But more importantly for out purposes, responsibility mediates our interpersonal relationships and influences how we respond to others—whether with kindness or antisociality, with trust or distrust, with avolition or engagement. Hence, responsibility enables us to maintain and promote relations of equality.

Further, responsibility on a social cognition model is a biopsychosocial capacity, sensitive to situational factors. Thus, while it can be impaired by ACEs, it can also be remediated by trauma-informed interventions, such as CBT, heathy relationships, community support, affordable housing, and so on. These interventions can enhance responsibility, and thus relational competency. When people experience responsibility deficits because of misfortune or injustice, they are entitled to community support and public health resources.

But people with functional childhoods and privileged lives can also have significant responsibility deficits. For example, many privileged white people with no history of trauma are high in implicit bias, and implicit bias can motivate prejudiced behaviour. This behaviour is unethical, and it can also have adverse epistemic consequences, such as prompting the hiring of unqualified white candidates (see Bertran & Mullainathan 2013); and it can have adverse pragmatic consequences, such as undermining corporate decision-making. (This is not to say that all privileged white people are high in implicit bias, but white people show higher implicit racial bias than other groups on the Project Implicit IAT, and they benefit from implicit bias against people of color, which creates de facto affirmative action for white people). Moreover, many privileged people also have explicit biases, whether due to ill will or indifference to the interests of disadvantaged groups. These biases similarly cause or constitute moral, epistemic, and pragmatic deficits, undermining the attainment of relevant goals. Unlike the role of ACEs, however, motivated irrationality and moral indifference are not public health problems that call for rehabilitative interventions. The government should intervene to reduce the prevalence of implicit bias in our social institutions (see Hurley 2006), but this is not because privileged people deserve public resources; it is because disadvantaged people do.

Deciding how to respond to responsibility deficits is not a straightforward matter, particularly as there are two oppositional approaches recommended by research on agency and public health. We can blame someone for a responsibility deficit, or we can offer a remediating intervention. While we can, in principle, do both, there are putative tensions between the blaming response and the remediating response. If someone is in treatment for an addiction, it may be counterproductive to blame the person for her addictive impulses or for past alcohol-induced behaviour, if blame would hinder the person’s recovery. Furthermore, blame may be unwarranted if the person’s deficits are due to oppressive circumstances such as ACEs. We would not blame someone for failing an academic test because the person was barred from attending school, and by parity of reasoning, we should not blame someone for lacking responsibility due to childhood trauma. ACEs are a paradigmatic example of a non-culpable deficit, as children have little autonomy or volitional control, so their psychological development is not up to them. For traumatized and oppressed people, the rehabilitative approach may be more fitting.

Privileged people who lack responsibility due to their own life choices, on the other hand, are better candidates for blame, as they may not want to be rehabilitated, they may not respond well to rehabilitative interventions, and they are the authors of their own destinies (if anyone is). Blame, exclusion, and sanctions are perhaps the best approach to such people.

These claims highlight important considerations, but they fall short of providing a systematic method for attributing blame and praise. I propose the following framework, which fits with the above impressions: blame and praise should serve the purpose of enhancing relations of equality (see Elizabeth Anderson 2013), and thus, of undermining oppression. This provides a way of systematizing our impressions across cases. Victims of ACEs are victims of a type of oppression—traumatic experiences and/or poverty—and to blame them, instead of their oppressors, may serve to reinforce systemic injustice, particularly if this is part of a broader victim-blaming narrative. Offering rehabilitative interventions, by contrast, may enhance the recipient’s ability to participate fully in relations of equal standing, esteem, and authority with others, if these interventions enhance the person’s responsibility. Privileged people who lack responsibility, on the contrary, have more than their fair share of status, respect, and resources, and may be insensitive to rehabilitative interventions, making blame the fitting response. A blaming response may also serve to condemn their role in hierarchies of oppression and alert others to their motivational deficits, contributing to an egalitarian social narrative, and protecting potential victims from their vicious behaviour. The role of praise and blame in these cases supports egalitarian aims, and this is what justifies its differential deployment.

These claims are still rather impressionistic, and require empirical support to be validated. If praise and blame, as I claim, ought to serve relational equality, we need to know more about how these attitudes affect people in light of their motivational profile, learning history, and social circumstances. Then we can draw accurate generalizations  about what types of response are fitting for what type of person and in what context. That said, when we hold people responsible in our daily lives, we typically do so on the basis of incomplete data. So, schematic, ambivalent attributions might be okay, and even inevitable, if we are acting under time constraints (as we do). That said, even if we cannot know everything about a person’s circumstances, we should at least be mindful of the purpose our reactive attitudes are meant to serve when deciding how to express them. On my view, that purpose is to construct and reinforce relations of equality. To be responsible critics, we should keep this in mind when blaming and praising people.

In sum, responsibilities are valuable because they enable us to participate in relations of equality; that is, responsible people are in a position to contribute to a society of equals, one in which people respect each other’s moral and epistemic standing, and take the initiative to pursue and protect egalitarian goals. Responsible people do not unfairly oppress others, or undermine their own agential capacities by pursuing irresponsible and counterproductive agendas. Responsibility is also valuable because it can improve health outcomes, if it enables us to respond to situations and relationships in an adaptive way; but positive health outcomes are a byproduct of responsibility, not its end goal.

Responsibility, Epistemic Confidence, and Trust


In my last post, I argued that severe deficits of epistemic confidence can undermine responsible agency by undermining a person’s ability to form resolutions and have a deep self. In this post, I want to discuss a related notion: trust. In writing about epistemic confidence, Miranda Fricker (2007) says that people who conspicuously lack epistemic confidence are perceived as less competent and less trustworthy. Being seen as less trustworthy undermines a person’s epistemic confidence, which in turn undermines the person’s agency or competency. Trust, epistemic confidence, and agency are thus related in a positive feedback loop. This is illustrated in the experiment on expectancy effect, in which certain students were randomly designated as academically gifted, and the teacher’s trust in the students’ academic competency actually improved their competency (as measured by test scores) over the course of the year (Rosenthal & Jacobson 1996, cited by Fricker 2007: 56).

In this post, I want to look more closely at trust and its relation to responsible agency.

Victoria McGeer also writes about trust. She argues the ‘substantial trust’—trust that goes beyond the evidence and abjures strategic judgment—enhances the trustee’s responsible agency (2008).[1] Substantial trust ‘goes beyond the evidence’ in the sense that it embodies a belief in the trustee’s moral worth that isn’t supported by the balance of evidence; and it ‘abjures strategic judgment’ in that it entails a refusal to evaluate the trustee’s worth on the basis of evidence. That is, trustors don’t meticulously scrutinize the evidence regarding their friend’s moral qualities; they take a leap of faith in favour of the friend’s potential to be good. To illustrate this epistemic state, if my friend is accused of bribery, I exhibit substantial trust if I’m biased in favour of her innocence, in spite of any evidence to the contrary. When we substantively trust someone, we refuse to judge her on evidential grounds.

A central element of substantial trust on McGeer’s view is hope: in trusting a friend, we hope the person will live up to our optimistic expectations of her moral worth, but we don’t know if she will. Yet substantive trust can’t (or shouldn’t) be delusory: if the evidence confirms our friend’s guilt beyond doubt, we shouldn’t trust in the person’s innocence; but it would still be reasonable in this case to trust in our friend’s capacity to improve. In this way, substantive trust is relatively resistant to disappointment: even if a friend fails several times, we can continue to trust in the person’s basic capacity to live up to our hopes. We trust that the person can gain new capacities or build on existing capacities to embody our ideal. Only in the face of repeated disappointment does substantive trust become irrational. ‘Irrational’ trust on McGeer’s view is pointless; it doesn’t reliably contribute to the trustee’s agency.

Substantive trust enhances the trustee’s agency because it “has a galvanizing effect on how trustees see themselves, as trustors avowedly do, in the fullest of their potential” (McGeer 2008: 252). That is, our trust inspires confidence in the trustee, who begins to believe in herself.

This picture of trust as agency-enhancing interests me for 3 reasons, which I’ll elaborate briefly here.

  1. Epistemic confidence: the mediating variable between trust and responsible agency

McGeer’s account helps to explain how epistemic confidence is related to responsible agency: substantial trust (when assimilated to Fricker’s moral epistemology) inspires epistemic confidence, which (in the right degree) facilitates responsible agency. The right degree, as per my last post, is midway between between epistemic insecurity and epistemic arrogance; it’s neither too much nor too little self-regard. Epistemic confidence, then, is the mediating variable between trust and responsible agency. McGeer doesn’t explicitly mention ‘epistemic confidence,’ but she’s interested in elucidating the psychological mechanism whereby trust enhances responsibility. She rejects Pettit’s theory (1995) that trust incites a desire for approval, as this isn’t a ‘morally decent’ motive, befitting of the trust relationship (2008: 252). Instead, McGeer proposes that trust ‘galvanizes’ the trustee to see herself in a more positive light—through the trustor’s eyes. The resultant state—let’s call it positive self-regard—motivates the trustee to aspire to a higher standard of conduct.

Positive self-regard can be seen as a weak form of epistemic confidence—an aspirational kind. Whereas epistemic confidence is a positive belief in one’s merit or abilities, self-regard (in McGeer’s sense) appears to be faith in one’s (as yet unproven) merits and abilities. But self-esteem and epistemic confidence are of a kind: one is just firmer than the other. So, we can see positive self-regard as a weak form of epistemic confidence, and both states as intermediary between two epistemic defects: epistemic insecurity and epistemic arrogance. These epistemic virtues—self-esteem and epistemic confidence—are positively correlated with responsible agency, in the following sense: they enhance the trustee’s confidence in herself, and thus her ability to have firm beliefs and values (or convictions) about herself, and to act on those states. Having convictions prevents people from being ‘wantons,’ akratics, and irresolute people—paradigms of irresponsibility or weak responsibility. Responsibility is enhanced by belief in oneself, and this belief tends to confer self-control, willpower, and resilience—competencies implicated in or constitutive of fully responsible agency.

These related virtues—positive self-regard and epistemic confidence—might serve slightly different purposes; specifically, self-esteem might be particularly adaptive in adverse circumstances where a positive outcome is unlikely (but possible), whereas epistemic confidence might be more fitting when success is reasonably probable; but both states facilitate responsibility. Trust is fitting, therefore, when it’s likely to enhance responsibility by either of these means. In other words, we’re rational to trust someone when our trusting attitude reliably confers agency-conducive epistemic virtues. This allows us to say (consistent with McGeer’s view) that trust is a ‘rational’ attitude even if it goes against the evidence, insofar as it tends to foster agency in the trustee. Trusting in someone ‘irrationally’ would mean trusting in someone who can’t reasonably be expected to live up to our ideal; in that case, we’re merely wishing (not trusting) that the person could be better. Trust is also irrational if the trustee is overconfident, since in that case, our trust is either wasted or positively harmful: it’s likely to increase the person’s epistemic narcissism.

On this (basically functionalist) account of trust, epistemic confidence is counterfactually dependent on trust in the following sense: it wouldn’t exist without some initial investment of trust, but it can become increasingly self-sustaining and self-perpetuating over time. That is, people who never receive trust probably (as a matter of statistical probability) won’t develop epistemic confidence, but people who do receive trust may become increasingly self-trusting and self-sufficient. This claim is based in part on facts about ordinary human psychology: As a matter of fact, trust tends to confer epistemic confidence in psychologically normal humans, which enhances responsibility as a measure of resoluteness, willpower, and resilience. This psychological picture is suggested (though not explicitly articulated) by McGeer and Fricker, who cite developmental and child education studies showing that trust from an adult inspires confidence and competency in children. (This is sometimes called ‘Pygmalion effect’). Fricker cites the famous teacher expectation study (Rosenthal & Jacobson 1996), and McGeer cites research in developmental psychology showing that children who receive support from parents—‘parental scaffolding,’ as she calls it (2007: 249)—develop stronger powers of agency than deprived and neglected children. This research suggests that agency typically, in ordinary humans, depends on positive self-regard, which depends on a non-trivial investment of trust, especially during a person’s formative years. Subsequent trusting relationships, however, can compensate for deficits in childhood, as other research indicates—for example, research on therapy showing how positive therapeutic relationships can remediate symptoms of childhood trauma (Pearlman & Saakvitne 1995). This is how I suggest we perceive the trust-epistemic confidence relationship: epistemic confidence is counterfactually dependent on a non-trivial investment of trust in psychologically normal people, but can eventually become relatively (though not completely) self-sustaining; epistemic virtues inculcated by trust typically confer strong(er) agency.

This discussion suggests a particular taxonomy of epistemic states related to trust and agency. Specifically, I’ve said that trust catalyses three closely-related epistemic virtues: positive self-regard, epistemic confidence, and epistemic courage. These states are increasingly robust epistemic virtues, which support our ability to form resolutions, exercise willpower, and act resiliently. At either end of thus spectrum is an epistemic defect: on one side, epistemic insecurity (a paucity of epistemic confidence), and on the other side, epistemic arrogance (a superabundance of epistemic confidence). These defects undermine agency for different reasons: epistemic insecurity undermines our ability to form and act on convictions, and epistemic arrogance undermines our ability to adequately consider evidence for and against our beliefs, inciting us to favour our prior assumptions come what may. (That is, it spurs self-serving bias and confirmation bias). These vices thus undermine our ability to have a deep self and to exercise moderate control over our deep self, respectively.

This is one possible epistemic framework for responsible agency—the one that I’ve settled on. I think that more work can be done here, viz., at the intersection of responsibility and epistemology (especially social/feminist epistemology, which is relational in nature). We can call this intersection ‘the epistemology of moral responsibility’. This is promising area for future research, I think, and it may be of interest to neuroscientifically-inclined philosophers, inasmuch as these epistemic states are amenable to neuroscientific description.

  1. Responsibility as ‘external’ or ‘distributed.’ 

I’m also interested in McGeer’s account because (I think) it poses a challenge to classic theories of responsible agency that are relatively ‘atomistic’ (Vargas 2013) or ‘internalist’ (Hurley 2011). Classic accounts include Frankfurt’s (1971), on which responsibility is a matter of being able to form higher-order volitions consistent with one’s lower-order desires, and Fischer’s (2006, 2011), on which responsibility is a matter of being moderately responsive to reasons. These are different types of theory (one is character-based and the other is control-based, as typically construed), but they both emphasize the internal properties of agents to a greater extent than McGeer’s theory of trust, and so they can be regarded as comparatively ‘internalistic.’ (I’ve adopted aspect of these theories here—the idea that responsible agency is a function of deep-selfhood and reasons-responsiveness—but I’m going to to suggest that these capacities are more ‘extended’ than classic accounts imply).

Internalism should be seen as a matter of degree: most theories of responsibility treat some background factors as responsibility-relevant—for example, neuroscientific intervention (Mele 1995). But classic theorists usually think that exogenous factors are only relevant insofar as they intervene on the ‘actual sequence’ of the agent’s deliberation. For example, Fischer holds that clandestine brainwashing impairs responsibility because it operates on the agent’s actual motivational profile, dramatically altering it; but a ‘counterfactual device,’ that would have intervened had the agent deliberated differently is ‘bracketed’ as irrelevant (for more on this, see Levy 2008). Frankfurt, too, sees these counterfactual conditions as irrelevant.

McGeer’s theory is comparatively ‘externalistic’ in that it (implicitly, at least) construes counterfactual interveners as relevant to responsibility (qua trust-fittingness). We can’t, on her view, ‘bracket’ these counterfactual conditions when considering whether someone is trustworthy. This is because when we substantially trust someone, we (implicity) judge the person by what she could be in a nearby possible world—one in which she’s better than she is. This is implied by the hopeful optimism intrinsic to substantial trust—we don’t see the trustee as she is (at least, in paradigm cases), but rather as she would be if she succeeded in translating our trust into ideal self-regard. Moreover, when someone fails to live up to our optimistic expectations, we don’t immediately withdraw our trust, since substantial trust is inherently resilient. Trust, then, doesn’t always track a person’s real-world capacity for control or real-world quality of will; it sometmes tracks the person’s potential to improve, not based on evidence but on hopeful optimism. Trust, then, is a form of responsibility (a reactive attitude) that isn’t constrained by considerations about a person’s real-world or actual-sequence capacities at the time of action—when the trustee did something good or bad. It considers the person as she is in a nearby possible world or as she may become in the future.

This sets McGeer’s account apart from classic ‘actual-world’ or ‘actual sequence’ theories, because substantial trust treats counterfactual possibilities—in which the agent has a different kind of self-regard—as morally relevant. The trust relationship itself can be seen as a ‘counterfactual enabler’ in Levy’s terms (2008), in that it enables the trustee to gain a capacity, if the person succeeds in internalizing the proffered trust. But these transformative effects aren’t countenanced as legitimate considerations on classic views of responsibility. Also importantly, the trust relationship is distributed between two people, not intrinsic to the trustee; if it’s withdrawn at a critical stage of development, it undermines the cultivation of positive self-regard and agency. This is another ‘externalist’ aspect to trust: it implicates two or more people’s agencies. So trust is ‘externalistic’ in at least these two aspects: it depends upon counterfactual scenarios and it implicates two agents.

  1. Responsibility as care-based (non-retributive) and forward-looking

Substantial trust also challenges two other familiar approaches to responsibility: the retributive view and the backward-looking view.

Retributivism is, in very simple terms, the view that those who commit a wrongful action deserve punitive attitudes (blame, disapprobation, resentment) and those who perform an excellent action deserve rewards (praise, approbation). (I won’t consider more complex versions of retributivism: this one will be my only target). This is a very natural way of thinking about the reactive attitudes, and it seems to be Strawson’s understanding. He implies that those who fail to conform to reasonable social expectations deserve punitive attitudes, unless there’s an excusing or exempting condition (e.g., hypnosis, severe psychosis).

Substantial trust challenges this neat binary by holding that a person who falls short of our aspirational norms still ‘deserves’ trust, if trust is likely to instil positive self-regard across a reasonable time scale. That is, continuance of trust is fitting when someone makes a “one-off” mistake, as substantial trust is an “on-going activity” that’s resilient in the face of moderate set-backs (McGeer 2008: 247). Hence, we can’t simply say that someone who surpasses our expectations thereby warrants praise and someone who breaches our trust thereby warrants blame, as per the standard desert-based picture. This doesn’t capture the essence of trust. Rather, we withdraw or modify our trusting disposition only when someone repeatedly or catastrophically disappoints us, rendering trust pointless and irrational. Since substantial trust is aspirational at its core, substandard conduct on the trustee’s part doesn’t compel us to automatically withdraw our trust and assume a retributive stance: we’re licensed to suspend blame in the hope that the person will improve.

This is related to the fact that substantial trust is a forward-looking attitude. Most theories of responsibility are backward-looking, meaning that they attribute responsibility (praise/blame) on the basis of an agent’s capacities at the time of action, i.e., some time in the past. Frankfurt’s and Fischer’s views are like this: if someone had (a) a certain motivational structure, or (b) reasons-responsiveness when performing a certain action A, the person is thereby responsible for A. Trust, however, isn’t deployed solely on the basis of someone’s past motivational psychology and conduct; it’s also deployed on the basis of the trustee’s ongoing and fluid potential: we can trust someone who doesn’t (presently) have the capacity to improve. Trust, that is, outstrips the trustee’s current capacities at any given time.

As McGeer points out, we don’t (paradigmatically) invest trust in someone on a calculated judgment that the person will ‘earn’ our trust (as Pettit thinks), as this would be perverse and ‘manipulative’ (2008: 252). Rather, we trust someone as a way of empowering the person. Another way of putting this, I think, is to say that we trust someone for that person’s own sake. This interpretation of trust has affinities with Claudia Card’s (1996) care-based approach to responsibility, on which responsibility serves the function of expressing care to the target agent. It also resembles Vargas’ agency-cultivation model (2013), which reflects a concern for the target’s wellbeing (at least, it’s amenable to this reading). This care-based orientation is very different from the retributive rationale, and it’s also not backward-looking: responsibility attributions are meant to enhance or empower the recipient, not to punish her for past misdeeds. McGeer’s account of trust thus fits better with consequentialist theories rather than retributive ones, and it seems to embody a care ethos—trust is an essentially caring attitude. It seems to be essential to trust that it be care-based—or at least forward-looking; any other interpretation is simply conceptually mistaken.

I think that this is the correct way to think about responsibility in general (i.e., as consequentialist); but even if this isn’t the whole story (arguably there are many incommensurable but correct theories of responsibility—see Doris 2015 on ‘pluralism’), this seems to be a necessary way of seeing at least one facet of responsibility: trust. This means (at a minimum) that not all of our responsibility-constitutive reactive attitudes are retributive.



[1] McGeer says that substantial trust fosters ‘more responsible and responsive trustworthy behaviour’ (2008). I’m just going to say that it fosters ‘responsible agency,’ and I’ll make a case for this more general claim in this post. It’s not hard to see how trust can enhance responsible agency: if we trust in our potential to achieve a desired outcome, we’re better able to achieve that outcome (under success-conducive circumstances, which I’ll leave vague).

Moral enhancements 2


In my last post on moral enhancements, I considered whether there is a duty to enhance oneself and others, and correspondingly, whether one can be blameworthy for failing to fulfil this duty. I said that this is a complicated question, but it depends to a great extent on whether the intervention is susceptible of prior informed consent, which in turns hangs on whether there are likely to be unknown (especially potentially adverse) side-effects.

Here, I want to consider whether intended moral enhancements – those intended to induce pro-moral effects – can, somewhat paradoxically, undermine responsibility. I say ‘intended’ because, as we saw, moral interventions can have unintended (even counter-moral) consequences. This can happen for any number of reasons: the intervener can be wrong about what morality requires (imagine a Nazi intervener thinking that anti-Semitism is a pro-moral trait); the intervention can malfunction over time; the intervention can produce traits that are moral in one context but counter-moral in another (which seems likely, given that traits are highly context-sensitive, as I mentioned earlier); and so on – I won’t give a complete list. Even extant psychoactive drugs – which can count as a type of passive intervention – typically come with adverse side-effects; but the risk of unintended side-effects for futuristic interventions of a moral nature is substantially greater and more worrisome, because the technology is new, it operates on complicated cognitive structures, and it specifically operates on those structures constitutive of a person’s moral personality. Since intended moral interventions do not always produce their intended effects (pro-moral effects), I’ll discuss these interventions under two guises: interventions that go as planned and induce pro-moral traits (effective cases), and interventions that go awry (ineffective cases). I’ll also focus on the most controversial case of passive intervention: involuntary intervention, without informed consent.

One of my reasons for wanting to home on this type of case is that there is already a pretty substantial body of literature on passive non-consensual interventions, or ‘manipulation cases,’ in which a futuristic neuroscientist induces certain motives or motivational structures in a passive victim. We can tweak these examples to make the interventions unambiguously moral (the intervener is tampering with the victim’s moral personality), to derive conclusions about passive moral interventions and how they effect responsibility. My analysis isn’t going to be completely derivative on the manipulation cases, however, because theorists differ in their interpretations of these cases, and specifically on whether the post-manipulation agent is responsible for her induced traits and behaviours. I want to offer a new gloss on these cases (at least, compared to those I will consider here), and argue that the victim’s responsibility typically increases post-manipulation, as the agent gains authentic moral traits and capacities by the operation of her own motivational structures. (Except in the case of single-choice induction, where there are no long-term effects). I will also say some words on the responsibility status of the intervening neuroscientist.

I’m going to assess three kinds of case, each of which we can find in the literature. First, Frankfurt-type cases (in fact, Frankfurt’s original case), in which an intervener induces a single choice. Second, ‘global manipulation’ cases like Alfred Mele’s (2010), in which an intervener implants a whole new moral personality. And finally, partial manipulation cases, in which the intervener replaces part (let’s say half) of a person’s moral personality. I’m not aware of philosophical ‘partial manipulation’ cases per se, but there have been discussions of partial personality breakdowns, as in (e.g.) split-brain patients, which we can use as illustrative examples. Partial manipulation cases are like that, only instead of generic cognitive deficits that impair performance, there are moral deficits (due to internal conflict) that may undermine the agent’s ability to make and carry out moral plans.

Let’s consider each of these manipulation cases in turn.

1. Induced choice (minimal manipulation)

In the original Frankfurt-type case (1979), a futuristic neuroscientist named Black secretly implants a counterfactual device in Jones’ brain, which would compel Jones to vote for a certain political candidate if he were to choose otherwise, but the device is never activated. It’s causally inert, so it doesn’t affect Jones’ responsibility status, according to Frankfurt. (Jones is responsible because he acts on his own decisive desire). This example has to be modified to fit our purposes. First, let’s imagine that the implant is casually efficacious as opposed to merely counterfactual, as this is a more interesting kind of case. And second, let’s suppose that the intervention is designed to induce a moral effect – say, to make Jones donate to charity. Finally, let’s consider two types of case – one in which the intervention produces the intended effect, and one in which it produces a counter-moral effect.

I’ll assess these cases after discussing global and partial manipulation scenarios.

2. Global manipulation 

Alfred Mele (2006) offers a case of ‘global manipulation,’ in which a person’s entire motivational system is changed. He asks us to imagine two philosophy professors, Anne and Beth, and to suppose that Anne is more dedicated to the discipline whereas Beth is more laid back. The Dean wants Beth to be more productive so he hires a futuristic neuroscientist to implant Anne’s values in Beth, making her Anne’s psychological double. Is Beth responsible for her new psychological profile and downstream behaviours? According to Mele, no, because Beth’s new values are practically unsheddable: they cannot be disavowed or attenuated under anything but extraordinary circumstances (e.g., a second neuroscientific intervention). That is, Beth lacks control over her implanted values.

This is again not obviously a moral example, so let’s imagine that Anne is a saintly person and Beth is a jerk, and the Dean doesn’t like working with jerks, so he hires a futuristic neuroscientist to implant Anne’s saintly values in Beth. The intervention works and Beth becomes Anne’s saintly doppelgänger. This is a moral intervention and it’s a causally efficacious one. In this example, Beth’s moral personality is radically transformed.

If Mele is right, Beth is not responsible for her saintly values (and downstream behaviours) because these values are practically unsheddable. There are two natural ways of explaining this judgment. (1) Beth lacks control over her new values because they are firmly inalterable (Mele’s explanation). And (2) Beth is no longer herself – she has become someone else (Haji’s preferred explanation [2010]). These interpretations exemplify the control view and the character view – competing views in the literature that I described earlier. But in this case they converge on the same conclusion – Beth is not responsible – though they provide different (albeit overlapping) grounds. On one account, Beth lacks control over her desires, and on the other, her desires are not authentic, because they did not emerge through a process of continuous reflective deliberation (diachronic control). The two criteria are related, but come apart (as Mele emphasises [2006]): we can imagine implanted desires that are susceptible to revision but not the agent’s ‘own’; and we can imagine authentic desires that are relatively fixed and impervious to control (e.g., think of Martin Luther King saying, ‘Here I stand; I can do no other’). Yet on a historical picture of control, the two conditions substantively overlap: a person’s mental states are not authentic if they were not amenable to continuous reflective deliberation. At least, this is the case for passive intervention cases, which we are considering here. If we consider a third account of responsibility – Vargas’ agency cultivation model (2013) – we find still more convergence: it’s typically not agency-enhancing to hold someone responsible if she lacked diachronic control and wasn’t herself at a particular point in time. These accounts do not always converge, but in the case of passive intervention, there is substantive agreement. So, conveniently, we don’t need to arbitrate them. Yet there is still room for debate about whether a manipulation victim is responsible on any of these pictures, since there is room for debate about how to treat this person. Vargas holds that a post-intervention agent is responsible for her behaviour, but is a new person, contra Haji and Mele.

I’ll return to this question after considering a third case: partial manipulation.

3. Partial manipulation

Imagine that instead of implanting Anne’s values in Beth, making her Anne’s double, the manipulator had only implanted half of Anne’s values, or a substantial portion, leaving part of Beth’s personality intact. This is a partial manipulation case. Now Beth is internally fragmented. She has some saintly desires, but she also has a lot of non-saintly desires. Is she responsible for only her pre-implantation desires, or also for her post-implantation desires, or none of her desires (and downstream effects)? This is a more complicated scenario (not that the other two are simple!) We can perhaps compare this situation to neurocognitive disorders that involve internal fragmentation, which have also drawn attention from responsibility theorists – examples like psychosis, alien hand syndrome, and split-brain (callosotomy) patients. Perhaps by assessing whether these individuals are responsible (to any degree), we can determine whether partial-manipulation subjects are responsible. (Spit-brain surgery, which partially or wholly divides the two hemisphere down the corpus callosum, induces a similar split-personality effect).

Let’s look at these cases separately, beginning with global manipulation.


  1. Global

If Mele is right, then Beth is not responsible in the global manipulation case because she lacked control, and if Haji is right, she is not responsible because she lacked authenticity. These accounts have some degree of intuitive plausibility, surely. But Vargas (2013) offers a different interpretation. Vargas suggests that we should see post-global-manipulation agents as different people, but responsible people. So post-manipulation Beth is Beth2, and she is responsible for any actions that satisfy relevant conditions of responsibility for her. (Vargas’ preferred condition is a consequentialist one, but we can remain neutral). Since Beth2 can control her post-intervention desires (as much as any normal person), and they reflect her character, and it seems efficient to hold Beth2 responsible for these actions, we ought to regard her as responsible. Beth, on the other hand, is dead, and responsible for nothing. This view, it must be admitted, also has a degree of plausibility. But it seems to ignore the fact that Beth2, in very significant ways, is not like the rest of us.

I’m going to suggest a middle-of-the-road view, and it’s a view that emerges from a focus on how we become moral agents – a historical view. It is somewhat indebted to Haji’s [2010] analysis of psychopaths, who, somewhat like Beth2, have peculiar personal histories. According to Haji, psychopaths are not responsible agents because, unlike ordinary people, from childhood onward they lack the capacity to respond to reason due to myriad cognitive deficits (which we do not need to get into here). Children are not (full) moral agents: they lack certain rational capacities, yet they have certain emotional capacities that psychopaths lack, which allows to them to pass the moral/conventional distinction, at least from 5-years-old onward. Still, their rational deficits impair their moral agency. It is not only psychopaths who lack moral agency: Haji suggests that anyone who is barred from acquiring moral competency from childhood (due to congenital or involuntarily acquired deficits) is not a moral agent (a ‘normative agent,’ in his terms), because such people are, in effect, still children. (This includes severely neglected and abused children who develop psychopathic traits not of their own choosing). These individuals lack control over their motives, as well as anything that could be called a moral personality. If we want to say that children are not full moral agents, we must grant that these individuals, who suffer from arrested moral development, also are not full moral agents.

Now consider Beth2, who has been globally manipulated to have a different moral personality. Beth2 has zero (authentic) personal history. She is, in this respect, like Davidson’s ‘swamp-person’ (though different in other salient respects) – she emerges ex nihilo, fully formed. In lacking a personal history, Beth2 is like a newborn baby – a paradigm case of non-responsibility. Yet unlike the newborn baby, Beth has intact moral capacities, albeit not her own moral capacities – they were implanted. Beth2, then, is not responsible in either an authenticity sense or a diachronic control sense. Nonetheless, her extant moral capacities (though not her own) allow her to reflect on her motivational set, explore the world, interact with other people, and live a relatively normal life after the intervention. In this regard, Beth2 differs from psychopaths, who can never acquire such capacities, and are in a permanent baby-fied state, morally speaking. Moreover, as time goes by, Beth3 will become increasingly different from Amy, warranting the judgment that Beth3 is a separate person – her own person. So over time, it becomes more and more reasonable to say that Beth3 is an independent moral agent with her own moral personality. Although Beth3 cannot completely overhaul her motivational system, she can make piecemeal changes over time, and these changes are attributable to her own choices and experiences.

With this in mind, I submit that we treat Beth2 as not responsible immediately post-intervention (since she lacks any authentic motives at that time), but increasingly responsible thereafter (since she has the capacity to acquire authentic moral motives and capacities over time, unlike psychopaths). This doesn’t mean that Beth2 will ever be as responsible as an ordinary (non-manipulated) person, but she is certainly more responsible than newborn babies and psychopaths, and increasingly responsible over time.

Another problem with saying that Beth2 is fully responsible for her post-manipulation behaviour is that this leaves no room for saying that the clandestine manipulator – the Dean or the futuristic neuroscientist or whoever – is responsible for Beth2’s behaviour, especially in the case that the moral intervention goes wrong and induces counter-moral effects.

Suppose that Beth2 goes on a killing spree immediately after being manipulated: is she responsible for this effect? Surely the intervener is to blame. One could say, like Frankfurt (1979), that two people can be fully responsible for a certain action, but this seems like a problematic case of overdetermination. Surely blame must be distributed, or displaced onto the one who preempted the effects. Indeed, Frankfurt’s proposal doesn’t fit with how we ordinarily treat coercion. Consider entrapment: if a law enforcement agent induces someone to commit a crime, the officer is responsible, not the victim. This might be because the victim lacked adequate control (due to coercion), acted against character, or because it’s not useful to blame victims – entrapment fits with all of these explanations. Shoemaker seems to appeal to the last consideration when he says of entrapment, we blame the public official and not the victim because “we think the government shouldn’t be in the business of encouraging crime” (1987: 311) – that is, we don’t want to encourage government corruption. But Shoemaker also appeals to fairness, which is tied to control and character: it’s not fair to blame victims who lacked sufficient control or weren’t themselves when they acted (which is also why it’s not useful to blame them). So on any standard criterion of responsibility, it’s not clear why we would blame a manipulation victim.

Now, suppose that the intervention worked and the Dean makes Beth2 a moral saint. If I am right, Beth2 isn’t praiseworthy for the immediate effects of the intervention because they’re not her own (on various construals). The intervener’s moral status is more complicated. While he might seem prima facie praiseworthy for Beth’s pro-moral traits, we also have to consider the fact that he’s patently blameworthy for intervening without informed consent or any semblance of due process (viz. decisional incapacity protocols), and this might vitiate or cancel out his prima facie praiseworthiness. If we consider praise and blame as competing attitudes, it makes sense to see the Dean as blameworthy on balance.

2. Partial

Next, let’s consider a partial manipulation case. Let’s s imagine that the Dean replaces half of Beth’s personality with Anne’s, creating Beth3. This is trickier than the global manipulation case inasmuch as Beth3 has some authentic mental states, but they are in competition with implanted states that are not her own. So we can’t say that Beth3 lacks authenticity or diachronic control entirely, yet she is deficient in both respects compared to an ordinary person. We might compare Beth3 to examples of neurocognitive disorders that cause internal fragmentation, such as alien hand syndrome and split-brain effects, which have attracted a lot of philosophical interest. These subjects have ‘alien’ motives like Beth3, but unlike split-brain patients, Beth2 can presumable enhance her control and motivational congruence over time – assuming that none of her implanted states are pathological. So Beth is somewhere between split-brain patients and non-pathological individuals.

There are three ways that we can go in assessing Beth3’s responsibility. (1) We can hold that Beth3 is not responsible (period) because she has insufficient control and insufficient depth of character to ground responsibility. (2) We can say that Beth3 is actually two agents, not one, and each agent is responsible for its own motives and downstream behaviours. King and Carruthers (2012) seem to suggest that, if responsibility exists (on which they seem unsure), commisurotomy patients and alien hand patients must be responsible for their aberrant behaviours, since the former have two “unified and integrated” personalities (205), and alien-hand gestures reflect a person’s attitudes, albeit unconscious ones. Furthermore, they think that consciousness cannot be a requirement of responsibility since most motives aren’t conscious anyways. If this is right, then perhaps Beth2 is two people, each responsible for ‘her’ own motives. This, to me, seems impossibly impracticable, because we can only address responsibility attributions to one self-contained agent. We cannot try to reason with one side of Beth3’s personality at a time, because Beth3 doesn’t experience herself as two people. She won’t respond to targeted praise and blame, aimed at adjacent but interconnected motivational hierarchies. And she might even find this kind of moral address baffling and/or offensive. So this won’t do, practically speaking. But there’s also a good case to be made that commisurotomy patients and the like lack control and character in the normal sense, given that they have significant cognitive deficits. So it’s reasonable to see them as impaired in their responsibility compared to cognitively normal people. And likewise for Beth3.

This leads to the third possibility, which I favour: the middle-of-the-road proposal. Beth3 is somewhat responsible immediately after implantation, and increasingly responsible thereafter. This is because Beth3 is subject to control-undermining motivational conflict and disorientation following the clandestine intervention, but she nonetheless has some intact moral (relevant rational, emotional, and motivational) capacities, which differentiate her from the psychopath, and which should allow her to regain psychological congruence over time, enhancing her control and authenticity. So Beth3 should be seen as increasingly responsible over time (under ordinary circumstances). That said, she will likely never be as responsible as someone who had an ordinary learning history, since most people never suffer from this degree of fragmentation. So she may have relatively diminished responsibility for a long time, even if she becomes more apt for praise and blame.

Once again, this analysis can be applied to effective interventions and ineffective ones. If Beth3 acquires pro-moral traits (as per the Dean’s intention), she is not immediately responsible for them, but she gains responsibility for any induced traits that persist and amplify over time – indeed, for all of her post-intervention traits. Regarding the intervener, he is not necessarily praiseworthy for the pro-moral effects of the intervention, inasmuch as he might be blameworthy for intervening without consent or due process, and this might outweigh any praise that he might otherwise warrant.

Worse still, if the Dean inadvertently implants counter-moral traits in Beth3, he is blameworthy for intervening without consent as well as the effects of the botched intervention.

3. Minimal

Finally, let’s consider the induced-choice subject. Call her Beth4. Suppose that Beth4 has been inculcated with a desire to give to charity, and she acts on this desire. (Note: presumably to be causally efficacious a desire must be combined with a supporting motivational structure, but for simplicity I’ll refer to this bundle, following Frankfurt’s example, simply as a desire. Assume than an induced desire comes with supporting motivational states). Is Beth4 praiseworthy? On my view, no, because her desire is not her own. But the intervener also is not praiseworthy, insofar as he is blameworthy for intervening without consent or due process, which (I think) cancels out the import of of any good intentions he may have had. (The road to hell is paved with good intentions, as they say). (Note that I am construing praise and blame as competing reactive attitudes; other people hold alternative conceptions of ‘responsibility,’ which I will conveniently ignore).

Next, suppose that the Dean intends to induce a pro-moral choice in Beth4, but inadvertently induces her to commit a murder. The Dean, I think, is blameworthy for the murder, because he is responsible for implanting the relevant desire in Beth4 without her consent. This can be construed as an omission case, in which the Dean was ignorant of the likely consequences of his choice, but is responsible because he failed to exercise proper discretion. (Compare this to Sher’s [2010] example of a soldier who falls asleep on duty; the Dean failed in his duties as a person, a citizen, a Dean… he failed on many levels. He is, as it were, a spectacular failure as a moral agent). The Dean acted in character, and on a suitably reasons-responsive mechanism, when he chose to surgically intervene on Beth4 without her consent, and he bears responsibility for this infraction and the fallout of his choice.



Moral enhancements & moral responsibility


I’m going to write a somewhat lengthy but off-the-cuff entry on moral enhancements, because I have to present something on them soon, in Portugal.

I’m going to write about (1) moral enhancement and moral responsibility, and how these things intersect (we might be responsible for failing to use moral interventions); (2) the possibility of duties to enhance oneself and others, and (3) passive moral interventions – the most controversial type – and whether we have a duty to use, submit to, or administer them.

A qualification: I don’t know very much about moral enhancement, so this is going to be a bit sketchy.

Moral enhancements and moral responsibility

Here’s what I know. A ‘moral enhancement’ is a “deliberate intervention [that] aims to improve an existing capacity that almost all human beings typically have, or to create a new capacity” (Buchanan 2011: 23). These interventions can be traditional (e.g., education and therapy), or neuroscientific (e.g., biomedical and biotechnological interventions). The latter are more controversial for various reasons. Examples include transcranial magnetic stimulation (TMS), transcranial direct current stimulation (tDCS), deep brain stimulation (DBS), and psychoactive drugs, along with futuristic ‘Clockwork Orange’-type methods that we can only imagine. (Typical science fiction examples are dystopic, but we shouldn’t let this prejudice the discussion). These enhancements are moral insofar as they aim to enhance the subject’s moral-cognitive capacities, howsoever construed. (I’ll leave this vague).

Some philosophers hold that the distinction between traditional and modern interventions is that one is ‘direct’ while the other is ‘indirect’: TMS, for instance, directly stimulates the brain with electric currents, whereas education ‘indirectly’ influences cognitive processing. Focquaert and Schermer argue that the distinction is better characterised as one between ‘active’ and ‘passive’ interventions: traditional interventions typically require ‘active participation’ on behalf of the subject, whereas neuroscientific interventions work ‘by themselves’ (2015: 140-141). More precisely, passive interventions work quickly and relatively unconsciously, bypassing reflective deliberation. The authors acknowledge that this is a rough and imperfect distinction, and that interventions lie on a continuum; but they think that this distinction nicely captures a normative difference between the two methods. Passive interventions, they say, pose a disproportionate threat to autonomy and identity because (1) they limit continuous rational reflection and autonomous choice, (2) they may cause abrupt narrative identity changes, and (3) they may cause concealed narrative identity changes. These effects undermine the agent’s sense of self, causing self-alienation and subjective distress (2011: 145-147).

For this reason, the authors recommend that we introduce safeguards to minimise the negative effects of passive intervention – safeguards like informed consent and pre- and post-intervneiton counselling, which help subjects integrate their newfound moral capacities into their moral personality, or come to terms with their new moral personality, in case of radical transformation. So although passive interventions have a higher threshold of justification, they can be justified if their adverse effects are managed and minimised.

(Some objections to this view, with responses, can be found here).

Now for a few words on moral responsibility. In earlier posts, I’ve discussed three models of responsibility, two backward-looking (character and control theory) and one forward-looking (the agency-cultivation model). These theories have something in common: they hold agents responsible (blameworthy) for omissions. On character theory, we are responsible for failing to act virtuously, if a reasonable person would have done better in our circumstances (see Sher 2010). So someone can be blameworthy for forgetting about her friend’s birthday. On control theory, we are ‘indirectly’ responsible for failures to exercise reasons-responsiveness responsibly, in accordance with people’s ‘reasonable’ expectations. As Levy remarks when discussing ‘indirect’ responsibility for implicit bias, a person can be “be fully morally responsible for [a] behaviour [resulting from implicit bias], because it was reasonable to expect her to try to change her implicit attitudes prior to t” (2014; see also Holroyd 2012). Since we can be responsible for omissions on both models, it stands to reason that we can be responsible for failures to enhance our moral agency, if this is what a reasonable person would have done or expected of us in the circumstances. In fact, remediating measures for implicit bias can be seen as ‘moral enhancements,’ though they are closer to education than biomedical interventions. But the point is that there are grounds for blaming someone for eschewing moral enhancements of either type – active or passive – depending on the particularities of the case.

On the agency-cultivation model (ACM), it’s even more obvious that a person might be blameworthy for failing to use moral interventions, since blame functions to enhance moral agency. If blaming someone for not using a moral intervention would entice the person to use it, thus enhancing the person’s agency, blame is a fitting response. And if forcibly administering the intervention would enhance the person’s agency, this might be fitting, too.

All of these theories as I have presented them invite questions about when blame for failing to use moral enhancements, or for using them inappropriately, is apt. But this leads somewhat away from responsibility (as an agential capacity or cluster of traits that warrant blame and praise), to the domain of duties: from the question of what we are responsible for, to what we are responsible to do. These two things are related insofar as we can be responsible (blameworthy) for omitting to fulfil our responsibilities (duties) on all three models. But they are distinct enough that they necessitate separate treatments. First we figure out what our duty is, and then we determine if we are blameworthy for falling short of it.

Moral duties and moral perfection

What are our duties with respect to moral enhancements? We can approach this question from two directions: our individual duty to use or submit to moral interventions, and our duty to provide or administer them to people with moral deficits. This might seem to suggest a distinction between self-regarding duties and other-regarding duties, but this is a false dichotomy because the duty to enhance oneself is partly a duty to others – a duty to equip oneself to respect other people’s rights and interests. So both duties have an other-regarding dimension. The distinction I’m talking about is between duties to enhance oneself, and duties to enhance other other people: self-directed duties and other-directed duties.

These two duties also cannot be neatly demarcated because we might need to weigh self-directed duties against other-directed duties to achieve a proper balance. That is, given finite time and resources, my duty to enhance myself in some way might be outweighed by my duty to foster the capabilities of another person. So we need to work out a proper balance, and different normative frameworks will provide different answers. All frameworks, however, seem to support these two kinds of duties, though they balance them differently. For Kant, we have an absolute (perfect) duty to abstain from using other people as mere means, so we have a stringent duty to mitigate deficits in our own moral psychology that cause this kind of treatment; and we also have a weaker (imperfect) duty to foster other people’s capabilities. On a consequentialist picture, we have to enhance the capabilities of all in such a way as to maximise some good. Our duty to enhance ourself is no greater or lesser than our duty to foster others’ capabilities. Aristotle, too, believed that we have a duty to enhance ourselves and foster others’ capabilities by making good choices and forming virtuous relationships. Rather than arbitrate amongst these frameworks, we can just note that they all support the idea that we have duties to enhance ourselves and others. The question is how stringent these duties are, not whether they exist. Of course, in a modern context these questions take on new significance, because the duty to enhance oneself and others no longer just means reflecting well and providing social assistance; it could potentially mean taking and distributing psychoactive drugs, TMS, etc.

One pertinent question here is how much enhancement any individual morally ought to undergo. How much enhancement is enough enhancement? We can identify at least four accounts of the threshold of functioning we ought to strive for: (1) the ubermensche view: we must enhance our own and others’ capabilities to the greatest extent possible; (2) prioritarianism: we should prioritise the capabilities of the least well-off or the most morally impaired; (3) sufficientarianism: we should enhance people’s capabilities until they reach a sufficient minimal threshold of functioning, after which it is a matter of personal discretion; and (4) pure egalitarianism: we should foster everyone’s capabilities equally.

No matter which of these accounts we favour, they all face potential objections. The more stringent forms (the ubermensche view) are more susceptible to these objections than the less stringent forms (sufficientarianism). If we are considering enhancing our own capabilities, the scope for criticism is minimised, but we might still worry that our attempts to enhance ourselves morally might go awry, they might interfere with our ability to foster others’ capabilities, they might impel us to neglect valuable non-moral traits, or (relatedly) they might prevent us from becoming well-rounded people, capable of living flourishing lives (see Susan Wolf’s ‘Moral Saints’ 2004). When considering whether we should foster others’ capabilities, there are still more worries. Namely, in seeking to enhance other people – particularly if we do this through institutional mechanisms – we risk insulting and demeaning those deemed ‘morally defective,’ stigmatising people, jeopardising pluralism (which may be inherently valuable), undermining people’s autonomy, or fostering the wrong kinds of traits due to human error. That is, by intervening to enhance moral agency, we might inadvertently harm people.

The most controversial type of moral intervention – passive – is the most vulnerable to these objections, and others that I haven’t mentioned. From now on, I’ll just talk about this type. I’ll consider Focquaert and Schermer’s cautionary advice, and then offer some suggestions.

Passive intervention, autonomy, and identity

The authors argue that passive moral interventions are particularly morally problematic because they disproportionally jeopardise autonomy and identity, insofar as they limit continuous rational reflection and precipitate abrupt and concealed narrative identity changes. To minimise these adverse effects, interveners should provide informed consent procedures and pre- and post-intervention counselling. Let’s suppose that this is right, and also that we have a duty to enhance ourselves and others, as per the above discussion. This seems to have the following implications. First, a duty to submit to a passive moral intervention obtains only on condition that informed consent and pre-and post- intervention counselling are available and reasonably effective. So, a person can be blameworthy for failing to undergo a passive moral intervention (if at all) only if the intervention came with those provisions. Secondly, there is (at minimum) a duty to provide these safeguards if administering a passive intervention; so, if an intervener fails to do so, the person is thereby blameworthy. This is the upshot of holding that there is a duty to enhance oneself and others, and this duty includes providing harm-minimising safeguards.

That said, there are other considerations that might vitiate the perceived duty to enhance one’s own and others’ moral capabilities. I think that the most forceful objection, which may underlie other objections to these interventions, is that we can’t be sure of their effects. We can call this the knowledge constraint. If we can’t know, in medium-grained terms, the probable effects of a passive intervention, we can’t obtain informed consent in a strong sense, since we can’t adequately assess whether the effects will be good or bad. Let’s reconsider Focquaert and Schermer’s worries, and tie it to this concern. They say that (1) passive interventions threaten continuous rational reflection and autonomy. When someone undergoes TMS, for instance, the person doesn’t have the capacity to endorse or reject the effects of the magnetic stimulus. But why is this a problem? I don’t think it suffices to say that it undermines autonomy, because many autonomy-enhancing procedures undermine the capacity for continuous reflection: I forfeit this capacity when I undergo a root canal under general anaesthesia, but I can still give prior informed consent unproblematically. This isn’t just because a root canal is localised to the mouth (although this is a related issue). It’s mainly because I know precisely what the effects of the root canal will be, so I can consent to them. If dental procedures induced unforeseeable side-effects, informed consent for them would be much more problematic. This suggests that the unforeseeableness of the side-effects of a procedure, as opposed to the reflection-undermining nature of the procedure per se, is what undermines autonomy and has moral import.

With neuroscientific interventions, we can’t very accurately predict the general immediate and long-term effects , because these interventions function very coarsely, across a broad range of cognitive systems and processes, and their mechanisms are not fully understood. Similarly, the brain is very complex, and its mechanisms are not very well understood. And so the interactions between these mechanisms are pretty murky. For these reasons, any targeted (intended) effect of neurointerventions comes with a range of possible side-effects. If we just focus on the immediate side-effects of TMS, they include fainting, seizures, pain, as well as transient hypomania, hearing loss, impairment of working memory, and cognitive changes (unspecified). In terms of long-term effects, NICE found that there is insufficient evidence to evaluate safety for long-term and frequent uses. That is, not enough is known to predict the long-range effects in a scientifically responsible way. Speaking more generally, TMS is used for multiple conditions; its activation is very un-specified, targeting various brain regions that serve various functions. It can’t be used to, say, ‘cure schizophrenia’; it acts on multiple brain regions, bringing about a cluster of effects, some intended and others unintended. All neurointerventions work this way: they induce a cluster of effects. Subjects can’t select just one desired effect: undesired side-effects are very likely, if not inevitable. Given these considerations, it’s far from clear that there is an obligation for anyone to submit to passive interventions that act directly on the brain, and even less clear that anyone has an obligation to induce them on other people.

So, it seems that the threat to continuous rational reflection is a problem because we don’t know, in medium-grained terms, what the effects of the procedure will be in advance, which threatens prior informed consent. If we could predict the effects relatively accurately, the loss of continuous rational reflection would not be a problem.

(2) Passive interventions cause abrupt narrative identity changes. Again, I think that this is a problem because the end result of the passive intervention is impossible to predict in medium-grained terms. If we could provide someone with an accurate specification of the effects of a proposed intervention and gain the person’s consent, the ‘abruptness’ of the narrative identity change wouldn’t be an obstacle, it would be an advantage. If I could get a root canal ‘abruptly’ rather than slowly, I would choose the abrupt method – but only if I understood the consequences. If I could quit smoking ‘abruptly’ as a result of voluntary hypnosis, I would choose that option. Abrupt changes are problematic only if we can’t predict the outcomes of those changes.

(2) Passive interventions cause concealed narrative identity changes. This is also a problem only because we can’t predict the effects of the intervention. If we could, we might even prefer that the procedure be concealed. I might want to have a ‘concealed’ root canal – a root canal that I completely forget about after the fact, but with my prior consent – but not if the procedure induces unwanted side-effects that I can’t detect until after the fact. If side-effects are an issue, I want to be conscious so that I can do what I can to mitigate them.

The fact that passive moral interventions impinge on the brain rather than the body (not that this is a hard and fast distinction) is also problematic because our brain is the seat of our personal identity, and changes to our personal identity are especially jarring. Still, the analogy between cognitive interventions and bodily interventions is informative. If I sign up for an appendectomy and wake up with a third arm, this is a violation of my autonomy. Waking up a different, or radically altered, person, is that much more autonomy-undermining. In both cases, the lack of prior knowledge about probable outcomes is what violates the person’s autonomy: an unforeseen transformation is an affront to the self. (This is true even if the transformation is an accident: if a piano player’s hand is crushed by a boulder, this is autonomy-undermining, even if no one is to blame [see Oshana 2008]).

Post-intervention counselling might mitigate some of these concerns, because it allows the agent to ‘consent’ retroactively. But this is problematic for the following reasons.

(1) For no other procedure is retroactive consent permissible. If my teeth are rotting and someone kidnaps me and performs involuntary dental work while I’m unconscious, this is a moral infraction. (Note that when I was discussing clandestine changes above, I meant that they were permissible only if there was prior informed consent). Even if I’m pleased with the dental work, the dental intervener could have performed the procedure less invasively – with my consent – and to this extent, he’s blameworthy: he chose the most invasive intervention instead of the least invasive one. Now, when moral enhancements are at issue, the intervention may be prima facie justified by the potential positive consequences, but we have to consider (1) the strength of the benefits, (2) the probability of adverse side-effects, and (3) whether a less invasive intervention was available. If passive interventions are used as a first response, this is clearly impermissible. If they are used as a ‘last resort,’ when no alternative is available or all alternatives have been tried, we have to consider the moral import of possible side-effects. Consider Alex (shown above) from ‘A Clockwork Orange’: the movie is a horror story because the Ludovico technique made him miserable and dysfunctional, unintentionally. And he was not informed beforehand of the risks. Interveners are, I believe, blameworthy for using interventions that may cause severe adverse side-effects, because there is a duty to protect people from these kinds of risks. This not to say that passive interventions are never permissible, but they are not permissible if they carry severe or unknown risks to wellbeing.

(2) Because moral interventions target a person’s moral identity, there is an argument to be made that post-intervention ‘consent’ is not consent at all, because the agent on which the intervention acted – the pre-intervention agent – no longer exists, or is so drastically altered, consent is not possible for that agent. Moral interventions have the ability to ‘kill off’ the pre-intervention agent. This places a massive burden of justification on those who would intervene. We need to consider whether it is permissible to ‘kill off’ unwanted moral personalities, or radically alter them in a ways that preclude obtaining consent. When a person is radically altered by an intervention without consent, the person cannot consent after the treatment, so any post-intervention consent given is a different person’s. This is a radical departure from the standard model of informed consent, and this is something we need to consider. We also need to weight whether society should be choosing which moral personalities have value, as this might threaten pluralism and self-determination.

I should make a qualifying statement here. I don’t want to stigmatize the use of TMS and other existing passive interventions. Some people voluntarily use TMS for personal (prudential) reasons, which I would deem morally unproblematic. I’m specifically concerned with whether people have a duty to use passive interventions – not just existing ones, but also interventions that may arise in the future – to enhance their moral capabilities; and whether society has a duty to impose these interventions on people without their consent. I’m saying that it’s much harder to justify a categorical duty (as opposed to a subjective desire) to undergo passive interventions, and still harder to justify a duty to impose (as opposed to merely offer) interventions on people deemed morally deficient.

Back to responsibility

Here’s how this bears on responsibility. On a simplistic reading, it might seem that we have a duty to use moral interventions to enhance ourselves and others to a substantial (sufficientarian, if not ubermenschean) degree. And maybe this is true for traditional moral enhancements, which we can reflectively choose and consent to with full information. But it’s not clear that this is the case for passive moral interventions, just because we can’t tell if those interventions are going to have overall positive consequences. And this is because neurointerventions target clusters of traits, some of which may be adaptive while others may be maladaptive. So, in trying to morally enhance someone, we might inadvertently harm the person and even introduce new counter-moral traits. If so, we might be blameworthy for the consequences of those interventions. And if there is no clear duty to submit to passive moral interventions that might produce harmful and even counter-moral side effects, then a person might not be blameworthy for refusing them.

Consider stories like ‘A Clockwork Orange’ and ‘Brave New World.’ They’re about attempts at moral enhancement and social engineering gone horribly awry. And this is a legitimate worry. I’ve suggested that when technology is morally problematic, it’s not because it threatens rational reflection and narrative identity per se. Rather, it’s because it has unpredictable side-effects, and these side-effects might be morally problematic. This isn’t mean to imply that we have no duty to enhance ourselves and others, but rather that when trying to enhance our capabilities with modern methods, we should proceed with caution.


Implicit Bias: The limits of control/character: Continued.


This post continues from the last one. I was saying that if we need to trace responsibility back to a suitable prior moment (t-1) at which the agent could have foreseen the consequences of his choices (following Vargas’s 2005 description of control theory), then we need to assess not only that agent’s internal capacities at t-n, but also the agent’s ‘moral ecology,’ and the relationships between the agent and the moral ecology. This follows from the fact that ‘indirect control’ is a matter of whether an agent could have exercised or acquired a capacity using available resources in her local environment. (Indeed, these two things are related: we acquire capacities in part by exercising more basic capacities: by taking piano lessons, I learn how to play piano well, i.e., I acquire piano-playing reasons-responsiveness. This is a result of exercising a more basic capacity – the basic human capacity to master a symbolic system with a combination of tutelage and practice). According to Holroyd, if someone is implicitly biased but could have avoided becoming implicitly biased or expressing an implicit bias, the person may be blameworthy. I am suggesting that to determine a person’s blameworthiness, we must evaluate, not just the individual person at t-n, but the person’s location in the social ecology.

This picture suggest a different kind of objection to Fischer’s notion of control – the dominant model. Fischer explicitly states that to appraise a person’s responsibility status, we must home in on the ‘actual sequence’ of deliberation, ignoring counterfactual circumstances in which the agent would have deliberated differently. That is, control for Fischer is ‘actual-sequence control.’ This strategy is (I believe) meant to undercut incompatibilist and nihilistic objections to control: objections to the effect that no one is responsible for anything because no one is capable of exercising ultimate control (being an unmoved mover) in a determinist universe (G. Strawson 1986), or because control is irrelevant in light of moral luck: whether you’re capable of exercising compatibilist control is just a matter of luck, not agency (Levy 2008). I think this last view places too much importance on the moral ecology and not enough on agency, but we can return to this later. By restricting control to the actual sequence, we cut off counterfactual circumstances in which the agent is metaphysically determined, as well as circumstances in which all agents are equally capable of control – distant possible worlds. But there is a case to be made that even if we include some counterfactual circumstances as morally relevant – and thus, some possible worlds – we do not need to include all counterfactual possibilities. Even if the buck doesn’t stop at actual-sequence control, it might stop in the next nearest possible world, cutting off the kind of slippery-slope objections that lead to nihilism.

I’m going to make this case in a moment, but first consider an existing objection to actual-sequence control, from Levy (2008). Levy argues that, while counterfactual disabling circumstances superficially appear to be irrelevant (consider, for example, Frankfurt-type cases in which the counterfactual device is never activated), it is possible to construct a counterfactual enabling circumstance that seems to make a difference. An enabling circumstance is one in which the agent gains a capacity, in contrast to the standard disabling scenarios, in which the counterfactual device would prompt the agent to go against the demands of morality if it were activated. Here’s one of Levy’s examples:


Jillian is walking along the beach when she notices a child growing. Jillian is a good swimmer, but she is pathologically afraid of deep water. She is so constituted that her phobia would prevent her from rescuing the child were she to attempt to; she would be overcome by feelings of panic. Nevertheless, she is capable of trying to rescue the child, and she knows that she is capable of trying. Indeed, though she knows that she has the phobia, she does not know just how powerful it is; she thinks (wrongly) that she could effect a rescue. Unbeknownst to Jillian, a good-hearted neurosurgeon has implanted her with a chip with which he monitors Jillian’s neural states, and through which he can intervene if he desires to. Should Jillian decide (on her own) to rescue the child, the neurosurgeon will intervene to dampen her fear; she will not panic and will succeed, despite her anxiety, in swimming out to the child and rescuing her. (2008: 170, 2008b: 234).

Levy actually presents this scenario variably as an objection to control theory and an objection to character theory. Suppose that Jillian decides not to rescue the child, in spite of (falsely) believing that she can. In the first place, it seems as if Jillian is responsible (says Levy) because she believes that she can rescue the child, and fails to act on this belief. But if Fischer is right, then Jillian isn’t responsible because she can’t succeed by her own (independent) means. Without the help of the benevolent intervener, success is impossible, and the intervener is external to Jillian’s motivational set. To vindicate the intuition that Jillian is responsible, we need to include the counterfactual intervener; so counterfactual scenarios seem relevant. (At least, this is my understanding). In the second place, if we include the counterfactual scenario as relevant, we have to accept that Jillian’s motivational set, and thus her character, include this scenario as a component part, and so Jillian’s character is “smeared” across time and space (Levy 2008: 179). Hence, locational externalism (i.e., the extended mind hypothesis) is true. These proposals call into question the viability of actual-sequence control and of character as traditionally conceived. I take it that this is supposed to support responsibility nihilism, i.e., the idea that responsibility (in a desert-entailing sense) doesn’t exist, as per Levy’s thesis in ‘Hard Luck.’

I think that that some of Levy’s claims are accurate and others need to be modified. Here’s what I endorse and what I dispute. I agree that counterfactual circumstances matter – at least, some counterfactual circumstances (specifically, those in which the agent could have intervened and succeeded with some kind of help), but I disagree with the ‘intuition’ that Jillian is responsible for an omission in this particular case. The reason is that, if someone, for bizarre reasons, thinks that she can save someone from drowning, but there is good objective reason to think that she can’t, the person doesn’t have a duty to intervene. Arguably, if we are not lifeguards, we don’t have a standing duty to save someone from drowning in any circumstance, because the risk of drowning ourselves in the attempt is too high, even if we are excellent swimmers for ordinary people (not lifeguards). So there’s no reason to think that Jillian ought to act on her belief that she can help. Normal people don’t have a standing obligation to risk their lives to save drowning victims. What we ought to do is alert the lifeguard, call emergency medical services, or look for an indirect means of intervening that doesn’t risk our own safety. So if we don’t have an obligation to risk our lives to save drowning victims, then people with water-related anxiety disorders certainly don’t, even if they have bizarre beliefs about their capacities and their moral duties.

I raise this point in part because this is the response I usually get when I present this case to other people. We could adjust the scenario by stipulating that the child is drowning in a wading pool. Normal adults have a duty to save children from drowning in wading pools, surely. But Jillian is not a psychologically-normal adult, so the worry recurs. Just as it is unreasonable for a normal adult to think that she has a duty to save a drowning victim in open water, it may be unreasonable for someone with a water-induced phobia, which could endanger her safety, to think that she has a duty to save a wading-pool victim. The worry is that the sacrifice required of the person is (objectively) too great to ground the existence of a duty, even if the person thinks that she has a duty for bizarre, subjective reasons.

But now consider a case in which it’s more obvious that a moral duty obtains. I’ll try to construct it to resemble the Jillian scenario, i.e., to include a protagonist who cannot achieve some end using her own (internal) capacities, but could succeed with external support. Suppose that Jack can’t contain his anger towards women, and regularly berates his female employees, family members, servers at restaurants, and so on. Jack (falsely) believes that he can control his misogynistic anger using willpower alone, and decides to exercise his willpower. But this is a false belief – he has much less willpower than he imagines. Unbeknownst to Jack, a local therapist specialises in anger-management problems, and would have been able to help him if he had sought out her help. But Jack never does. So Jack tries to exercise his willpower and fails, and continues to demean women on a daily basis.

If we home in on Jack’s internal capacity for control, we have to excuse Jack, since he lacked the internal resources to suppress his misogynistic urges. But if we consider the counterfactual scenario (in which Jack visits the therapist) as relevant to Jack’s responsibility status, then we can hold him responsible. It seems very reasonable to say that Jack had the capacity, in a very basic sense, to look for resources to control his misogynistic anger. But Jack does not have unassisted actual-sequence control over his anger.

I think that this scenario presents a plausible argument for the idea that, although Jack lacks actual-sequence control over his misogynistic anger, he has counterfactual-sequence control over it, and this counterfactual control is relevant to Jack’s responsibility status. Jack would not be risking his life by seeking out therapy. In fact, he wouldn’t be sacrificing anything of moral value. And if psychiatric counselling is free, as it is in Canada and some of the moral socialist countries, he isn’t even sacrificing anything of prudential value. By not seeking help, he’s not exercising his (basic human) capacities in the way he should. And because he’s not exercising these capacities responsibly, he has a character flaw.

Now, assuming that all of this is plausible, there’s an argument to be made that the proposed counterfactual-control model initiates a slippery slope into responsibility nihilism. Once we allow that counterfactuals are relevant, we have to admit that determinism precludes agency. But that’s only true if we hold all counterfactual possibilities – or at least, very many counterfactual possibilities – to be morally relevant. Yes, in a deterministic universe no one is incompatibilist-responsible for anything. But why go back to the metaphysical basis of reality – the metaphysical underpinnings of human behaviour? Why engage in ‘panicky metaphysics’ at all, when we can stop at the moral ecology? All I’m suggesting in constructing the above example is that some counterfactual possibilities matter – the ones that the agent could have availed himself of relatively easily, without sacrificing anything of moral (and in this case, even prudential) value.

Recall that in my last post I noted that control theorists typically espouse an implicit ‘reasonableness’ constraint in their conception of ‘indirect responsibility’: they hold an agent responsible for omissions for which it is reasonable to complain against the agent. As Levy says when commenting on implicit bias, an agent might “be fully morally responsible for [a] behaviour [resulting from implicit bias], because it was reasonable to expect her to try to change her implicit attitudes prior to t” (2014). This implies that only counterfactual circumstances that were reasonably available to the agent are morally relevant – scenarios in which the sacrifice demanded of the agent is not overly stringent. I have argued that in the Jillian scenario, the moral demand is too high, but in the Jack case, it’s not. This gives us a foundation on which to say that someone can be responsible for failing to exercise counterfactual control if doing so was reasonable. Of course, ‘reasonableness’ is a vague concept, and I won’t precisify it here, but it’s a concept that makes intuitive sense, and one that we regularly rely on without clarification, to constrain the scope of normative concepts. (Consider Scanlon’s account of moral principles as those that no suitably-motivated person could reasonably reject; we get the point without delving into the semantics).

Character theorists of Sher’s stripe similarly hold that reasonableness is critical to responsibility. A person is responsible for an omission just in case a reasonable person with relevantly similar capacities would have done better. So once again, we are to judge the agent’s responsibility status by what it would be reasonable to expect of her, in light of certain counterfactual possibilities – what the agent could have achieved under conditions C.

If this is right, then counterfactual circumstances do matter morally. But not all, or just any, counterfactual circumstances are relevant. Just those that were reasonably available to the person, at relatively low personal cost. This goes back to what I was saying about the moral ecology, and about tracing. Counterfactual scenarios are part of the moral ecology, external to the person’s material brain. So when evaluating a person’s responsibility status, we have to consider the person’s relevant brain states, and the reasonably available counterfactual circumstances supported by the agent’s moral ecology. We have to look at the person’s capacities and the person’s moral ecology and the potential interaction between those two variables, to see if they support the possibility of agency cultivation. And with regard to tracing, we need to trace responsibility to that possibility. We need to assess whether Jack, for example, had the (general) capacity to acquire the (specific) capacity to remediate or suppress his misogynistic anger, given the resources of his moral ecology.

This suggests the following revisions to control theory and character theory. We need to see control as more than actual-sequence control to account for the possibility of indirect responsibility for omissions that were reasonably avoidable. Specifically, we need to include as morally relevant, counterfactual circumstances in which the agent could have interacted with the moral ecology in such a way as to bring about a new capacity, or (more precisely) to leverage an existing basic (undifferentiated) capacity into a more specific (specialised) capacity. And it suggests that we ought to regard character as diffuse or locally extended, i.e., co-constituted with agency-supporting or agency-enhancing social supports. (This is not a revision to Sher’s view, in fact, but it emphasises that aspect of it). And finally, it suggests that when ‘tracing’ back to control, we need to trace beyond the agent’s actual sequence, to reasonably available aspects of the agent’s moral ecology, which would have enhanced the agent’s capacities if the agent had taken the right kind of initiative. And similarly with character theory, we need to trace character to relevant features of the local ecology, to determine if the agent is using those features to the best of her ability. If she is not, she may be culpably indifferent. (In this way, tracing applies to character theory as well, though its reach is more limited – we don’t need to trace as far back).

These considerations strike against any theory that is too narrow in its conception of responsible agency, particular the actual-sequence control model. And it suggests that control theory and character theory are perhaps more similar than they may initially appear, in that control theory admits a greater scope for blame under the category of ‘indirect responsibility,’ properly understood. These considerations build on Levy’s objections to actual-sequence control and character internalism. But I recruit them to show that there is a broader scope for blame than we tend to think, and he does the opposite – he recruits them in support of responsibility nihilism. Our views are technically compatible, though, because he’s refuting a desert-based notion of responsibility, and a relatively harsh form of desert-based responsibility, on which blame (1) is justified by reference to an agent’s actual sequence of deliberation or internal traits, and (2) entails fairly punitive sanctions. I also reject this notion of responsibility, because it combines a metaphysically tenuous conception of agency with dubious assumptions about the kind of thing blame is (punishment) and proportionality (harshly punitive). I think that the same objections can be taken to support a modification of responsibility rather than a rejection of it.

Here’s one substantive alternative to the actual-sequence control model – the ‘limited counterfactual-sequence control model.’ People are responsible for (1) intentional infractions (like explicit bias), (2) failures to properly exercise control (e.g., manifestations of implicit bias that could have been avoided by suitable reflection), and (3) failures to enhance the capacity for control (e.g., failing to search for remediating measures available in the local moral ecology, when it would be reasonable to do this). And here’s a viable version of character theory: people are responsible for character defects just in case those defects could have been remediated with reasonable effort, using the resources of the local ecology. In case it’s not obvious, here’s why the local ecology matters for character theory. If Smith is a misogynist because he lives in 1950s middle America and doesn’t have access to good examples of egalitarian behaviour, while Jones is a misogynist in present-day New York just because he hates women, Jones has worse character than Smith, because he’s not only a misogynist, he’s also indifferent to women’s interests. That is, Jones exhibits a greater degree of indifference than Smith (see my 2015 paper and my 2013 paper for lengthier examples, and see also Fricker 2012). Some theorists assume that tracing doesn’t apply to character theory, but that’s false. We have to trace the causal source of a character defect, to see if the character defect is amplified by indifference to available reasons.

These are two viable versions of control theory and character theory that present plausible alternatives to responsibility nihilism. But there’s a third option that may seem to be a better fit with what I’ve said so far. It’s a consequential approach, along the lines of Vargas’ agency cultivation model (2013). On that view, we’re responsible to the extent that praise or blame is likely to enhance our agency (very crudely put). Here’s how this view works. Suppose that Smith is a misogynist who berates all the women in his life, but Smith is still a moral agent (not a full-blown, unresponsive psychopath or something of that nature; he has some vestige of the capacity to respond to reasons). Blame might function to enable or enhance Smith’s capacity to respect women (and it might function that way for all misogynists – let’s suppose that this is its general effect). So Smith is blameworthy. This approach, because forward-looking, might seem to eliminate the need for tracing, which might seem to be a desideratum, since tracing is hard. But I don’t think it does. First, we need to know if Smith is a misogynist, as opposed to, say, a foreigner who doesn’t know he’s using a misogynistic slur, or a brain-washing victim, or someone whose family is being held hostage on condition that he demean his female acquaintances, etc. I artificially stipulated that Smith is a misogynist above, but in real life, we need to discover things about a person’s circumstances to make correct moral appraisals, so we need to get to know people and inquire into their lives. Second, we might want to consider whether Smith had opportunities to develop a more egalitarian sensibility – control-based considerations. The point is, even on a forward-looking account, we need to know things about an agent’s capacities and environment, and so we need to do some non-negligible amount of tracing. We can’t just guess what someone is like on the basis of one time-slice. People are notorious for jumping to conclusions, but to be responsible in our responsibility attributions, we need to be committed to giving fair consideration to relevant data.

The upshot is that there are convincing arguments against narrow versions of control theory and character theory, but they don’t force us down a slippery slope to responsibility nihilism. There are viable (extended) versions of control theory and character theory that we can adopt; and consequentialism is also an option. But we should not, I think, assume that we can do away with tracing on any of these alternatives. If anything, once we grant that the moral ecology is relevant to responsibility –  more relevant than we might have previously thought – we have to extend the scope of tracing beyond the agent’s material brain. But I think that we implicitly do this anyways (in our ordinary judgments of praise and blame), which is why we consider people from past times and foreign cultures to be less blameworthy for certain infractions. Yet some accounts of responsibility don’t adequately explain this kind of contextual assessment – they don’t sufficiently appreciate the significant of context. Circumstances matter because they co-constitute, enable, and support – or conversely, impair – the capacities that underwrite responsible agency.


Here’s how all of this relates back to implicit bias. As I said in my first post on implicit bias, it’s not at all clear how wide the scope of control has to be for responsibility to obtain. If we think that reasonably-available counterfactual circumstances are morally relevant, we can hold people responsible for exhibiting implicit biases if they failed to use remediating measures that were locally available, provided that this was a reasonable expectation. People with special duties – people on hiring committees, for example – have stronger reasons to use these measures, and thus are more susceptible to blame for relevant omissions. And this is true even if they lack responsiveness to such measures now, provided that they could have acquired suitable patterned sensitivity at some time in the past, by a reasonable effort. Ordinary people can be blamed if we think that it was within their ability to avoid manifesting implicit bias through some reasonable act of will. The case for blame is stronger if we admit counterfactual circumstances into the equation, because then we have grounds for saying that someone is (indirectly) responsible for an omission, just in case a counterfactual enabling circumstance was within reach. This brings the view somewhat closer to modern character theory in its scope for attributing blame.


Implicit bias: The limits of control/charater



I’m going to write an entry pursuant upon my previous post on responsibility and implicit bias. (Find it here). There, I noted that there is no agreed-upon definition of ‘control’ (see Fishcer 2012, Levy 2014). And there’s also no agreed-upon definition of character (see, e.g., Angela Smith 2005, Holly Smith 2014, David Shoemaker 2014, and George Sher 2010 for different definitions of character). And this is likely why there are debates about the scope of moral responsibility within in each camp, as well as between camps. Some control theorists think that we can be responsible for the manifestation of implicit bias and others think that we can’t. And ditto for character theorists. It is debatable whether implicit biases, which are neither explicitly valued by the agent nor under the agent’s direct control, are blameworthy.

Let’s start with control, which is, I think, the easier metric. How ‘narrow’ or ‘internal’ is control? That is, is control over motive M merely a matter of whether we have conscious access to M? Certainly not, since we lack access to a great many mental states that are patently blameworthy. We take it that Smith can be responsible for killing a pedestrian when driving under the influence, because Smith was in a sound state of mind when he started drinking. This is a ‘benighted omission,’ in which Smith lacks control at t only because he impaired his reasons-responsive capacity at t-1. In this case, Smith bears ‘indirect responsibility’ for hitting the pedestrian because he had ‘indirect’ (prior time) control. The only thing that would excuse Smith would be if he were drunk only because he had been subjected to clandestine brainwashing by an evil neuroscientist who induced him to drink, or something of that nature – or so people tend to assume. (There are innumerable examples of ‘Frankfurt-type’ brainwashing scenarios, including in Fischer 2006, 2012).

The point is that direct conscious control at the time of action is not needed. Virtually every theorist agrees on this point. Only some kind of ‘indirect control’ is necessary. But it is still an open question how broad the scope of indirect control should be. According to Levy (2014), we aren’t directly responsible for the effects of implicit bias because we have at most ‘patchy’ control over these states, but we might be indirectly responsible for their effects. He doesn’t say much about indirect responsibility, however, which is a shame because this is arguably the more interesting question. If Smith selects a White male job candidate due to implicit bias, did he have indirect control over his choice? As Holroyd points out, we have indirect control over implicit biases through the use of strategies such as implementation intentions, countersterotypical exposure, explicit belief formation, and so on. So in principle, Smith could be held responsible on this basis. But much of this debate hangs on how reasonable it is to expect someone to preemptively use these strategies. As Levy states,

“an agent may lack direct responsibility for an action caused by their implicit attitudes, because given what her implicit attitudes were at t, it would not be reasonable to expect her to control her behavior, or to recognize its moral significance, or what have you. But the agent might nevertheless be fully morally responsible for the behavior, because it was reasonable to expect her to try to change her implicit attitudes prior to t” (2014).

So indirect responsibility ends up being tied to reasonable expectations. Is it reasonable to expect someone to use implementations intentions etc., prior to taking a certain action? It’s natural in our current climate to expect employers to take steps to remediate implicit bias, but is it reasonable to expect this of ordinary people? Suppose that Jones, an ordinary person with no special responsibilities, make a sexist joke, thinking that he’s being witty. He has no idea that he’s being sexist and is explicitly committed to egalitarianism. Is he responsible because he could have used remediating measures against implicit bias in the past?

If so, the scope of direct responsibility turns out to be very slight, while the scope for indirect responsibility turns out to be enormous. In fact, the control theory veritably collapses into character theory! Character theorists (for the most part) hold that we can be responsible for unconscious omissions such as implicit biases, provided that those mental states are caused by certain parts of our motivational set (as opposed to extraneous factors). So a character theorist might hold that someone is not responsible for acting under hypnosis (because this is an ‘alien’ motivation, external to the agent), but responsible for deeply-entrenched implicit biases that reliable cause patterns of behaviour. Well, that’s what control theorists would hold someone indirectly responsible for, too.

Consider for a moment Sher’s version of character theory (2010). He says that we’re responsible for an omission if it’s one that a reasonable person in our position would not have committed, provided that the omission stems from our own motivational psychology. So, we’re responsible for omissions for which a reasonable person could complain against us. That sounds a lot like the control view as I have just described it. On the latter, we’re responsible for omissions if we could have acquire the capacity to avoid them, by taking reasonable measures – by doing what a reasonable person would do in the circumstances. Construed this way, control view and character view turn out to be very similar.

Moving forward, let’s assume that a person can be responsible for A just in case it would have been reasonable to expect the person to acquire the capacity to control A at some point in the past. This somewhat circumscribes the scope of control, forbidding acquisition measures like killing someone, or severely injuring oneself, to enhance control. What’s ‘reasonable’ is still very debatable, but I won’t settle it here – let’s just grant that it rules out killing, death, and severe injury. Still, this doesn’t circumscribe control very much, because we still need to determine whether someone had the capacity within reason to acquire a better capacity for control, and doing this still requires careful analysis.

These remarks are somewhat speculative, because control theorists haven’t said much about indirect control. Fischer and Levy write a lot about direct responsibilility, but not much about indirect responsibility, except to note its existence. Yet direct responsibility, I think, turns out to be pretty irrelevant. Does it matter if we’re directly responsible for something, versus being indirectly responsible? If someone acts on implicit bias only because she failed to critically reflect on this mental state, is she more blameworthy than someone who fails to use remediating measures? Not obviously. Levy says that ‘indirect responsibility can underwrite a great deal of praise and blame; it is not necessarily a lesser kind of responsibility.’ We don’t have an argument to the effect that direct responsibility is a special, worse kind of responsibility, just because it is a more remote kind. Maybe direct responsibility matters because it tells us something about a person’s capacities: someone who is unaware of her implicit biases might be less capable of suppressing those biases. But now this seems irrelevant, since the critical question is not, ‘is S reasons-responsive now,’ but, ‘was S capable of becoming reasons-responsive at some ‘suitable prior time’ (to quote Vargas 2005)? We must trace reasons-responsiveness back – maybe way back – to prior times when S had the opportunity to improve herself. If indirect responsibility is a thing, the buck no longer stops at the agent’s direct capacity for control; we need to see if she was ‘indirectly,’ as some prior time, capable of acquiring direct control.

So for this reason, I don’t really understand why we need to be concerned with consciousness or direct responsibility at all. Correct me if I’m wrong here.

Indirect responsibility is slippery. It’s epistemically slipper. Here’s why: it requires that we ‘trace responsibility’ back to a ‘suitable prior time’ at which the agent could have foreseen the long-range consequences of her choices. This is what Vargas calls ‘the trouble with tracing (2005). Fischer subsequently responds, basically, that tracing isn’t really a problem, because we can generally foresee in ‘coarse grained’ terms what the consequences of our choices will be (2012). But I doubt this. I think we’re pretty terrible at predicting the future, and we’re terrible because we’re susceptible to all kinds of cognitive biases, including but not limited to implicit bias. (See also: self-serving bias, confirmation bias, confabulation, rationalisation, and all these). So, we’re generally bad at forecasting outcomes.

This is one dimension of the epistemological problem, but I’ve just adverted to another one: we not only need to determine whether someone satisfies the forecasting condition directly, but we also need to determine whether the person could have acquired a better forecasting capacity at some ‘suitable prior time.’ So we need to trace back beyond the person’s immediate capacity for forecasting (and beyond the person’s immediate capacity for control more generally), to see if the person had a prior opportunity to refine that capacity (or those capacities), in light of her circumstances. (Our circumstances are the resources that we use to acquire and hone our capacities, so they’re relevant here). Suppose that Jeff, the middle-aged middle manager, is a jerk, and that Vargas is right that Jeff could not foresee in his halcyon youth that he would one day become a jerk. This doesn’t settle the question of whether Jeff is all-things-considered responsible, because we need to trace back further still to see if Jeff could have acquired the capacity to foresee that he would become a jerk. Because if he could have, then he may be indirectly responsible for failing to acquire or hone that capacity, and thus indirectly- indirectly responsible for becoming a jerk. He’s responsible at n degrees of remoteness for his jerkiness.

If this is right, then control entails even more tracing than Vargas’ 2005 paper suggests. What I mean is, we can’t just home in on the agent’s brain at time t-n (some time in the agent’s personal history), and assess whether the agent had a particular capacity at that time. We need to abstract away from the agent’s capacity, to see if the agent could have enhanced, modified, or refined that capacity, using the resources of his environment. The capacity for control is not just a matter of an agent’s internal (physical, psychological, cognitive) capacities, but the relationship between those capacities and the world. So we can’t just trace back and look at the agent’s endogenous capacities at t-n. We have to look at those capacities, and the local environment, and the relationship between those two things.

Maybe Vargas intends for tracing to mean exactly this. In his later book (2012), he places a lot of emphasis on ‘the moral ecology,’ and the moral ecology consists of the conditions that support moral agency (2013). This view is a corollary of his agency-cultivation model. But I think that it’s informative to frame this perspective as a response to control theory, since it implies that, while control is relevant to responsibility, control is not an endogenous capacity, in the agent’s mind-brain, i.e., something that we can evaluate, in principle, with a brain scanner. Control is a relational property, between an agent and the environment.

Similarly with character: if character is a reliably-manifested property of an agent, then it’s a feature of the interaction between the agent as an individual and the environment. Sher’s view captures this insight: he sees character as a set of physical and psychological states that interact to produce an agent’s “characteristic patterns of thought and action” (2010: 124), and he describes this account as ‘suitably interpersonal (2010: 125). But insofar as a person is responsible for only ‘characteristic patterns,’ we need to identify the cause of a person’s action, to see if it is part of his character or not. If Jeff is curt with an employee, is it because he’s a jerk or just because he’s sick? If the latter, this isn’t part of his character. So character theory also requires tracing, though to a much more limited degree.

I’m going to continue this train of thought in the next post.

Responsibility and sympathy?


Tonight I was thinking about the relationship between responsibility and sympathy, and more specifically, whether there is one.

I could be wrong, but I don’t think that a lot of people have written about whether sympathy should affect our reactive attitudes. A lot of people have written about whether sympathy (or empathy) is required for someone to be a responsible agent – that is, for someone to be susceptible to moral address at all; but not much has ben said about whether, when contemplating someone who is unambiguously a moral agent and has done wrong, sympathy ought to mollify our blame, indignation, or resentment toward that person.

It might be natural to think that sympathy ought to mollify blame, but this isn’t obvious at all. On a strict retributivist view, a person deserves blame just in case she has done wrong and possesses certain agential capacities (rationality, reasons-responsiveness, second-order reflection, or what have you); whether the person has suffered pitiable hardships (which have not undermined her agency) has no bearing on whether she deserves blame. On a strict retributivist view, if you know that Smith is a moral agent and Smith has committed a murder, and you find out that Smith has had a wretchedly abusive childhood, and you feel sorry for him on this account, this is no reason for you to revise your original moral stance. Your natural sympathy is just a meaningless emotional effluence. Smith deserves blame because he satisfies all the retributivist criteria for blame: he committed a violation and he’s a moral agent. Your feelings about his plight are irrelevant.

As I’ve noted in earlier posts, Strawson did think that ‘peculiarly unfortunate formative circumstances’ could attenuate blame, but not because they elicit our sympathy. (At least, he never talks about sympathy). His rationale seems to be that these kinds of circumstances can prevent someone from developing any moral personality at all, and from being responsive to the reactive attitudes. Such people don’t satisfy the agency conditions of retributive blame – they’re not moral agents. But here, I’m talking about people who are responsible agents, but who had really challenging lives. Not necessarily such bad circumstances that they turned out to be psychopaths, but still really hard lives.

Watson (1987) wrote an interesting paper on the reactive attitudes with a section called ‘sympathy and antipathy,’ where he notes that murderous psychopaths like Robert Harris (a famous case) elicit a mix of sympathy and antipathy in us. We feel ambivalent because we see Harris as both a villain and a culprit. But Watson doesn’t take this to mean that sympathy should move us to excuse Harris. He thinks that whether we should blame him depends on whether he’s capable of responding to moral address, which is a lot like Strawson’s view. So it doesn’t seem like sympathy should play any role. All that matters is (a) whether the individual has done wrong, and (b) whether the individual is a moral agent. Sympathy doesn’t play a role in this equation. It’s just an irrelevant moral remainder.

But what if we don’t take a retributive stance? What if, like Vargas (2013), we think that blame functions to enhance moral agency? Well, then I think that sympathy is relevant, because if we lack appropriate sympathy toward someone, this can damage the person’s moral agency. Suppose that Salima has had a wretched life, but is still, by all appearances, a moral agent. In fact, Salima is surprisingly sensitive. But Salima has committed an infraction. If we blame Salima unsympathetically, it might make her a worse person; it might alienate her further from the moral community. People who have had wretched lives are usually already alienated, since wretched lives typically have alienating features: neglect, abuse, etc. Being unsympathetic to such people can push them of the fold. Towards traumatized people, we might decide to attenuate our blame, or suspend it altogether.

Should we adopt a stance of universal sympathy, and eschew the reactive attitudes altogether? Strawson seemed to think that this was impossible. On a forward-looking account, it might not be desirable, inasmuch as some people might benefit from blame. Maybe over-privileged people who are too self-entitled. Maybe people with ‘affluenza’ like Ethan Couch, who have never been properly held responsible. This doesn’t mean that they’re not moral agents, just that they’re very irresponsible moral agents. Perhaps, unlike people with wretched lives, they suffer from too much privilege, and blame would correct that.

This is obviously very speculative. Whether blame accomplishes anything is an empirical question. We need to study the effects of blame – blame as a reactive attitude – to see if it’s justified on an agency-cultivation model. This is a separate question from whether punishment accomplishes anything – punishment such as locking someone up in prison. I doubt that punishment is ever as effective as rehabilitation, and so I doubt that punishment is ever strongly justified. But theorists who write about the reactive attitudes aren’t talking about punishment, they’re talking about cognitive-conative states. It would be worth studying and writing about whether deploying these kinds of attitudes can be effective in certain types of cases, and if so, which types. Unfortunately I can’t do that here.


Moral responsibility and implicit bias



Today I want to talk about moral responsibility and implicit bias. I have to confess that I’m not quite up-to-date on the implicit bias literature, including this new book, “Implicit Bias and Philosophy,” eds. M. Brownstein & Jennifer Saul (2016), so this will be an informal treatment.

As I noted in previous posts, some of the debates in moral philosophy centre on whether a person can be responsible for unconscious/implicit/system-1 states or processes. There are disagreements in both major camps: character theory and control theory. That is, there are disagreements between camps, and also within camps. Here’s an example of some fractures.

On the control theory side, Jules Holroyd (2012) argues that we cannot be responsible for having implicit biases, since we lack direct control over them, but we can be responsible for manifesting implicit biases insofar as we failed to put into place strategies for preventing them from influencing our behaviour – strategies like anonymizing job applications. Neil Levy similarly (2014, 2014b) argues that we cannot be responsible for manifesting implicit biases because we don’t have personal-level control over them.

Amongst character theorists, Angela Smith (2005) thinks that we can be responsible for implicit biases because they can reflect our evaluative judgments – our character – whether we explicitly endorse them or not. On the other hand, Holly Smith (2014) (not to be confused with Angela) argues that we cannot be responsible for implicit biases because they don’t reflect our ‘full evaluative structure’: they’re recalcitrant, alien states, that are outweighed by our explicit commitments. Levy similarly holds that implicit biases are not blameworthy on character theory because they’re ‘too alien to the self to ground responsibility.’

Obviously this doesn’t exhaust the possible responses to this question. I just wanted to point our some of the fault lines in the literature and then throw in my own two cents.

I feel like we can be responsible for acting on implicit biases, even if we don’t have direct reflective control over them, and even if they run afoul of our explicit commitments. I’m not going to make a theoretical case for this intuition: I’m just going to tell a kind of story.

When I was searching for pictures for my post on responsibility and attachment theory the other day, I did an image search for ‘attachment theory’ to find some stock photos. I had already decided that I wanted to use fairly generic, mostly cartoon-y, images for all of my posts. These are the first three images of human figures that I found, and they’re typical examples:


Notice anything? These are all pictures of White women with White babies. (The third picture is of a White woman’s hand holding a White baby’s hand – another typical image). The frequency of this particular type of image is troubling for several reasons, including the following. First, it marginalizes interracial families (like my own extended family). Second, when interpreted against the backdrop of western patriarchal colonialism, it implies that White people are more concerned to ensure that their (White) children grow up with a healthy attachment style: that this is a special concern for White mothers. And third, it suggests that women (not men) are primarily responsible for ensuring that their children develop secure attachment. And conversely, when children turn out badly, mothers are to blame.

The pictures of White dads and babies were way down the list. And the pictures of Black dads and babies were way, way, way down the list. No one curated this list, so I’m not saying that there’s a biased curator behind the results. But the list reflects the top hits of Google users, so it reflects a culture-wide implicit bias – a composite sketch of our implicit biases.

It’s worth noting that attachment theory was originally a theory about women’s parenting abilities. John Bowlby’s original formulation of the view was called ‘maternal attachment theory,’ and it posited a lack of proper bonding with the mother as the main source of later emotional problems in the child. In a report for the World Health Organization titled ‘Maternal Care and Mental Health,’ Bowlby wrote that “the infant and young child should experience a warm, intimate, and continuous relationship with his mother in which both find satisfaction and enjoyment” or else the child was at risk of serious mental health problems (1953). To be fair to Bowlby, women were obliged to do a majority of the care-taking in his time, but only because they lived in a patriarchal culture where men were generally unwilling to do their fair share. This doesn’t mean that women were exclusively, or primarily, blameworthy for their children’s emotional problems. Bowlby also lived at a time when it was natural to assume that parents formed heterosexual pair-bonds, which explains his assumption that there would even be a mother, and only one (cisgender) mother, paired with a (probably cisgender) father. In other words, the theory, in its original formulation, is archaic. That’s why people now call it just ‘attachment theory,’ and expunge all of the (implicitly) heteronormative, racist, and sexist parts of the original theory.

It’s problematic that in 2016, the Google image page for attachment theory still reflects Bowlby’s original version of the theory: the idea that women have a responsibility to foster secure attachment in their babies in the context of a heteronormative same-race family.

Now, here’s how this discussion is related to responsibility. Yesterday when I was browsing the pictures, I unreflectively chose the first image in the list: the first one posted above. I was prepared to insert it at the top of my entry, under the title. I wasn’t being very discriminating: I just needed stock images to divide my entries, to make it easier for the reader to distinguish one from the other. I didn’t much care about the image. Then I scrolled down and noticed that most of the pictures were of White mothers and White babies. And then I thought about Bowlby’s implicitly sexist ‘maternal attachment theory’ from the 1950s, and how those Google images implicitly reaffirm Bowlby’s problematic conception of the ‘healthily-attached American family.’ Then I decided not to use that picture, and instead chose a generic black-and-white image of two hands. Then I did a special search, and paired the original picture with a less ambiguous one of a Black man’s hand holding a child’s hand. I didn’t want to perpetuate Bowlby’s archaic conception of the family.

I think that if I had hastily posted a stereotypical stock image, it would have been a manifestation of implicit bias, because it would have reflected my tendency to see this kind of stereotypical image as unproblematic. If someone had said, ‘you chose a stereotypical image for your picture, and it reflects implicit bias on your part,’ I would have been inclined to agree. But even if I hadn’t been acting on implicit bias, I would have been making a morally problematic choice, something for which I might be blameworthy provided that I satisfy other conditions – having the capacity for control, being a certain kind of person.

On either interpretation, the choice is problematic.

Now, when I made my original selection, did I have control over my decision? In one sense, no. I didn’t realize at the time that I was making a criticizable choice. On the other hand, with more reflection, I could have made a better choice. (Fortunately, I did reflect more; but this was something of an accident. What if I had been tired, or hungry, or in a hurry?). The idea of control is elusive. Do we have control over a certain decision if we make the decision in a hurry, and we could have made a better decision with more time? Do we have control if we act on a bad decision, when we could have solicited advice from another person prior to so acting?  What if we solicit feedback from the wrong person, when we could have gotten better advice from someone else? Is this scenario too remote to say that we had ‘control’ over our decision? In other words, how context-bound is control? Is our capacity for control ‘narrow’ or ‘extended’?

Here’s one of the problems with the notion of control (reasons-responsiveness). Our reasoning capabilities are ‘bounded,’ i.e., limited by available information, available time, and the mind’s information-processing ability (H. A. Simon 1982). This is why people perform differently in different contexts, as situationist psychology shows: moral competency is situation-sensitive (e.g., Doris 2008). A person might be better able to notice and respond to moral reasons in context A than in context B. We might be in a better position to make a decision after we have taken a nap, or had a good meal, or consulted with a trustworthy advisor, or taken a class, and so on and so forth.

It’s instructive to consider that both food and sleep (amongst other creature comforts) affect reasoning ability. Judges tend to give more lenient verdicts after lunch (Danziger et al. 2011). Sleep enhances memory processing and emotional brain reactivity (Walker 2009), and facilitates creative problem solving (Mednick 2009). Robert Louis Steven claims to have come up with the plot for ‘Dr. Jeckyl and Mr. Hyde’ during a dream, and this is also how Mary Shelley conceived of Frankenstein’s monster. Perhaps people are more responsive to relevant reasons when their basic necessities have been met – when they are at peak cognitive performance. Situations also make a difference. A person might be unwittingly sexist until taking a university course in feminist philosophy. The person was, in a broad sense, ‘capable’ of responding to reasons not to be sexist all along, but he needed the right circumstance to realize that capacity. He wasn’t capable of responding appropriately using only his own cognitive faculties prior to taking the class.

The point of this discussion is to point out that the correct notion of control – whether it should be narrow or extended – is debatable. Should we hold someone responsible for manifesting implicit bias only when her physical needs are met and she is at peak cognitive performance? Or also when she is hungry, tired, sick, etc.? Or also when she failed to take an interest in the patriarchal, colonial, heteronormative history of her own culture?

I don’t think this question has been satisfactorily decided, and it might underlie some of the disputes in the literature.

This also connects with character theory. Is someone who acts on implicit biases – someone who would unreflectively post problematic content on social media, or unreflectively say problematic things – ultimately not responsible for those behaviours because they are at odds with her full evaluative structure? Or does her behaviour reflect a lack of appropriate concern? If the latter, there is a case to be made that her indifference makes her responsible for manifesting implicit bias, because her indifference is really what defines her evaluative structure. She’s more indifferent than she is concerned about her moral integrity.

I can’t pretend to be able to resolve these issues. This was mostly a reflective exercise for my own benefit, and a way of understanding some of the debates in the responsibility literature.

I am aware of some good work that makes inroads into this debate. First, Levy has a couple of articles that argue that an agent’s responsibility-relevant capacities are extended beyond the agent’s mind-brain (2008, 2014); and Doris has a book that argues that agency is socially-embedded and interpersonal (2015). We can see earlier precedents for these views in feminist philosophy, including relational accounts of autonomy (e.g., Code 1991, Friedman 2003, Oshana 2006). Unfortunately, feminist philosophy hardly ever gets cited in non-specialized journals. But a relational/extended account of responsibility owes a lot to feminism.