Agents are often blamed for the evaluative outlook their behavior manifests. Call this attributability blame. On standard accounts, an agent’s evaluative outlook is taken to reveal who that agent really is, that is, to reveal the commitments or values central to their character.1 However, a large body of research in social psychology suggests that human agency is often radically shaped by social dynamics. Under conditions of peer pressure, internalized social norms, and other social influences, agents attend to and respond to the world in ways that diverge dramatically from who they, supposedly, really are. These responses remain exercises of agency, and we routinely assess and criticize their evaluative content. Think, for example, of a spontaneous slur or chant. This poses a challenge for standard accounts of attributability blame. When the evaluative content of an action is sustained primarily by social dynamics rather than an agent’s character, it becomes unclear how, and to whom, attributability blame should apply.

Riots provide a particularly vivid illustration of this problem.2 Decades of research suggest that in riot contexts, powerful social forces reshape how individuals attend to and respond to their environment, often leading people to act in ways they later disavow and that do not seem to reflect their own commitments or values.3 At the same time, riot behavior typically expresses clear evaluative content—hostility toward certain social groups, indignation with institutions, or solidarity with a political cause—and we ordinarily judge rioters based on that content. The resulting puzzle for attributability blame comes into sharp focus once we distinguish two types of riot participation. Some participants, whom I will call Diehards, act in ways that reflect their own evaluative outlooks, even if their behavior is intensified by group dynamics. Others, whom I will call Tagalongs, are swept up by the moment and come to express evaluative content that does not manifest their own evaluative outlooks. Diehards are clearly blameworthy for the evaluative content of their actions. But what about Tagalongs? Intuitively, they are blameworthy, and not merely for being present in the riot or for their general suggestibility, but for the evaluative content their actions express. Yet on standard accounts, this kind of social influence is often seen as excusing. How, then, should attributability blame be understood when the relevant evaluative outlook is sustained primarily by social forces rather than by the agent’s character?

In response, I develop what I call the Outlook View of attributability blame. The Outlook View retains the core insight of existing accounts—that attributability blame concerns how evaluative outlooks are manifested in people’s actions—but allows that evaluative outlooks can be sustained not only by an individual’s psychology but also by social dynamics. This enables us to account for the kind of attributability blame targeted at what I shall call socially scaffolded evaluative outlooks. More precisely, what matters for attributability blame is not that an action expresses some psychological structure of an agent but, more generally, whether an action manifests an evaluative outlook and whether the sustaining basis of that outlook, whether individual or social, supports the characteristic and integrated functions of blame.

The Outlook View has broad implications for any context in which evaluative outlooks are socially scaffolded—from peer pressure and organizational cultures to longer-term political climates. I focus on riots, however, because they serve as an epistemologically privileged test case. The social dynamics that shape riot behavior are empirically well-described, riot actions are widely treated as expressive and blameworthy, and we typically have sufficient evidence about the evaluative content of those actions to ground moral assessment. This makes riots an ideal context in which to develop and test the account before extending it to harder cases.

The rest of the paper proceeds as follows. Section II draws on social psychological research to show how riot behavior is shaped by powerful but temporary forms of social influence, and why riots pose a distinctive challenge for accounts of attributability blame. Section III argues that existing accounts struggle with such cases because they assume that blameworthy evaluative outlooks must be grounded primarily in individual psychological structures. Section IV introduces the Outlook View and its two conditions on attributability blame, Manifestation and Functional Support. Section V explains how riot behavior can satisfy Manifestation even when the relevant evaluative outlook is socially scaffolded. Section VI then shows how such socially scaffolded outlooks can support the characteristic functions of attributability blame in a distinctive but coherent way. In the Conclusion, I take stock of the account and sketch some of its wider implications.

In this section, I review some prominent theories in social psychology that are commonly used to explain social behavior. Focusing specifically on riot cases, I show how these theories support the idea that social dynamics can sustain evaluative outlooks. I then identify three core intuitions (Evaluative Content, Differentiation, and No Full Excuse) that any adequate theory of attributability blame must account for.

Central to the social psychological literature is the idea that social forces can change how we attend and respond to our environment and, hence, the evaluative content of our actions. Several partly overlapping frameworks have been developed to explain this phenomenon.

One influential framework is self-categorization and social identity theory. On this view, riot dynamics often involve a switch between personal and group identity.4 Intergroup conflict (e.g., us vs. the police) typically serves as a catalyst for this shift, causing people to see themselves as group members and prioritize collective over individual concerns. This often leads to increased ingroup solidarity and defensive or aggressive responses, even among those who would not act like this in isolation.5

A second framework, deindividuation theory, highlights how anonymity and immersion in a crowd diminish self-awareness and a person’s sense of personal responsibility.6 In such conditions, people become less constrained by moral or personal standards and more susceptible to external situational cues.7

A third framework, emergent norm theory, suggests that people in chaotic situations look to others for guidance on how to act, which results in the rapid formation of new norms that guide collective behavior.8 These norms often diverge from agents’ preexisting evaluative frameworks.9

These frameworks appeal to different psychological mechanisms to explain riot behavior, but they converge on the idea that social dynamics can influence what agents take to be significant and thus shape the evaluative content of their actions.

The common picture of attributability blame holds, roughly, that agents are blameworthy for the evaluative content of their actions only if that evaluative content reflects or expresses a morally relevant part of their individual psychology. Social psychology challenges these theories by showing that social dynamics may undermine, distort, or amplify these aspects of our individual psychology. To illustrate, consider two types of rioters.

Some riot participants are deeply committed to the values expressed in their actions, even if their behavior is subtly shaped by crowd dynamics. Let us call these participants Diehards. A Diehard is a riot participant whose actions have evaluative content that manifests their own evaluative outlook, even when the expression of that content is enabled or triggered by some social interaction.

Others act on what they take to matter in the moment but later disavow the action or feel estranged from it. Let us call these participants Tagalongs. A Tagalong is a riot participant whose actions have evaluative content that does not manifest their own evaluative outlook. Both Diehards and Tagalongs may be subject to social powers, but the cases raise different challenges. For instance, a normally reserved Diehard may join a chant only in a crowd, yet the chant expresses convictions they already hold. A Tagalong, by contrast, may also join a chant, but here the evaluative content does not cohere with their characteristic values and commitments. They are therefore likely to regret what they said or did, maybe even be uncertain why they acted in that way.

Intuitively, the differences between Diehards and Tagalongs matter for our responsibility practices. They give rise to three intuitive judgments:

Evaluative Content The moral relevance of riot actions depends in part on their evaluative content.

Differentiation Tagalongs and Diehards are not all morally on a par with respect to the evaluative content their actions express.

No Full Excuse The social dynamics of riots do not, by themselves, fully excuse riot participants from attributability blame.

These judgments are, I take it, widely reflected in how riot behavior is assessed. Evaluative Content captures the idea that riot actions manifest evaluative content. For example, participants may express hostility toward certain groups or sympathy for a political cause. Our moral assessments of rioters are partly determined by this evaluative content. We tend to judge racist rioters who block the street very differently from climate activists doing the same.

Differentiation tracks the fact that when we blame rioters, one of the things we care about is how the evaluative content expressed in action relates to the agent’s evaluative outlook more generally. For example, when sentencing rioters, judges routinely cite evidence concerning whether a defendant expressed a preexisting evaluative outlook or acted in a way that was somehow “out of character.”10

Finally, No Full Excuse reflects the fact that we often do not treat social scaffolding as fully excusing riot participants from attributability blame.11 Even when a participant’s evaluative outlook is strongly shaped by social dynamics, we may still hold them responsible for the evaluative content of their actions. At the same time, as Differentiation suggests, social scaffolding is not normatively irrelevant. More specifically, I’ll argue that the shape that attributability blame takes depends on the sustaining basis of the evaluative outlook, so that some forms of attributability blame are inappropriate while others are not.12

Although these judgments are clearest in riot contexts, they reflect a broader problem concerning socially complex cases. In many social situations, agents act under the influence of social powers that shape what they take to matter and, hence, the evaluative content their actions express. Sometimes the influence of such social powers excuses (e.g., in a chaotic panic scenario), but in other cases attributability blame seems to remain appropriate even when the evaluative outlook expressed in an action is shaped, in whole or in part, by social dynamics rather than by the agent’s own values and commitments.

Existing theories of attributability blame hold that agents are blameworthy for the evaluative content of their actions only if it manifests an evaluative outlook that is suitably based on the relevant parts of their individual psychologies. In this section, I examine whether standard accounts can accommodate riot cases by way of Evaluative Content, Differentiation, and No Full Excuse.

First, however, it is helpful to clarify some notions that I have used liberally in the above and that figure widely in theories of attributability blame. What is it for an agent to have an evaluative outlook? For an action to have evaluative content? And for some evaluative content to manifest an outlook? These notions are rarely articulated in detail in the literature and could presumably be developed in different ways. I use them as follows. An evaluative outlook is a pattern of evaluative dispositions that reliably shapes what an agent sees as fitting or unfitting ways of acting or responding. The evaluative content of an action is the manifestation of an evaluative outlook. And an action manifests an evaluative outlook when it behaviorally indicates that some response is regarded as fitting or unfitting in light of that outlook. This is the case when a situational trigger causes some action and the outlook helps us explain why that trigger leads to that kind of evaluatively significant response rather than another.13

With this terminology in place, let’s turn to the main strands of attributability blame, namely, the deep self view and attributionism. Both deep self theorists and attributionists hold that attributability blame is appropriate only when an agent’s action expresses their evaluative outlook. In keeping with the above, both can be understood as identifying evaluative outlooks with a set of characteristic evaluative dispositions, but they differ on what sustains those dispositions. For the deep self theorist, the relevant dispositions are based on an agent’s “deep self,” typically understood in terms of second-order volitions,14 cares,15 desires,16 or approval.17 For the attributionist, by contrast, the relevant dispositions are based on the agent’s evaluative judgments, i.e., a set of internal states that the agent can be asked to defend, revise, or abandon. Because both views tie responsibility to evaluative outlooks rather than to voluntary control, they allow that agents may be appropriately blamed for actions like a spontaneous outburst as long as they manifest the relevant evaluative dispositions.

Both deep self theorists and attributionists can readily accommodate Diehard cases. This is not to say that the Diehard rioter is necessarily unaffected by the crowd, but only that the social dynamic does not alter the fact that their actions have evaluative content that expresses their own evaluative outlooks. The crowd may simply trigger the manifestation of the Diehard’s preexisting dispositions—much as a vase’s fragility is manifested when struck—or it may amplify the intensity with which the evaluative outlook is expressed—much as a fire burns brighter with more oxygen. But on both frameworks, the evaluative content of the Diehard’s behavior is typically explained by the agent’s own characteristic dispositions. Diehards are therefore straightforwardly blameworthy for the evaluative content of their actions.

Tagalong cases, by contrast, pose a clear challenge to both frameworks. For the deep self theorist, the fact that the evaluative content of the Tagalong’s actions does not align with their deep self-based evaluative outlooks places their behavior in the same category as out-of-character acts, which do not warrant attributability blame. Shoemaker, for example, discusses a caring and loving person who snaps and yells at her spouse after being involuntarily subjected to hallucinogens. She would be excused from attributability blame, he argues, “[b]ecause we view her outburst as having been genuinely out of character” and as not expressing anything about “who she really is.”18 Similarly, Arpaly and Schroeder note that if a person commits murder under the influence of a sinister cult, “her action would be seen as poorly integrated and she would not receive full blame,” and elsewhere they even allow that “an act which is extremely poorly-integrated may be one that, as one might say, is better attributed to circumstances, a drug, or a nefarious neurosurgeon than to the actor, and so be an act for which the actor is not responsible.”19 In both cases, behavior that expresses evaluative content is nonetheless treated as excused because it does not express the agent’s own outlook. It seems that deep self views must treat Tagalongs in the same way. This directly violates No Full Excuse, since we ordinarily would blame Tagalongs for the evaluative content of their actions and expect them to acknowledge their faults and express remorse.

Attributionism faces a related difficulty. An act may be out of character in the deep self sense (say, if it doesn’t reflect a second-order volition) but not in the attributionist sense (if it remains judgment-sensitive) and vice versa. Attributionists therefore hold people responsible for some actions that deep self theorists do not. But this difference doesn’t solve the problem. The Tagalong may still pass judgment in some sense, but their evaluative outlook, which structures what counts as a reason and how much weight a reason should carry in deliberation, remains socially scaffolded. The Tagalong’s judgment is made from within the outlook of the riot rather than their own, so to speak. This leaves the attributionist with a dilemma. If they say that the action doesn’t reflect the agent’s own judgments, because the evaluative outlook within which the judgment was formed was socially scaffolded, they effectively excuse the Tagalong, thereby contradicting No Full Excuse. If, on the other hand, they insist that blame is appropriate because the action reflects the Tagalong’s judgment in some looser sense, it seems that they are bound to attribute full responsibility to the Tagalong in the same way as the Diehard, thereby contradicting Differentiation.

There are two ways in which the deep self theorist or attributionist may try to respond, but they both fail.20 The first response tries to accommodate Differentiation by suggesting that the suggestibility of the Tagalong may itself constitute an evaluative outlook for which we can appropriately blame them. This would go some way toward satisfying Differentiation because, on this reading, the Diehard is blameworthy for, say, their anti-democratic views, whereas the Tagalong is blameworthy only for their suggestibility. This response does get something right, since we sometimes do blame people for being suggestible. The problem with this strategy is, however, that it fails to appreciate the content-sensitivity central to Evaluative Content. When we blame Tagalongs for what their actions express during a riot, we are not simply blaming them for being suggestible but for the outlooks they manifest. If suggestibility were the real target, the specific content of their riot behavior shouldn’t matter to us (or only matter as evidence of that more general suggestibility). Indeed, we would often blame people for being suggestible and for the evaluative content they manifest. These are distinct attributions. To illustrate, there is a clear difference between someone saying, ‘Look at the sheeple,’ and shouting, ‘Fascists!’21

A related but distinct response sidesteps the content-sensitivity problem by holding the evaluative content of the riot action fixed while arguing that Differentiation is a matter of different degrees of blameworthiness. Maybe the Diehard has a deep disregard for law enforcement, and the Tagalong displays a similar but much shallower disregard. This would allow us to say that the content of our attributability blame is similar across the two cases, but that the Diehard is blameworthy to a higher degree than the Tagalong.

I grant that attributability blame comes in degrees, yet I think this response faces a dilemma, because the idea of a shallower disregard is ambiguous between two readings. On the first reading, the Tagalong has a weaker but genuine first-order evaluative outlook directed at law enforcement. Suppose, for instance, that they occasionally make disparaging remarks about the police but normally respect their authority. On the second reading, the Tagalong has a second-order disposition toward acquiring such a first-order outlook under certain social conditions. But the first reading misidentifies the Tagalong, since it stipulates that the agent’s disregard for law enforcement is sustained by their own psychological structures, and in that case, the agent is a Diehard Lite and not a Tagalong at all. And the second reading collapses into the previous objection, since the evaluative content of the second-order disposition would be a general suggestibility, not the first-order outlook manifested in the riot.22

Alternatively, one might try to avoid these difficulties by adopting a more permissive account altogether according to which agents are responsible for the evaluative content of their actions regardless of whether that content manifests an evaluative outlook. But this strategy is equally unsatisfactory. It not only violates Differentiation by erasing the moral difference between Diehards and Tagalongs; it also throws the baby out with the bathwater by abandoning the core insight of attributability blame, namely, that responsibility for evaluative content depends on how that content manifests an evaluative outlook. The challenge, then, is to preserve what is compelling about attributability-based theories while rejecting the idea that only individual-based evaluative outlooks can serve as the basis of such blame. This is the task of the next section.

In this section, I develop the Outlook View as a more general account of attributability blame that avoids the limitations of deep self views and attributionism. The guiding thought is, roughly, that attributability blame tracks evaluative outlooks and their manifestation, not any of the specific structures that sustain them.

It is worth noting from the outset that the Outlook View is a friendly amendment to existing accounts, which already, implicitly or explicitly, emphasize the importance of evaluative outlooks. Sripada, for example, characterizes cares in terms of functional profiles that closely resemble what I call evaluative outlooks. In this sense, deep self theorists have always been implicitly concerned with dispositions that structure what an agent finds fitting but have focused on the deep self because they believe that this is what normally sustains evaluative outlooks.

The importance of outlooks becomes even clearer in cases where the deep self is preserved but no longer gives rise to a characteristic evaluative outlook. Suppose Rashid has a deep and long-standing contempt for a particular ethnic group. As a result, he does not support businesses owned by members of that group. Rashid is also meticulous and plans his errands carefully, writing in his diary which exact shops to patronize and which exact routes to take. One day, however, a neurosurgeon implants a push/pull inverter chip in Rashid that systematically reverses the valence of his evaluative responses toward other people. His contempt for the ethnic group is now reversed into sympathy and goodwill. Later, he follows the plan he prepared so carefully in his diary, without noticing the underlying pattern.

In this case, Rashid’s actions are causally dependent on the relevant deep self state (his contempt, which caused him to make the plan) and their content is in harmony with it (systematic avoidance).23 Yet the actions no longer reflect what Rashid currently takes to matter. His evaluative outlook, shaped by the chip, regards people from the ethnic group with sympathy rather than contempt. In such a case, attributability blame for the evaluative content expressed in Rashid’s behavior seems mistaken. Plausibly, a deep self theorist would agree with this verdict and reply that the deep self grounds moral responsibility only when it actually produces an evaluative outlook. This shows that deep self views are already committed to evaluative outlooks doing the real normative work.24

A similar point applies to attributionism. Although attributionists frame responsibility in terms of evaluative judgment, the most plausible interpretation of such judgments treats them not as isolated occurrent states but as dispositional patterns that guide what the agent takes to be reasons and how they regard different responses as fitting or unfitting. As Smith says, evaluative judgments are “not necessarily consciously held propositional beliefs, but rather tendencies to regard things as having evaluative significance.”25 In this respect, evaluative judgments are a specific kind of evaluative outlook. Characteristically, attributionists hold that the morally relevant outlooks are those that the agent can be asked to defend, revise, or abandon.

Recognizing this similarity suggests a more general approach. Deep self theorists and attributionists disagree about how evaluative outlooks must be sustained, not whether responsibility depends on the expression of such outlooks. The Outlook View makes this explicit:

The Outlook View An agent is an appropriate target of attributability blame for an action only if

(Manifestation) the evaluative content of an action manifests an evaluative outlook, and

(Functional Support) the basis on which that evaluative outlook is sustained supports the integrated functions of attributability blame.26

Manifestation states that attributability blame is appropriate only when an action’s evaluative content manifests an evaluative outlook. This follows directly from what has been established above and is shared by both deep self views and attributionism. Manifestation, however, departs from these accounts by allowing that evaluative outlooks may be sustained in different ways.

Functional Support introduces a further constraint specifying that not every evaluative outlook that is manifested is one for which attributability blame is appropriate. For blame to be appropriate, the sustaining basis of the outlook must be such that the central functions of attributability blame can be coherently exercised together. Functional Support is therefore not a claim about which bases suffice for blame but a constraint on what an adequate sustaining basis must (enable us to) do. Deep selves and evaluative judgments are familiar candidates for satisfying this condition. The Outlook View does not deny this. Whether other sustaining bases can also satisfy Functional Support is a further question that I will take up in the later sections.

I postpone a full discussion of what these functions are and how they are integrated until Section VI, but it is worth noting how Functional Support clarifies the familiar disagreement between deep self theorists and attributionists. Deep self theorists emphasize the appraisal dimension of attributability blame and therefore require that the relevant evaluative outlook reliably characterizes the agent. Attributionists, by contrast, emphasize address and alteration and therefore require that the basis of the outlook consist in attitudes that are judgment sensitive. On my reading, these differences concern not the metaphysics of agency as such, but which functions of attributability blame each tradition regards as fundamental.

One might worry that the Outlook View is circular. Are the functions of attributability blame in Functional Support not themselves defined in terms of when attributability blame is appropriate? But this is not the case, since the key functions are analyzed independently as features of our responsibility practices rather than as features of what makes blame appropriate. They describe what this form of blame does, not what makes it justified. In this way, Functional Support holds that the sustaining basis of an outlook (whatever it is) must be related to an agent (or set of agents) so that these functions can operate without undermining each other.

Taking stock, the Outlook View allows evaluative outlooks to be sustained in multiple ways. This includes the possibility, central to what follows, that an outlook may be sustained by social dynamics rather than an individual’s character. On my proposal, determining whether the evaluative content of an action warrants attributability blame requires asking two questions. First, does the action manifest an evaluative outlook? Second, is the way in which that outlook is sustained such that the characteristic functions of attributability blame can be appropriately applied?

Not all forms of social influence satisfy these conditions. Some (such as the coordination of movement during a marathon) fail to generate evaluative outlooks at all and thus fail Manifestation. Evaluative outlooks, recall, are not just behavioral patterns but dispositions that reliably shape what someone sees as fitting or unfitting ways of acting or responding. Others (such as hypnosis or panic flight) do give rise to such evaluative patterns but do so in a way that cannot sustain the relevant blame functions. Riot contexts, or so I shall argue, often allow us to answer both questions in the affirmative. Rioters typically express evaluative outlooks and the social scaffolding of those outlooks supports the relevant blame functions. This preserves Evaluative Content and No Full Excuse. At the same time, the way these conditions are met differs systematically between Diehards and Tagalongs, which yields the moral difference required by Differentiation.

Socially complex cases like riots put pressure on standard models of attributability blame by questioning the individualist assumption on which these rely. The Outlook View, in contrast, allows that evaluative outlooks can also be socially scaffolded. This aligns well with our ordinary talk of movements, organizations, and mobs as having “cultures” or “atmospheres” that shape what their members take to matter. I suggest that we take this talk seriously. In this section, I’ll defend the claim that social interactions can sustain evaluative outlooks, before I, in the next section, argue that they can support key blame functions.

A socially scaffolded evaluative outlook is a pattern of evaluative dispositions that reliably shapes what one or several agents see as fitting or unfitting ways of acting or responding, where these dispositions are generated and sustained by ongoing social interaction rather than by a particular individual’s psychological states alone.27 In other words, an evaluative outlook is socially scaffolded when the group context itself makes a difference to what the agent sees as the right way of responding to the world.28 This also shows that the claim is not that psychological states play no role in socially scaffolded evaluative outlooks. Indeed, for social dynamics to generate and sustain a first-order evaluative outlook, they must actually alter what the agent feels, perceives, and takes to matter; otherwise, there would be no sense in which the social dynamics sustained an evaluative outlook. What is distinctive about socially scaffolded outlooks is therefore not that they bypass the agent’s psychology altogether, but that the relevant psychological states are themselves produced and sustained by the social dynamics rather than by the agent’s own psychological structures.

This raises an immediate question. If psychological states ground the agent’s suggestibility, what prevents us from saying that the Tagalong simply has a preexisting disposition that is triggered by riot conditions, much as the Diehard’s preexisting evaluative outlook is sometimes triggered by certain social settings? On the model developed here (see also Section III), there is an important structural difference between the two cases. The Diehard has a preexisting first-order disposition (say, hostility toward police) that the riot (among other things) triggers. The Tagalong, by contrast, acquires a first-order disposition with similar content only when placed in the riot context.29 This means that even if they both act similarly (say, by shoving a police officer), their underlying dispositions may have different structures. Two points are worth emphasizing.

First, since the Diehard’s first-order disposition is not sustained by the social dynamics of the riot but by the Diehard’s psychological structures, we would expect it to manifest across a broader range of contexts (in conversations, online, at other protests), whereas the Tagalong’s first-order disposition is context-specific: it exists only when the relevant social dynamics are present. The Tagalong’s second-order suggestibility may itself manifest across a broader range of social contexts, but, as I argued in Section III, this suggestibility has a different evaluative content from the first-order outlook generated in any particular context. Second, and relatedly, since the Tagalong’s second-order suggestibility makes them vulnerable to a broad range of first-order evaluative outlooks, we would expect them to manifest different first-order dispositions depending on the relevant social dynamics. For instance, they might express solidarity in one crowd and hostility in another. The Diehard’s evaluative outlook, by contrast, will manifest more robustly across contexts because it is sustained by their own standing psychological structures rather than by the dynamics of any particular social setting.

We must also distinguish socially scaffolded evaluative outlooks from other changes in behavior brought about by social dynamics. Social settings often alter behavior in ways that do not generate an evaluative outlook at all. Just as an individual can be caused to act in ways that fail to manifest their personal evaluative outlook (say, when a muscle spasm causes me to knock over your drink), social influence can likewise produce behavior without shaping what the agent takes to be fitting or unfitting. For example, I may yawn after seeing you yawn. What is missing in such cases is not coordination or causal influence but evaluative uptake. The social influence must shape what agents experience as a fitting or appropriate response to the situation.

The same point applies to many forms of crowd behavior. Consider a marathon. Runners typically have preexisting evaluative dispositions. They care about their level of fitness, pacing, personal records, etc. They manifest these cares during a race, including in how they adjust their behavior in response to others. They match pace with someone, overtake another, and move aside for someone else. Their interactions are coordinated and causally interlinked, but this interaction does not sustain a socially scaffolded evaluative outlook. What matters to each runner, and how different responses are evaluated as fitting or unfitting, is fixed independently of the group interaction. This contrasts with riot contexts, where interaction itself structures what counts as an appropriate or called-for response at least for some rioters.

A natural worry about the Outlook View in general, and about socially scaffolded evaluative outlooks in particular, concerns robustness. Standard accounts implicitly assume that the psychological structures that they emphasize as sustaining bases are relatively stable, and that the evaluative outlooks they sustain persist over time. Socially scaffolded evaluative outlooks, by contrast, are often brittle and short-lived. Perhaps the robustness of our evaluative outlooks is somehow decisive for our responsibility practices?

In response, it is important to distinguish a metaphysical from an epistemological version of this worry. Metaphysically, neither deep self views nor attributionism typically require evaluative outlooks to be temporally robust.30 Deep self states or evaluative judgments could, in principle, flare up briefly, sustain an evaluative outlook for a moment, and then disappear. On standard accounts, an agent would still be responsible for actions manifesting such momentary outlooks. This reflects a general feature of dispositions: their identity is determined by their modal profile, not by their duration. If evaluative outlooks are correctly understood as evaluative dispositions, then temporal brittleness is not itself morally excusing. Indeed, riot cases can thus be understood as empirically well-described instances of momentary evaluative outlooks.

This brings us to the epistemological worry. If evaluative outlooks can be short-lived, how can we tell when an action genuinely manifests such an outlook rather than being an expressive misfire (as in Shoemaker’s example of a caring person who snaps after exposure to hallucinogens)? This is a serious worry, reflecting a difficulty in distinguishing dispositions from accidental surface behavior. For our purposes, the key is to examine whether the relevant patterns occur reliably when both the sustaining basis and the triggering conditions are present. The question is not whether the disposition is stable across time in general, but whether, relative to its sustaining basis, it manifests with the right kind of reliability under the relevant conditions. By analogy, we attribute to a light bulb the disposition to emit light even when it is not powered, precisely because it manifests that disposition reliably whenever the sustaining basis (the filament) and the trigger (electricity) are present. The epistemological point is then that in order to distinguish genuine manifestations from expressive misfires, we must look at how reliably the relevant behavior occurs when the sustaining basis can reasonably be assumed to be present.

This epistemic limitation has direct consequences for the Outlook View in terms of when attributability blame must be withheld:

The Epistemic Corollary Attributability blame should be withheld whenever Manifestation cannot be established with sufficient epistemic confidence.

Given this corollary, the Outlook View predicts that it will often be inappropriate for us to blame agents in cases where an action could, in principle, manifest a momentary evaluative outlook, simply because we lack sufficient evidence to distinguish a genuine manifestation from an expressive misfire. Presumably, there will also be cases of socially scaffolded evaluative outlooks for which this epistemic problem arises. Riot cases, however, are epistemologically privileged. Although the social dynamics that sustain riot-based evaluative outlooks are short-lived, they manifest reliably across participants. The coherence and convergence of riot behavior thus provide evidence for the presence of a socially scaffolded evaluative outlook.31

We can now see how riot behavior may satisfy Manifestation. Diehards manifest evaluative outlooks based on their characters, and the crowd merely triggers or intensifies the dispositions they already possess. Tagalongs, by contrast, manifest an evaluative outlook whose sustaining basis lies in the group dynamics themselves: the interaction alters what they take to be fitting or unfitting in the moment, and their joint behavior manifests this outlook reliably across participants, thereby meeting the evidential threshold the Epistemic Corollary requires. Riot actions therefore count as manifestations of evaluative outlooks for both Diehards and Tagalongs, although these outlooks are sustained in different ways.

In the previous section, I argued that riot crowds can satisfy Manifestation in the sense that riot actions can manifest evaluative outlooks even when those outlooks are not based in an individual agent’s own psychology but are instead socially scaffolded. This still leaves us with the Functional Support condition. Can socially scaffolded evaluative outlooks support the characteristic functions of attributability blame? The task of this section is to show that they can, though not in the same way as evaluative outlooks sustained by an agent’s character.

The point of Functional Support, recall, is to qualify which evaluative outlooks warrant attributability blame, since not all outlook manifestations are morally relevant. A panic flight, for example, could be seen as a manifestation of a momentary evaluative outlook, yet we normally do not blame people for panicking. Standard theories typically explain this by appealing to the outlook’s sustaining basis: panic is not based on an agent’s deep self or evaluative judgment but somehow bypasses them. There is something right about this diagnosis. According to the Outlook View, however, panic flight is excused not because deep selves or evaluative judgments are metaphysically privileged, but because the sustaining basis of the outlook fails to support the characteristic functions of blame. Formulated more generally, the Functional Support condition ensures that we can distinguish excused from non-excused outlook manifestations without assuming that blameworthy outlooks must be generated in one particular way.

This suggests that the dispute is not primarily about the metaphysical preconditions of moral responsibility, but about what our practices of attributability blame are for and how they function. This shift is especially important in socially scaffolded cases, where it is unclear whether blame remains intelligible when an evaluative outlook is sustained by social dynamics rather than by an enduring psychological structure. To make sense of blame in these cases, we therefore need a functional profile of attributability blame.

Existing theories differ in emphasis and detail, but there is a rough convergence around four core functions that attributability blame typically performs:32

These functions are analytically distinct but practically interconnected. For blame to be apt, the sustaining basis of the relevant outlook must be such that these functions can be jointly supported in a coherent way.

It is easy to see how these functions are supported in paradigmatic cases where an action expresses an agent’s deep self or evaluative judgments. Standard theories of attributability blame differ, however, in how they weight these functions and in which of them they take to be most central. This helps explain why they emphasize different sustaining bases. Views that place particular weight on the address function are naturally inclined to require some form of judgment sensitivity or responsiveness to reasons, since addressing an agent as a moral interlocutor presupposes that they are able to recognize their faults and change. By contrast, views that emphasize attribution and appraisal are more permissive about the sustaining basis of the relevant evaluative outlook and are correspondingly more willing to ground blame in features of the agent’s psychology that escape reflective reach. On this picture, disagreements about the metaphysical basis of blame can be traced back to disagreements about which of its functions are primary.

In what follows, I’ll assume that all four functions must be supported for blame to be appropriate. One could adopt a leaner interpretation of the Functional Support condition, yielding a more permissive account of attributability blame. I will argue, however, that even this maximalist account (requiring support for all four functions) can be satisfied in riot contexts.

The challenge in thinking about socially scaffolded evaluative outlooks is that they can sustain all four functions of blame even when these four functions do not straightforwardly converge on a single individual agent. The crucial step is to recognize that the different functions of blame need not have the exact same object, as long as they are functionally integrated. These functions are integrated when the sustaining basis of the evaluative outlook constrains how each blame function is appropriately realized, such that they can be intelligibly exercised as part of a single normative response. That is, the sustaining basis determines not only where attribution properly falls but also what forms of address and relationship alteration are appropriate. Functional integration is satisfied when the sustaining basis provides a unified explanation of how each function is realized and how their respective objects are related.

Stephanie Collins makes a related point by distinguishing the target and the audience of blame.37 In paradigmatic individual cases, these coincide. But, as Collins argues in the context of structural injustice, they can come apart without rendering blame unintelligible. In cases of structural injustice, our blame targets an enduring social structure (such as systemic racism), while its audience consists of the individuals who materially constitute and sustain that structure. Although she focuses on a different type of blame and its corresponding functions, the underlying lesson is that a social rather than psychological sustaining basis can support the functional profile of blame even when its functions are directed at distinct objects.

However, the Functional Support condition is not satisfied merely by assigning different objects to different functions of blame. The sustaining basis must constrain how these objects are related. In Collins’ cases, the target of blame and its audience are connected by relations of constitution and maintenance. This structured connection is what makes it intelligible to target the structure while addressing the individuals who uphold it. Absent such a connection, blame would fragment into disconnected responses rather than forming a unified normative practice.

With this functional framework in place, we can now ask how socially scaffolded evaluative outlooks such as those expressed in riot participation can sustain these functions. Consider:

Lina Lina joins a protest outside the parliament building to voice her disagreement with a newly passed law. At first, the atmosphere is tense but peaceful. People mainly hold signs and chant. As more protesters arrive, the mood begins to shift. Shouting grows louder, bodies pack tighter, and emotions run high. Police officers form a barrier at the top of the marble steps. Lina feels the crowd surge forward and moves with it. Around her, people begin to shove the officers. Without much deliberation, she follows suit. Months later, in court, her lawyer argues that Lina was swept up by the moment and that this behavior does not reflect who she is.

How should we assess Lina’s actions? In Section II, I suggested that our intuitions about cases like this take roughly the following form. First (Evaluative Content), our assessment does not turn solely on the harm caused by Lina’s action (the injury to the officer) but also on the evaluative content it expresses (disregard for authorities or democratic procedures). Second (No Full Excuse), even if Lina’s lawyer is right that this behavior does not reflect a characteristic of Lina’s character, this fact does not by itself excuse her from blame for the evaluative content her action expressed.38 Third (Differentiation), at the same time, it clearly matters whether the evaluative content manifested in Lina’s action reflects her own standing evaluative outlook or is sustained primarily by the surrounding social dynamics. This difference concerns not whether blame is appropriate, but which shape it must take.

The Outlook View captures these intuitions by holding that, even if Lina’s lawyer is correct, the functions of blame come apart without fragmenting. Rather than attributing the evaluative outlook to Lina in virtue of some element of her character, we attribute it to her in virtue of the social dynamic of the riot that her action helped sustain. In line with the findings from social psychology, the Outlook View holds that Lina’s shove manifested the outlook (or, metaphorically, the atmosphere) of the riot as a whole rather than her character.

This helps explain why courts in riot cases often attend closely to evidence bearing on whether a defendant’s conduct manifested a preexisting outlook or one sustained by temporary social dynamics. Judges and prosecutors routinely ask whether similar evaluative content is expressed outside the riot context. Thus, in sentencing Scott Kevin Fairlamb, who shoved and punched a police officer during the January 6 attack on the U.S. Capitol, U.S. District Judge Royce Lamberth emphasized that Fairlamb’s conduct was “out of character for his whole history and his philosophy for living his life.”39 By contrast, in the trial against Peter Schwartz, prosecutors argued that his violence was “not at all out of character,” but just “the most recent episode in a three-decade history of assault, violence, and weapons.”40

But even when people act “out of character,” this does not mean that we cannot appraise the evaluative content of their actions. Returning to Lina, we can still find her disregard for the officer and democratic procedures objectionable.

Were blame limited to attribution and appraisal, our response would appear oddly impersonal. But the Outlook View also makes sense of the address and alteration functions of blame. In analogy to Collins’ account of blame in cases of structural injustice, these functions appropriately target Lina as someone who helped realize and manifest the socially scaffolded evaluative outlook. The address function involves holding Lina to normative expectations. In light of the attribution and appraisal, this cannot mean that she must take ownership of the outlook as a characteristic of her psychology but that we expect her to recognize that she helped realize an objectionable socially scaffolded evaluative outlook, to disavow it, and reaffirm her commitments to some other evaluative outlook. This expectation concerns a public act of acknowledgment rather than a change in private attitudes, which is why it remains appropriate even if Lina, ex hypothesi, already privately regrets what she did. This too is often reflected in judicial practice, where defendants are expected to reflect on their actions, their evaluative content, and to disavow that content. In support of the Outlook View, this practice is not limited to cases where the evaluative outlook is sustained by a stable psychological structure.

Relatedly, the alteration function concerns how our moral relations to those who help realize socially scaffolded evaluative outlooks change. In Lina’s case, it may still be appropriate for us to adjust how we relate to her—for example, by treating her future participation in emotionally charged group contexts with caution or by expecting reassurances that she would not act similarly again. In this way, the address and alteration functions sustain a form of blame that is morally serious yet appropriately responsive to the social scaffolding of the outlook Lina helped enact.

Importantly, when applied to Lina, the different functions of blame remain integrated because the sustaining basis of the evaluative outlook constrains how each function is appropriately realized. It determines to whom the outlook is attributed, what form of address can reasonably be demanded, and what kind of relationship alteration is warranted. Because the sustaining basis of Lina’s action lies in the social dynamics rather than her character, attribution falls on Lina as a riot participant rather than on Lina as an individual, address cannot demand that she revise her own evaluative outlook but only that she disavow what she helped realize, and alteration is appropriately local and context-sensitive rather than global. Each function is realized differently than in a paradigmatic individual case, but they remain coherent as a unified normative response precisely because they all answer to the same explanation.

As noted above, not all evaluative outlooks that satisfy Manifestation will also satisfy Functional Support. Some fail because the sustaining basis does not allow the functions of blame to hang together as a unified normative response in the required case. In those cases, social influence excuses people from attributability blame. The point can be brought out by briefly considering some limiting cases, where the sustaining basis fails to provide the unified explanation that functional integration requires, causing the blame functions to fragment rather than cohere.

First, consider psychosis. Here an agent may act from an internally coherent evaluative outlook, yet the sustaining basis of that outlook undermines the attribution, address, and alteration functions of blame. Although the behavior may have evaluative content, there is no intelligible way to hold the agent to normative expectations or to revise our moral relations in light of it. The case thus fails to satisfy Functional Support.

Second, consider panic flight. Even if such cases involve a minimal sense of fittingness and thus plausibly satisfy Manifestation, they do not satisfy Functional Support. Why not? As in the psychosis case, there is no intelligible way to hold the agent to normative expectations or to revise our moral relations in light of those manifestations. There are two reasons for this. First, appraisal breaks down. Even if the behavior of someone fleeing in panic carries evaluative content in a minimal sense, that content does not present itself as something that calls for concern, endorsement, or repudiation by others. It therefore lacks the kind of moral salience that appraisal in attributability blame is meant to register. Second, address and alteration are unavailable. Because the outlook expressed in panic flight is sustained by an automatic survival mechanism, it bypasses the agent’s capacities for normative uptake and self-regulation entirely. As a result, there is no coherent way to demand acknowledgment, disavowal, or realignment, nor to justify relationship-altering responses. The functions of blame therefore fail to hang together as a unified normative response.

Finally, consider a stampede. This is superficially similar to riot cases in the sense that behavior is coordinated and socially influenced, and that social forces bypass agents’ personal evaluative outlooks. In reality, however, a stampede is best understood not as a distinct phenomenon but as a collective and socially amplified form of panic flight. The fact that a stampede occurs collectively does not change the nature of the sustaining basis: the behavior remains driven by automatic survival mechanisms rather than by a socially sustained evaluative outlook. It thus fails for the same reasons as cases of panic flight.

In riot cases, by contrast, the social dynamics reorganize people’s evaluative outlooks in a way that preserves the functional integration of blame, rather than causing its constituent functions to come apart or undermine one another.

This paper has defended what I have called the Outlook View of attributability blame. According to this view, attributability blame is not tied to any particular psychological structure. Instead, what matters is, first, whether an action manifests an evaluative outlook (Manifestation) and, second, whether the sustaining basis of that outlook supports the characteristic functions of attributability blame (Functional Support).

Using riots as an epistemologically privileged test case, I have argued that existing theories of attributability blame struggle to accommodate the morally salient differences between cases in which an action manifests an agent’s own evaluative outlook and cases in which it manifests an evaluative outlook sustained by social dynamics. Standard deep self and attributionist accounts tend either to over-excuse socially scaffolded cases or to simply treat them as normal individual cases.

The Outlook View avoids this dilemma by allowing that evaluative outlooks can be sustained in different ways while insisting that blameworthiness depends on whether the relevant outlook supports the functions of blame. In riot cases, socially scaffolded evaluative outlooks can satisfy Manifestation and Functional Support, but they do so in a distinctive way. Attribution and appraisal may properly target an outlook sustained by group dynamics, while address and alteration are directed at individual participants in light of their contribution to realizing and manifesting that outlook. This explains why socially influenced rioters are not exempt from attributability blame while still accounting for the moral differences we ordinarily recognize between different forms of participation.

Although riots have served as the paper’s focal point, the Outlook View naturally extends to other cases. This includes online mob dynamics and instances of peer pressure, both of which may temporarily change what participants take to matter, but there are also more long-lived cases, where organizational cultures, political climates, or broader dynamics of social cohesion act as scaffolds for certain evaluative outlooks. A paradigm case of longer-term social scaffolding is the ordinary citizen under Nazi Germany, where the political climate acted as a powerful scaffold for antisemitic evaluative outlooks.

A noteworthy complication of extending the analysis to long-term social scaffolding is that these cases sometimes involve internalization, in the sense that a socially scaffolded evaluative outlook gradually becomes integrated into an agent’s psychology, generating a hybrid case that is neither a pure Tagalong nor a pure Diehard. Many agents will be suggestible to and go along with, say, a sexist work culture or a hostile political climate, and over time internalize it. Indeed, we might have an agent who, at first, has an evaluative outlook sustained by social scaffolding alone, who later internalizes it so that it is sustained by both social scaffolding and their own psychology, and who finally outlives the social scaffolding with an evaluative outlook now sustained entirely by their own psychology.

The Outlook View predicts that the shape of attributability blame shifts across these stages. In the early stage, attribution falls on the individual as a participant in a social dynamic and address targets the individual’s contribution to realizing that socially scaffolded evaluative outlook. In the fully internalized stage, attribution falls on the agent’s own psychology and address can demand genuine revision of a standing evaluative outlook. In the hybrid case, where both sustaining bases are operative simultaneously, both shapes of blame are appropriate such that attribution can fall on both the individual qua member of the social dynamic and the individual qua bearer of certain psychological structures, address can demand both public disavowal and genuine self-revision, and alteration can be both local and global.

In this way, the Outlook View illuminates both when attributability blame is appropriate and what shape it should take, regardless of what sustains our evaluative outlooks.