2. Contemporary moral responsibility judgment

If moral responsibility judgment evolved, then there is a path that goes from its beginning to its current form. This section offers a characterization of contemporary moral responsibility judgment and describes some of its key components.

Moral responsibility judgment is part of the broader practice of holding people responsible. That practice involves a number of social and psychological elements, including values and beliefs, as well as emotional and behavioral patterns that lie at the core of what we commonly refer to as human morality. I use the term “responsibility episodes” to refer to those events in which someone is held responsible for having done something that is taken to be either good or bad according to some normative standard. It is hard to characterize, in general terms, what makes something a responsibility episode (Zimmerman Reference Zimmerman, Clarke, McKenna and Angela2015, 52–54), but there are some standard examples. Some responsibility episodes are negative, such as expressions of blame or punishments, and others are positive, such as praise and reward. The person who holds someone responsible is the author of the responsibility episode, and the one who is held responsible is its target. The author of a responsibility episode is usually different from its target, but there are self-directed responsibility episodes as well, such as self-blame or guilt.Footnote ¹ Also, the target of a responsibility episode is usually held responsible for having acted in some way, but other categories of human phenomena can also motivate responsibility episodes, such as omissions, attitudes, emotional responses, and character traits. For simplicity, I focus on actions. A central role moral responsibility judgment plays is to regulate responsibility episodes.

What is a moral responsibility judgment? At its core, saying that a person is morally responsible for an action means, at the very least, that the person can be the target of some responsibility episode that can be described as deserved, appropriate, or simply justified (I will shortly distinguish between two layers of the justification under consideration). Virtually all philosophical accounts of moral responsibility mention that general aspect. I look for further details on empirically informed accounts, since they provide a more accurate view of how moral responsibility judgment figures in present-day human moral psychology. Two such accounts are Malle, Guglielmo, and Monroe’s (Reference Malle, Guglielmo and Monroe2014) path model of blame and Hoffman and Krueger’s (Reference Hoffman, Krueger, Menon, Nagaraj and Binoy2017) model of the neural bases of third-party punishment.

According to Malle et al., blame judgments (which they call “cognitive blame”) arise in situations where a norm violation is detected, is taken to have been caused by an agent, and the agent has violated the norm either intentionally (but for no good reason) or unintentionally (but had the obligation and capacity to prevent it). Even though the authors avoid using the word “responsibility,” they take overt expressions of blame (which they call “social blame”) to require a justification or warrant (148). Moreover, according to them, the justification of a blame episode coincides to a great extent with the basis for cognitive blame (149).

Hoffman and Krueger’s model, in turn, postulates that blame judgment and third-party punishment rely on three neural networks: the salience network, the default mode network, and the central executive network. The salience network “is involved with detecting and responding to norm violations or threats of norm violations” (215). The default mode network “integrates the assessment of the wrongdoer’s mental state with the assessment of the harm from the SN [salience network]” and, as a result, produces a “blame signal” (215). And the central executive network is involved in decisions about punishment; here “the blame signal from the DMN [default mode network] is converted into a punishment decision after integration with a wide variety of context-dependent circumstances” (216). In sum, Hoffman and Krueger’s neural model says that a punishment episode starts with the detection of a norm violation, proceeds to an assessment of the psychological involvement of the target with the norm violation, and then, under the influence of multiple and context-dependent considerations, culminates in a decision about whether and how to deliver the punishment. I take Hoffman and Krueger’s description of the salience and the default mode networks to be largely consistent with Malle et al.’s model of cognitive blame.

The models just considered allow for an understanding of the psychological components of contemporary moral responsibility judgment, at least as it figures in negative responsibility episodes. I take those components to involve at least the following:Footnote ²

1. 1. a capacity to detect norm violations,

2. 2. a motivation to respond in some way to a target who has violated a norm,

3. 3. a capacity to assess the mental involvement of the target with the violation in a way that can affect the motivation to respond, and

4. 4. a capacity to assess contextual factors that are relevant to whether and how to respond to the target in face of the relevant motivation to do so.

In light of those four components, I distinguish between two senses of justification that are present in contemporary moral responsibility judgment. First, a responsibility episode can be said to be justified in a demand sense, meaning that there is a positive motivation or demand for realizing the episode (component 2). That sense of justification is emphasized, for example, when we worry about impunity. I believe Malle et al.’s description of cognitive blame and Hoffman and Krueger’s characterization of a blame signal are accurate descriptions of the demand sense of justification involved in a negative judgment of moral responsibility (I leave it open whether or not the same characterization works for positive episodes).

In a second sense, a permission sense, one can regard a responsibility episode as justified because it is permitted for someone to realize it. The second sense is emphasized when we worry about getting a responsibility episode wrong, something that is captured, for example, in the presumption of innocence. I take the “warrant” Malle et al. (Reference Malle, Guglielmo and Monroe2014, 148) describe as involving the permission sense. And I take the transition from Hoffman and Krueger’s blame signal to a punishment episode through the central executive network to leave room for a justification in the permission sense to be part of the contextual circumstances considered (component 4).

In sum, contemporary moral responsibility judgment essentially involves considering whether an actual or potential responsibility episode is justified in two related senses. And at least four psychological components underlie a moral responsibility judgment, two of which are directly linked to the two senses of justification. The next section briefly reviews some of the literature on the evolution of the practice of holding people responsible (especially punishment) and then focuses on an account of the evolution of moral responsibility judgment developed by Matteo Mameli (Reference Mameli2013).

3. Mameli on the evolution of moral judgment

Among the components of the practice of holding people responsible, punishment has got most of the attention. In contrast, moral responsibility judgment has seldom been the subject of evolutionary theorization. This section reviews evolutionary perspectives on the practice of holding responsible and punishment, and then focuses on Mameli’s (Reference Mameli2013) work, which targets the evolution of moral responsibility judgment in a more direct and detailed way. I argue that Mameli’s account suffers from two problems and take that as the main motivation for an alternative elaborated in section 4.

Responsibility episodes are ubiquitous in human relations. Most current societies have formal institutions in charge of punishment, which themselves have a long history (see, e.g., Morris and Rothman Reference Morris, Rothman, Morris and David1995). And human relations, from family interactions (Laforest Reference Laforest2002) to interactions in the workplace (Aquino, Tripp, and Bies Reference Aquino, Tripp and Bies2001) to interactions among strangers (Svennevig Reference Svennevig2012) can involve forms of moral appraisal characteristic of moral responsibility, such as thankfulness, resentment, indignation, acknowledgments, among many others (Strawson Reference Strawson and Pereboom1962; Malle, Guglielmo, and Monroe Reference Malle, Guglielmo and Monroe2014). There is evidence that children as young as 8 months have a primitive understanding of patterns associated with moral responsibility (Hamlin et al. Reference Hamlin, Wynn, Bloom and Mahajan2011). And even other species, such as chimpanzees, can sometimes deliver punishments and rewards (Apicella and Silk Reference Apicella and Silk2019, R449).

The ubiquity of responsibility episodes, and especially the fact that even other species seem to have rudiments of them, has led many to hypothesize an evolutionary origin. Available theories describe responsibility episodes as part of the mechanisms that allowed for the evolution of the peculiar capacities for cooperation, altruism, and morality found among humans (see, e.g., Boehm Reference Boehm2012; Tomasello Reference Tomasello2016; Boyd and Richerson Reference Boyd and Richerson1992). Responsibility episodes—punishment, in particular—usually figure in those theories as a potential solution to the free rider problem. Because cooperation and altruism can be costly to their authors, and given that non-cooperators can benefit from the cooperation of others without incurring any costs, it may seem puzzling that cooperation ended up being naturally selected. Punishment promises to offer a solution to the puzzle by eliminating the advantage free riders might otherwise obtain.

In contrast to responsibility episodes, judgments of moral responsibility were less often the subject of evolutionary theory. To be precise, Malle et al. (Reference Malle, Guglielmo and Monroe2014), and Hoffman and Krueger, do make connections between their models and evolutionary considerations. Hoffman and Krueger (Reference Hoffman, Krueger, Menon, Nagaraj and Binoy2017, 211, 217), in particular, hypothesize that third-party punishment was an adaptation to more complex forms of social life and an alternative to the faster and more automatic second-party punishment. Tomasello (Reference Tomasello2016, 33–34, 61, 67–70) also touches on themes such as resentment, desert, and protest, which are central to moral responsibility judgment. But those connections fall short of providing a more systematic account of the evolution of moral responsibility judgment.

Mameli (Reference Mameli2013) addresses the evolution of moral responsibility judgment more directly, in the context of a more general account of the evolution of moral judgment. He understands moral judgment as consisting in a set of emotional dispositions. Simply put, his expressivist account says that a judgment that A (a type of action) is morally required involves four emotional dispositions (see Mameli Reference Mameli2013, 905):

• D1: a disposition to feel anger at those who act in ways that violate A;

• D2: a disposition to feel guilty about having oneself acted in ways that violate A;

• D3: a disposition to feel anger at those who do not have dispositions D1 and D2; and

• D4: a disposition to feel guilty about not having dispositions D1 and D2 oneself.

For the present purposes, it is important to note that Mameli sees an understanding of moral responsibility judgment as embedded in the dispositions for meta-anger (D3) and meta-guilt (D4). Those dispositions give rise to what he calls “meriting”:

According to this view, responsibility episodes (at least the negative ones) are ultimately based on anger directed at those who violated some expected behavioral standard. And a judgment of moral responsibility, which is a component of moral judgment according to Mameli, is taken to consist in dispositions for meta-emotions (meta-anger and meta-guilt) directed at those who do not feel anger at first-order violators.Footnote ³

Before assessing the plausibility of Mameli’s understanding of moral responsibility judgment and his hypotheses about its evolution, it is worth noting that his account and the one I offer later on have slightly different goals. Mameli starts with an account of a fully-developed moral judgment and explores its evolutionary origins (i.e., he is concerned with the evolution of what he takes to be distinctively moral capacities in humans). My account, in contrast, is more centrally concerned with the early evolution of a moral responsibility judgment (i.e., with a judgment that guides certain types of responses, especially in the sphere of morality) without addressing its distinctively moral character. Given these differences, some components may be essential for Mameli’s account but unnecessary for mine. In other words, it is possible that what I characterize as an original moral responsibility judgment would not count as fully moral by Mameli’s standards. I nonetheless expect my account to capture something that eventually became part of a fully developed morality, whatever that may be.

Mameli (Reference Mameli2013, 921) situates his hypotheses about the evolution of moral judgment in a broader view according to which human cooperation and altruism evolved in the context of large-game hunting (Boehm Reference Boehm2012). According to that view, around 500,000 years ago our ancestors became collaborative large-game hunters and were thus able to access new sources of food that were unavailable for solitary initiatives. The evolutionary pressure for that change could have been a much earlier scarcity of our ancestors’ preferred type of food (plants) due to climate change and competition with other species (Tomasello Reference Tomasello2016, 44). The products of large-game hunting, however, opened the door for free riding: uncooperative individuals might take advantage of the collaborative efforts of others. Mameli follows Boehm in counting bullies and cheaters as examples of free riders. In this context, according to Mameli (Reference Mameli2013, 921–22), D1 dispositions emerged as a mechanism that motivated the delivery of punishments, which in turn favored the selection of both cooperators and individuals with D2 dispositions that could work as a form of moral conscience or self-control.

Delivering punishments, however, is another example of a costly activity that opens the door for exploitation, now in the form of a second-order free rider problem: individuals can be better off by benefiting from punishment implemented by others, and hence each individual has an incentive not to punish despite the collective benefits of someone acting otherwise (see, e.g., Henrich and Boyd Reference Henrich and Boyd2001, 80; Boyd and Richerson Reference Boyd and Richerson1992). It is in the context of the second-order free rider problem that Mameli places judgments of moral responsibility. He hypothesizes that meta-emotional dispositions D3 and D4 would be able to ensure that second-order free riders are punished just like first-order ones.

Mameli’s account is a valuable attempt to develop a fine-grained description of the components of morality within the broader context of evolutionary theories. That virtue notwithstanding, the account suffers from two main problems that are crucial for an account of the evolution of an early moral responsibility judgment. One is that the evolutionary hypothesis about the selection of second-order dispositions is weakly supported. And another is that analyzing meriting in terms of second-order emotional dispositions is conceptually implausible.

On the first problem, Mameli relies on Boehm’s work to a great extent, but he disagrees about how exactly punishment practices evolved. According to Boehm, the solution to the first-order free rider problem that evolved in our species was already able to prevent the emergence of a second-order problem. Boehm relies on ethnographies of what he calls “Late Pleistocene Appropriate” (LPA) groups, which are a selection of present-day foraging societies taken to represent a close approximation of how humans lived in Africa around 45,000 years ago (Boehm Reference Boehm2012, 79). As the ethnographies attest, LPA groups can control actual and potential cheaters and bullies through a collective punitive enterprise. By acting as a group, greater punishment power is available and retaliation is discouraged. Also, Boehm notes that the ethnographies have no mention of second-order punishment: “so far in my survey of group punishment among fifty LPA hunter-gatherers […], the punishment of nonpunishers is never mentioned in the hundreds of ethnographies even though punishment does take place so regularly—and even though there are plenty of abstentions” (206). If first-order dispositions evolved and made first-order punishment so common, why are manifestations of second-order dispositions so uncommon?

In response, Mameli says that

I think the passages emphasized above include a misinterpretation of what Boehm says (1) and an unsupported hypothesis (2). Contrary to the first part, Boehm does not seem to suggest that most, or even some, people are second-order free riders just because they do not get involved in some particular punishment episode, despite supporting its execution. More than that, Boehm (Reference Boehm2012, 206) explicitly acknowledges that eventual “abstentions need have no relation to free-rider genes.” Thus, his view seems to be that, even though some experimental studies suggest that second-order punishment can stabilize cooperation, observations in settings with greater ecological validity suggest that it is not required.

In the second passage, Mameli claims that everyone is expected to disapprove of a deviant and that violations of that expectation would be met with negative consequences. But he fails to provide any evidence (ethnographic or otherwise) in support of that claim, which is, in addition, somewhat at odds with some cases discussed by Boehm. In an instance of collective punishment, for example, members of the group that lived close to the one being targeted “were obviously staying to one side and appeared to be neutral” (Boehm, Reference Boehm2012, 206); the explanation offered is that “close relatives or associates of a deviant may choose to stand back and let others deal with him harshly.” Boehm suggests that other group members could reasonably accept that type of situation as part of social expectations related to certain roles, including family relations. Abstentions like that suggest that second-order dispositions were not needed to stabilize first-order punishment or, at the very least, that some evidence is needed to establish that they were.

Again, Mameli’s account purports to capture an essential feature of contemporary moral judgment, and it is possible that second-order dispositions are some of those features. Also, I do not dispute that second-order dispositions could, in principle, help to stabilize first-order punishment. Even so, the reasons Mameli offers to support the claim that those dispositions were part of the evolutionary history of moral responsibility judgment and punishment episodes seem uncompelling.Footnote ⁴

A second difficulty for Mameli’s account, insofar as it addresses moral responsibility judgment, has to do with his characterization of meriting. As he rightly says, meriting consists in viewing responsibility episodes “as deserved or merited” (907). But it is conceptually implausible to identify a judgment that a response to some behavior is deserved with a disposition to feel anger at some (potentially different) individual who is not disposed to feel anger at that behavior. For example, Mameli says that “regarding guilt and anger as merited reactions toward a violation just is having dispositional anger at meta-violators” (907). If meta-dispositions are necessarily involved in full-blown moral judgment, they would also be present in judgments saying that a response to a genuine moral violation is deserved. The problem is that judgments of moral responsibility are first and foremost about first-order violators themselves and only secondarily about how others should react to a violation. For example, the models by Malle et al. and Hoffman and Krueger, which aim to provide an empirically accurate portrait of moral responsibility judgment in humans (and the sense of “warrant” it involves), do not mention an assessment of individuals other than first-order violators. In the same way, when we say a person deserves a response, we are primarily saying the response is correct for that exact person.

In short, Mameli’s account ultimately describes moral responsibility judgment as a disposition for second-order anger (a disposition to feel anger at those who are not disposed to feel anger at first-order violators) whose evolutionary function was to prevent second-order free riding. More concretely, a sense of meriting consists in an expectation that everyone disapproves of deviants, and is willing to execute an act of punishment, if requested by the group, and to support those who execute punishment against retaliation. I have questioned, on two related grounds, how credible this specific part of Mameli’s account is. The next section offers an alternative which I take to rest on a more accurate description of moral responsibility judgment and to be better aligned with available works in evolutionary ethics.

4. A two-step account of the evolution of moral responsibility judgment

This section offers an account of how the two senses of justification involved in contemporary moral responsibility judgment may have evolved. The account shares some assumptions with Mameli’s account, although there are important differences in the details. I agree with Mameli that moral responsibility judgment was part of the evolutionary mechanism that stabilized responsibility episodes. I also agree that part of the reason why moral responsibility judgment (in my view, in the demand sense) was selected is that it motivates the realization of responsibility episodes. Beyond that, the main difference of the present account is that it takes the permission sense of justification as more relevant to avoid the second-order free rider problem. My main hypothesis—the social support hypothesis—says that the permission sense arose as part of a mechanism that allowed for the social support enjoyed by a potential responsibility episode to be assessed in a way that could influence the realization of the episode. The account relies to a great extent on Tomasello’s two-step theory of the evolution of human morality and on Boehm’s description of punishment practices in LPA bands.

Tomasello (Reference Tomasello2016) postulates that two steps were crucial for the evolution of the specific form of cooperation found among humans, both of which resulted from an increased interdependence among individuals. Early humans, who lived around 400,000 years ago, had as their main innovation the collaborative hunt in the context of a dyad. That accomplishment depended on new capacities, such as joint intentionality, second-personal agency, and joint commitment, which enabled individuals, for example, to pay attention to common objects in their environment, and to understand the perspective of their partners and the roles they each should play in a collaborative activity (2016, 50). Modern humans, in their turn, who began to develop around 150,000 years ago, achieved the capacity to live in larger and more diverse groups, where cooperation required, among other things, better communication capacities. While individuals in the first step learned to cooperate with well-known partners within small groups, humans in the second step needed to learn how to collaborate with less well-known members of groups that were forging their own cultural identities and that eventually needed to compete with rival groups (2016, 85).

It is relevant to emphasize how Tomasello’s account attempts to explain how human morality and cooperation may have become stable and immune to free rider problems. Tomasello focuses on processes of mutualism and reciprocity that operate at the individual level in contexts of increased interdependence among individuals, i.e., where collaboration benefits everyone involved (Tomasello Reference Tomasello2016, 13–19; Tomasello et al. Reference Tomasello, Melis, Tennie, Wyman and Herrmann2012). One advantage of the mutualistic explanation, in particular, is that free riding, although it still poses some challenges, becomes less salient (see, e.g., Tomasello Reference Tomasello2016, 13, 61). The account also posits mechanisms of social selection according to which individuals who do well in collaborative enterprises—e.g., by communicating well and being helpful to partners—may become preferred partners of joint collaboration and have, as a consequence, increased access to the benefits of collaboration. Tomasello argues that his account is more plausible than classical alternatives, including hypotheses that involve group selection through, for example, competition between groups (Tomasello Reference Tomasello2016, 12).

My account centrally hypothesizes that moral responsibility judgment began to develop in the first evolutionary step Tomasello describes. During the first stage, moral responsibility judgment in the demand sense consisted in a capacity to detect violations of proto-norms coupled with a motivation to deliver a (sometimes punitive) response. The permission sense also began to develop in the first step as part of a channel of mutual understanding among cooperators, which enabled them to coordinate a punitive response against free riders. A secondary hypothesis (see section 5) says that moral responsibility judgment is unlikely to have achieved its contemporary form, including its characteristic sensitivity to the mental states of its targets, before the second evolutionary stage. Within the second step, responsibility practices as a whole became more complex as human groups became bigger and cultural. Scrutinizing the mental lives of potential targets of responsibility episodes makes sense in that context, I argue, even though it remains open how close to current assessments of intentions and omissions that scrutiny would be.

In the context of dyadic hunting, responsibility episodes could fit in two main types of situations. First, one could face a greedy hunting partner’s attempt at taking more than half of the spoils. Tomasello thinks attempts at non-equal sharing in this context violate an implicit agreement that underlies the whole collaboration. The tacit commitment is that each party should play their role properly for both to achieve their common goal; it also involves accepting that both partners have authority to initiate sanctioning when their joint commitment is violated (2016, 68). The paradigmatic response to violations of this type is a form of resentful, albeit respectful, protest, which “does not seek to punish the partner directly, only to inform her of the resentment, assuming her to be someone who knows better than to do this” (69). That kind of protest requires simple communicative skills and can be expressed by “a simple ‘Hey!’ or a squawk” (69). As a result, the violator is expected to acknowledge and repair her own fault—thus keeping her cooperative identity—or, less likely given the implicit agreement, to face the threat of being excluded from future collaboration on which her survival may depend.

I take the expression of protest in the scenario just described as an early instance of a moral responsibility episode. But I think that scenario plays a less central role in the evolution of moral responsibility judgment than it plays in the whole story about human morality, as told by Tomasello. A second type of situation is more relevant for moral responsibility judgment, one that Tomasello also describes, although in less detail.

The second type of situation humans of the first step faced was the threat of free riders, i.e. individuals who would attempt to take some of the spoils without having taken part in the hunt themselves. This type of situation arises only after collaborative hunting is achieved: “others could come up after the kill, and they were essentially competitors—from outside the collaboration—so at some point humans also evolved the tendency to deter […] free riders by denying them a share of the spoils” (2016, 61). Tomasello elaborates on how the control of this type of free riding could have produced a sense of desert that later could also guide the division of resources among collaborators. But he does not elaborate on how the denial of a share of the spoils to non-collaborators could be carried out.

In contrast to the situation in which a greedy partner attempts to take more than an equal share, the threat of a non-collaborator is potentially more dangerous. In the context of a collaborative dyad, a vocal protest might suffice because the greedy individual may just need a reminder of a tacit agreement already at play. But, lacking one such agreement, the interaction among collaborators and external free riders might easily involve physical violence from either part. My suggestion is that it would be advantageous for cooperators to have, in addition to a motivation to respond negatively to the violator, a mutual understanding of the fact that they both shared (or not) that motivation.

Returning to Hoffman and Krueger’s model, I propose that a primitive moral responsibility judgment in the demand sense arose when early collaborators developed two dispositions that were reactive to having one’s own or one’s partner’s food threatened by external individuals. The first is a disposition of the salience network to display an aversive emotional response, which would work as a proto-norm against food robbery. The second is a disposition to display an aggressive response against violators of that proto-norm, in a way that can influence the central executive network. I propose that the assessment of a violator’s mental states (e.g., intentions) that is constitutive of contemporary moral responsibility judgment was absent at this early stage. That is to say, even though some understanding of the mental states of the external violator could be present, it was not, at this point, a factor that could make someone refrain from realizing a responsibility episode. Accordingly, the initial situation was somewhat closer, albeit not identical, to Hoffman and Krueger’s (Reference Hoffman, Krueger, Menon, Nagaraj and Binoy2017, 217) characterization of second-party punishment “as being blame plus an automatic punishment response, all rolled into one, and without the cognitive restraints we see with third-party punishment.” In a sense, early humans’ punitive response to a free rider was a second-party response, as Tomasello describes them as acting as a collective “we.” If this same “we” reacts, then it is implementing second-party punishment, according to the definition (Hoffman and Krueger Reference Hoffman, Krueger, Menon, Nagaraj and Binoy2017, 207). But, in another sense, the situation was different, because the reaction against the violator was less individualistic and automatic than Hoffman and Krueger take second-party punishment to be.Footnote ⁵ Thus, the initial moral responsibility judgment provided early large-game hunters with the ability to detect a threat to their collective goal and with the motivation to respond aggressively.

On its own, however, a moral responsibility judgment in the demand sense is not likely to become evolutionarily stable, as it does invite the second-order free rider problem. Given the possibility of retaliation from the free rider, each collaborator would be better-off by refraining to punish, if the other party punished alone. A potential solution is available if collaborators, in addition to detecting a violation and being motivated to respond negatively, were also able to assess the motivation of their partners. That would be the birthplace of moral responsibility judgment in the permission sense. The potential authors of responsibility episodes would have not just the motivation to realize those episodes, but also a sensitivity to the social support enjoyed by the episode, in a way that makes them less open to exploitation.

Even though I assume there was not, at this early stage, an assessment of the mental states of the violator, there are good reasons to assume that collaborators in the first stage could have a capacity to assess the support for a responsibility episode. The context of the joint activity involved a channel of mutual understanding among collaborators themselves (Tomasello Reference Tomasello2016, 53). Within that context, then, and under the influence of a first-personal motivation to punish, collaborative hunters could form a further joint goal to protect their collective achievements by punishing eventual attackers. In a sense, that would be simply another intermediary step within their whole joint activity: just as their mutual understanding allowed them to coordinate a set of actions to hunt successfully, their understanding of each other’s motivation to punish would allow them to respond to the attacker as a collective “we.” One such coordinated sort of response would increase the safety and success of collaborators by reducing the risk of retaliation and by increasing punishment capacity. And because the punitive response was part of a broader collaboration for food each one was already invested in, defecting would be self-defeating. Therefore, even if a moral responsibility judgment in the demand sense alone was not likely adaptive and stable, the addition of the permission sense, in the form of an assessment of the social support for the responsibility episode under consideration, makes adaptiveness and stability more likely. This hypothesis, I propose, offers a more plausible starting point for moral responsibility judgment than the one involving Mameli’s second-order dispositions.

Within Tomasello’s second evolutionary step, humans formed larger and culturally structured groups. The social support hypothesis fits in that context as well. Collective life in bigger and cultural groups likely affected the practice of holding people responsible. First of all, there were new opportunities for free-riding, as cheaters can be harder to detect in bigger groups, where interactions among strangers become more frequent (Tomasello Reference Tomasello2016, 98). A cultural organization also means that more collective norms and social control are required to maintain more complex forms of cooperation. In addition to new free-riding opportunities and more norms, two other aspects of life in bigger cultural groups could have been relevant. One is what Tomasello refers to as a “common ground,” i.e., a shared knowledge base which “meant that everyone in a group knew that everyone in that group had had certain kinds of experiences—and thus skills, knowledge, and beliefs” and that individuals knew “many important things about the minds and likely behavior of others, often without ever interacting with them directly” (93). Another element of cultural life is knowledge of the linguistic conventions that guide communication within the group (95). A shared knowledge base that includes knowledge about mental and behavioral tendencies of other group members and linguistic communication paved the way, among other things, for a new way of assessing the social support enjoyed by responsibility episodes. Boehm’s account of gossiping among hunter-gatherers helps to illustrate both points.

Following studies by anthropologist Polly Wiessner, Boehm (Reference Boehm2012, 240) notes that the !Kung people “gossip intensively when trouble is shaping up and collective action may have to be taken”:

According to these passages, gossiping helps the group both to detect violations of norms and to coordinate a collective (and consensual) response. The details of that type of communication are illustrated by a specific episode of norm violation Boehm describes, this time based on Colin Turnbull’s work on the Mbuti Pygmies. In a collective net hunt—where each hunter would get for his family what he could catch on his net but where cooperation would allow for everyone to succeed—an egoistic man named Cephu quietly repositioned his net in such a way that animals driven by other group members would run first into his net. The hunt worked well for Cephu, but not the aftermath, as his cheating had been witnessed. The episode ended up with Cephu being strongly blamed by the group, facing the threat of expulsion, apologizing, and having to return all of the spoils that he and his family intended to eat. The details of the episode are relevant to the social support hypothesis. Just after the hunt, when the group was returning to the camp in a quiet and bad mood because of Cephu’s behavior,

This is a vivid example of a punishment episode within a group of hunter-gatherers, one that exemplifies a moral responsibility judgment, in the permission sense, playing a role in coordinating a collective response that becomes safer for its authors. While individuals could be previously motivated to respond negatively to Cephu, realizing that other group members were similarly motivated helped the group to execute the punitive episode.

In summary, the present account tells a story about how different components of moral responsibility judgment evolved along the two evolutionary stages Tomasello describes. The demand sense evolved as an individual motivation toward realizing negative responsibility episodes against free-riders. The permission sense evolved in a way that enabled individuals to coordinate their responses against free riders by assessing the social support enjoyed by a responsibility episode they felt motivated to realize. The permission sense made punishment episodes safer for those implementing them and less open for exploitation. In the first evolutionary stage, social support assessment was part of the mutual understanding among dyadic cooperators; during the second stage, it became linguistic and spread over the group as a whole through the practice of gossiping.

5. Limitations and explanatory potential

The account just presented has its own limitations, but it also has some explanatory power. Both are discussed below.