2. Appropriateness of emotion

In response to a situation, certain emotions may be appropriate to feel, while others are not. What constitutes the appropriateness of an emotion? Following D’Arms and Jacobson (Reference D’Arms and Jacobson2000), we may distinguish two grounds for emotional appropriateness. First, it concerns whether the emotional response is fitting. To ascertain this, we ask whether the emotion accurately represents its target (i.e. its particular object) in light of evaluative judgements. Secondly, it pertains to whether experiencing that emotion in the given situation is morally acceptable. I construe the appropriateness of emotion strictly in terms of fittingness. To this end, I argue that formal objects establish the evaluative framework that determines both the kind of emotion experienced and its appropriateness relative to its particular object.

The particular object of an emotion is the target, that is, the specific entity, event, or state of affairs at which the emotion is directed (de Sousa Reference de Sousa1987: 16; Deonna and Teroni Reference Deonna and Teroni2012: 5–6). In a way, particular objects allow us to differentiate between distinct instances of the same type of emotion (Teroni Reference Teroni2007: 396). For example, in Draisine, Avis may feel sadness about the child’s death and sadness about losing her job due to the accident. She would recognise these as distinct instances of sadness owing to their different particular objects. Despite this difference, the targets at which Avis directs her emotions share a commonality rendering them particular objects of sadness. This is because, for Avis, both constitute a loss of something valued, which serves as the formal object of sadness (Prinz Reference Prinz2004: 62–63). As such, there exists a specific formal object available in both cases. But what exactly is the formal object of an emotion?

The formal object of an emotion is a specific second-order, evaluative property (e.g. a value) ascribable to the targeted particular object (Kenny Reference Kenny1963: 132–135; Prinz Reference Prinz2004: 62; Teroni Reference Teroni2007; Deonna & Teroni Reference Deonna and Teroni2012: 41 & 76). By being ascribed to the particular object, it determines, and thus defines, each emotion type to be experienced (de Sousa Reference de Sousa1987: 126; Rossi and Tappolet Reference Rossi and Tappolet2019: 549). Taking the formal object as such, I assert that when it is correctly ascribed to a particular object, the related emotion fittingly represents that object, thereby becoming appropriate.

Teroni presents a counter-argument to the thesis that formal objects uniquely individuate emotion kinds (Reference Teroni2007: 401–402), positing that ‘danger’ could be the formal object of both fear and excitement. However, although ‘danger’ might appear common to both emotions, their true formal objects could be closely related yet distinct, such as ‘danger likely to cause harm’ for fear and ‘danger unlikely to cause harm’ for excitement. Hence, a formal object should be as specific as possible so that it can uniquely individuate emotion kinds (Salmela Reference Salmela and Tissari2008: 29). For example, if the loss of something valued is the formal object of sadness (Prinz Reference Prinz2004), that formal object can belong only to sadness. When such a loss occurs, the individual experiencing the loss is expected to feel sadness directed at that loss. If this is true, by generalising the case, I argue that a specific emotion is appropriate to direct at a particular object when the emotion’s corresponding formal object genuinely exists or is instantiated by that particular object.

With these considerations in mind, how may we approach faultless group harms such as Draisine? What emotion is then appropriately at play for Avis? I propose to identify this emotion through its formal object, but doing so first necessitates examining the specific dynamics of faultless group harms.

3. Nature of faultless group harm

To identify the formal object of the emotion appropriate for faultless group harm, I begin by examining its distinctive features.

In cases of faultless group harm, the harm arises from multiple voluntary individual actions, and the contributor recognises at least one other agent as having also contributed to the harmful outcome. I call this the Multiple Feature (MULTIPLE). MULTIPLE requires, from the contributor’s perspective, that she sees herself as one among multiple contributors to the harm. It does not presuppose any particular relationship between the contributors. They need not be coordinated, need not share an intention, and need not even be aware of one another prior to the harm.

This formulation is deliberately broad, meaning that MULTIPLE is satisfied regardless of how the contributors’ actions relate to one another. The contributors may be engaged in shared intentional activity (Bratman Reference Bratman2022), bound by a joint commitment (Gilbert Reference Gilbert2000), acting in the ‘we-mode’ (Tuomela Reference Tuomela2007), acting within a corporate structure (French Reference French1984; List & Pettit Reference List and Pettit2011), forming part of the activities of random collections (Held Reference Held1970; Hindriks Reference Hindriks2019), or participating in unintentional aggregate action (Chant Reference Chant2007). My account does not have to adjudicate between these frameworks, for MULTIPLE is a perspectival condition rather than a metaphysical one.

On a sufficiently broad construal of group action, such as that offered by Schmid (Reference Schmid2023), even the lorry driver and the child might be said to act together in some sense, since their actions converge to produce the outcome. However, the lorry driver scenario differs from cases akin to Draisine in a structurally significant respect. In the lorry driver case, the child is the victim of the harm. The expressive dimension of agent-regret (its characteristic orientation towards recompense and restitution) is directed from the driver towards the child or the child’s family (Williams Reference Williams1982: 28–29). The child occupies the position of the harmed party, not that of a fellow contributor. By contrast, in Draisine, Ernest is not the victim. He is a co-contributor to harm that befalls a third party, namely the child. MULTIPLE conceptually isolates this distinction between a dyadic structure (agent and victim) and a triadic structure, which refers to evaluative positions rather than sheer numbers. These positions comprise the contributor herself, the fellow contributors (whether one or millions), and the victim or victims.Footnote ¹

Secondly, neither Avis nor Ernest intends to run over the child, nor do they form such an intention together. Moreover, it is unforeseeable that the child will run onto the railway. This yields the Faultlessness Feature (NO-FAULT), which concerns the unintentional and unforeseeable character of the incident from the contributor’s own perspective.

NO-FAULT differentiates faultless group harm from other group actions, whether harmless (e.g. jointly moving a piano) or deliberately harmful (e.g. two assassins intentionally murdering a victim). But it is important to note that NO-FAULT attaches to contributors individually. That is to say, even if some other contributors to the same group action intended or foresaw the harm, the harm remains faultless for those who did not. In such cases, a contributor who neither intended nor reasonably foresaw the harm still confronts it as a faultless group harm from her own standpoint.

The remaining features concern contributors themselves. The Contribution Feature (CONTRIBUTION) specifically concerns the contributor’s recognition of her own causal relevance to the harmful outcome. While the child’s death constitutes a faultless harm, Avis’s and Ernest’s individual actions are both intentional and voluntary in that each operates the draisine deliberately, albeit without any intention to harm. CONTRIBUTION requires that the contributor recognises her own voluntary action as having helped bring about the harmful outcome.Footnote ² This recognition is not restricted to cases where the contributor’s action is individually necessary for the outcome. The contributor may recognise her action as having contributed to it in some way, broadly construed. In particular, she must recognise her action as forming part of the actual causal process that generated the outcome, even if it was not a counterfactual difference-maker. What is essential is that she does not regard herself as causally disconnected from the outcome. Without this recognition, the contributor would have no ground for seeing herself as anything other than a spectator, and the fitting response would simply be spectator-regret.

Yet, while the contributor must recognise her causal involvement, the structure of this involvement differs fundamentally from that of an individual agent. For instance, the lorry driver’s individual action stands in a direct causal relation to the child’s death. Nevertheless, in Draisine, the causal path from any individual contributor’s action to the harmful outcome intersects with the contributions of others, meaning that the harm is fully accounted for by the group action. Therefore, Avis’s causal involvement is not self-standing like that of the lorry driver. Instead, it pertains to the causal path that includes Ernest’s contribution.Footnote ³

Finally, the Biperspective Feature (BIPERSPECTIVE) builds on MULTIPLE and CONTRIBUTION taken together. Because the contributor recognises her own causal role (CONTRIBUTION) alongside that of at least one other agent (MULTIPLE), she occupies a dual evaluative standpoint. She sees the harm both as something she helped bring about, which places her in what I call a ‘quasi-agential’ position, and as something others helped bring about, placing her in what I call a ‘quasi-spectatorial’ position. These are not alternating perspectives between which she shifts depending on what she recalls. As I construe them, they are simultaneous aspects of a single evaluative standpoint that arises from her recognition of the triadic structure in which she finds herself: she is one among several contributors to harm suffered by a further party. I term this dual standpoint ‘biperspective’.

None of these features requires that each contributor’s action be individually necessary for the harmful outcome. In Draisine, each operator’s contribution indeed happens to be necessary for propelling the vehicle to the speed at which the collision becomes unavoidable. But my account is not restricted to such cases.

This is precisely where Methane matters. Griffith recognises himself as one among millions of agents whose aggregated emissions cross the atmospheric threshold (MULTIPLE), thereby standing in a triadic structure alongside innumerable fellow contributors and the distinct victims of accelerated climate change. He drives deliberately yet neither intends nor can reasonably foresee that aggregated emissions will trigger rapid permafrost thaw at that specific juncture (NO-FAULT). Although his individual commute is neither necessary nor sufficient for crossing the threshold, he does not regard himself as causally disconnected from the outcome. Rather, he recognises his action as one among the innumerable actions whose emissions together carried the atmosphere past that point (CONTRIBUTION). Recognising both his own causal involvement and that of innumerable others, he occupies the same dual evaluative standpoint as Avis (BIPERSPECTIVE). With all four features obtaining in Methane no less than in Draisine, Griffith shares Avis’s structural position. Therefore, given that this position, rather than the scale of the group or the degree of coordination among its members, determines the fitting emotion, whatever emotion is appropriate for Avis must be appropriate for Griffith.

How do these features determine the emotion appropriate to faultless group harm? By drawing on them, we can identify the relevant formal object as ‘comparative disvalue of group harm to which one faultlessly contributed’. I shall talk about related formal objects in Section 4 and defend this formulation in Section 5, but for present purposes, let us take it for granted.

Even without going into details, we can see how this formulation captures each of the features effectively. Yet is there an emotion individuated by this formal object? If so, we must examine what kind of emotion it is. Given the analogous character of Williams’s lorry driver case and Draisine, I propose that the emotion individuated by the present formal object is a species of regret commonly discussed in relation to faultless harms (Williams Reference Williams1982; Bagnoli Reference Bagnoli2000; Sussman Reference Sussman2018; Wojtowicz Reference Wojtowicz2022b). Bernard Williams’s own remarks make this especially clear:

Would someone involved in a faultless group harm think that it would have been better if things had been otherwise? That seems plausible. Yet McQueen contests the reasonableness of regretting harms for which one bears no fault (2024: 43–44). Indeed, he rejects the very notion of agent-regret directed at unintended and unforeseeable harms and considers it irrational to expect the lorry driver to regret his action (2024: 100–101). Nevertheless, I assert that Williams’s constitutive thought of regret can still be reasonably ascribed to the lorry driver.

Notwithstanding McQueen’s reservations, it is worth asking how the driver himself would reflect on the incident. He would surely wonder what might have happened had he not driven that day and would judge that matters would have been better had he refrained. Such reflection, in which he recognises that his presence precluded a preferable outcome, naturally gives rise to a form of emotional distress. Williams identifies this distress as a species of regret, and there seems nothing unreasonable in doing so (Bagnoli Reference Bagnoli2000). Therefore, because faultless group harms can be viewed analogously, regret likewise seems the fitting response. However, the distinctive features of faultless group harms compel us to refine our conception of regret in this context.

4. Regret and faultless group harm

Recall that Williams may have referred to group actions when he asserted that one experiences agent-regret not only for one’s own past actions but also for actions in which one had participated (1982: 27). This claim is somewhat ambiguous, as Williams leaves unclear whether he associates agent-regret specifically with group wrongdoing (foreseeable, non-accidental, or intended) or also with faultless group harm. To my knowledge, no one has yet addressed this ambiguity.

I shall not attempt to resolve Williams’s ambiguity. Instead, I examine the phenomenon of faultless group harm through his unambiguous claims. He maintains that a form of regret, namely agent-regret, is fitting when an individual causes faultless harm. I accept this claim and similarly contend that faultless group harm calls for some form of regret. But if Williams intends to extend the appropriateness of agent-regret to cases of faultless group harm,Footnote ⁴ I must disagree. In my view, agent-regret is ill-suited to such cases, and so is spectator-regret. Even so, these two emotions still warrant scrutiny in this investigation.

Four key differences separate agent-regret from spectator-regret (Williams Reference Williams1982: 27–31). Recognising these distinctions can clarify how regret functions in cases of faultless group harm. First, and most importantly, spectator-regret may concern any event or action, while agent-regret relates specifically to one’s own past actions (Baron Reference Baron1988: 261; de Wijze Reference de Wijze2005: 459–463). In addition to this core difference, a second distinction emerges in the nature of the associated wishes. While spectator-regret is accompanied by a wish that events had unfolded differently, agent-regret involves wishing that one had acted differently (de Wijze Reference de Wijze2005: 459–460; Wojtowicz Reference Wojtowicz2022a: 72). Third, spectator-regret can arise in response to occurrences wholly outside one’s control; by contrast, agent-regret is restricted to past actions within one’s causal control (Wojtowicz Reference Wojtowicz2022b: 502–503).

Causal control over one’s action must be distinguished from causal control over the outcome of that action. The lorry driver exercises control over his own behaviour in that he drives the lorry deliberately. However, he does not exercise control over the fatal outcome, since the child’s appearance on the road is unforeseeable. Agent-regret attaches to the action, which lies within the driver’s control, notwithstanding the fact that the harmful outcome does not. Thus, one may be causally responsible for an outcome one could not control, and, coupled with the lack of fault, this direct causal responsibility renders agent-regret fitting in Williams’s account. The lorry driver’s causal responsibility for the child’s death is direct in that it stretches from his own action to the outcome, which a spectator lacks.

The fourth and last distinction is that spectator-regret may be expressed simply by feeling sorry about the incident, but agent-regret is expressed through acts aimed at recompense or restitution. This expressive dimension of agent-regret is directed specifically at the harmed party. For instance, in the lorry driver case, the driver’s regret orientates towards the child’s family, as it is their loss that demands acknowledgement and, where possible, redress. In Williams’s framework, this dynamic is straightforward because the structure is entirely dyadic in the sense that the agent causes the harm, the victim suffers it, and agent-regret flows from the former to the latter. As I shall argue in what follows, this dyadic structure breaks down in cases of faultless group harm. Where a fellow contributor is present alongside the victim, the contributor’s position can no longer be subsumed under the agent’s role in Williams’s formulation. Hence, a different emotional response is called for.

Taking the differences between agent-regret and spectator-regret into account and considering the defining features of faultless group harm, I argue that neither agent-regret nor spectator-regret is fitting for such harm. Yet, before outlining my position, I should emphasise a distinction.

In a case of faultless group harm, a contributor confronts not one but several potential particular objects of regret. For instance, in Draisine, Avis may direct her attention to what she herself did, to what Ernest did, or to the group harm as such, namely the child’s death resulting from what they did together. Directed at her own individual action, agent-regret may in principle be warranted, since it concerns her own past action. By the same token, directed at Ernest’s individual action, spectator-regret may in principle be warranted, since it concerns another’s past action. I do not deny this.Footnote ⁵ However, the group harm itself constitutes a further particular object that is not reducible to either individual action considered in isolation.

I contend that when directed at this particular object, neither agent-regret nor spectator-regret is fitting, since, as I shall show, the formal object of neither emotion properly applies. Furthermore, because Draisine is triadic rather than dyadic, Avis’s relation to Ernest’s contribution cannot be accounted for by agent-regret, which concerns one’s own past action, or by spectator-regret, which presupposes causal disconnection from the harm.

If neither agent-regret nor spectator-regret alone suffices for the group harm, one might argue that Avis and Ernest experience both at once, directed at the group harm as such. Owing to Ernest’s contribution, Avis realises that the harm was not produced solely by her own action and therefore feels spectator-regret directed at the group harm. At the same time, she is aware of her own causal role and on that basis experiences agent-regret directed at the group harm. Hence, one might suggest that the most fitting response for a contributor is a simultaneous combination of agent-regret and spectator-regret directed at the group harm.

I find this simultaneous combination similarly inappropriate, and the inappropriateness once again turns on the formal objects of the emotions. Although I do not contest the possibility of experiencing multiple emotions simultaneously when they are directed at different particular objects, I contend that where two emotions are directed at the same particular object and their formal objects conflict, they cannot both be fitting. Agent-regret and spectator-regret directed at the group harm constitute precisely such a case. To show this, I shall clarify the formal objects in question.

Matheson proposes that the formal object of agent-regret is simply ‘a doing’ (Matheson Reference Matheson2017: 672). In his own example, where he regrets purchasing a seat on the earliest flight from the furthest airport, he treats himself as the particular object of regret and his decision to buy that seat as the formal object (Matheson Reference Matheson2017: 666–667). This reading departs from the standard account of emotions. First, the target of regret is what one regrets, namely purchasing the seat, and therefore constitutes the particular object. Secondly, the formal object is not the ‘purchasing’ itself but the property ascribable to this object that renders it regrettable. On these grounds, Matheson’s view is non-standard.

A more standard interpretation is offered by Brown, who treats the formal object of regret as the ‘loss of something good’ (Brown Reference Brown2006: 12). Although her discussion of formal objects is conceptually attractive, this formulation does not fit regret well. The same description suits sadness (Prinz Reference Prinz2004), and it fails in cases where there is no loss or where what is regretted was never good.

No author has yet offered a precise formulation of the formal objects of regret, agent-regret, and spectator-regret. I undertake to supply one. Based on Williams’s analysis of the constitutive thought of regret (Williams Reference Williams1982: 27), I hold that, for regret in general, the relevant property is that of being a state of affairs that is worse than a genuinely possible alternative. Thus, I propose that the formal object of regret is ‘comparative disvalue of a past circumstance’. The property is comparative because we regret a circumstance only by measuring it against a superior counterfactual. Even when no better alternative actually obtained, a better counterfactual remains available, namely the circumstance’s non-occurrence. Finally, I speak of a ‘past circumstance’ because what may be regretted can be any item located in the past such as a state of affairs, an action, a situation, an event, or the like.Footnote ⁶

One might wonder whether the ‘comparative disvalue’ clause in my formulation is superfluous if harm is itself analysed in counterfactual comparative terms, as is common in the literature on the metaphysics of harm (see Feinberg Reference Feinberg1984; Hanser Reference Hanser2008). I am open to a counterfactual analysis of harm but do not commit to one. However, even on a counterfactual analysis, the ‘comparative disvalue’ clause performs a role that ‘harm’ alone does not, for harm concerns whether the victim is worse off, not whether the circumstance itself is worse than its alternative. What renders a circumstance the fitting object of regret is not merely that harm occurred, but that the circumstance is evaluatively worse than a genuinely possible alternative. ‘Comparative disvalue’ corresponds to this evaluative property. A faultlessly harmful action may, for instance, unforeseeably spare others a greater injury: its victim is worse off, yet the circumstance is not worse than its alternative. ‘Harm’ characterises what happened to the victim, while ‘comparative disvalue’ characterises the property in virtue of which the emotional response is rendered fitting, and the latter is not made redundant by the former.Footnote ⁷

Because agent-regret and spectator-regret are species of regret, their formal objects must accord with the formal object of regret in general while preserving the distinctions noted earlier. They cannot depart from the comparative disvalue of a past circumstance but must refine that description. Therefore, I propose that the formal object of agent-regret is ‘comparative disvalue of one’s own faultlessly harmful past action’, and that of spectator-regret ‘comparative disvalue of anyone else’s faultlessly harmful past action’. I think each formulation can reflect its respective emotion and be accurately ascribed to the particular object of the respective emotion.

For both, the adverb ‘faultlessly’ is indispensable, since, without it, the formulations might entail other emotions, such as remorse or indignation, respectively. Nonetheless, the difference lies chiefly in whose action is in question. While the former concerns the agent’s own deed, the latter concerns only those of others. Hence, the formal objects of agent-regret and spectator-regret conflict because an action is either one’s own or another’s. Therefore, when directed at the same particular object, the two emotions are mutually exclusive in terms of appropriateness in that the object either instantiates the formal object of agent-regret or that of spectator-regret, but not both. However, when directed at different particular objects, I contend that no conflict arises.

Returning to the group harm, for the formal object of agent-regret to be instantiated, the particular object must be one’s own action, and that action must be faultlessly harmful. Yet the group harm, as I have characterised it in Draisine, is not Avis’s own action. It is the outcome of what both Avis and Ernest did. If so, the group harm does not instantiate the formal object of agent-regret.Footnote ⁸ Likewise, since the group harm is not purely another’s action, but rather one to which Avis herself contributed, it does not instantiate the formal object of spectator-regret either. In this regard, neither emotion, when directed at the group harm, meets the appropriateness condition established before.

Let me be precise about the scope of this claim. I do not deny that Avis can feel agent-regret about her own individual action and spectator-regret about Ernest’s individual action simultaneously, so long as each targets a different particular object. I shall argue in Section 5 that these emotions can coexist with the appropriate emotion for the group harm. What is unfitting is directing agent-regret or spectator-regret at the group harm as such, for the formal objects of those emotions do not apply to it.

One might object that even if agent-regret and spectator-regret are unfitting when directed at the group harm simultaneously, a contributor could direct them at the group harm at different times, alternating between them. At one moment, she adopts the agent’s perspective and feels agent-regret directed at the group harm, and at another, she adopts the spectator’s perspective and feels spectator-regret directed at the same object. However, I find this alternation problematic but do not rule it out entirely.

The group harm is a fixed past event whose features do not alter between the moments of alternation. If so, all that changes is the perspective from which the contributor views the same unaltered object. Yet to direct agent-regret at the group harm, she would need to regard it as though it were her own action, and to direct spectator-regret at it, she would need to regard it as though it were purely another’s. Nonetheless, neither characterisation is accurate, since the group harm is neither wholly hers nor wholly another’s. To alternate between treating it as one and then the other would accordingly involve a form of perspectival ambivalence that misrepresents what she knows about her own situation. I do not contend that such ambivalence is psychologically impossible (see Razinsky, Reference Razinsky2016). Rather, I contend that it would not constitute a fitting response, since each perspective, when applied to the group harm, rests on an inaccurate evaluation of the particular object in question. This calls for a single emotion whose evaluative standpoint matches what the contributor actually knows about her situation: that she is one among several contributors to harm suffered by a further party.

The foregoing analysis reveals that neither agent-regret nor spectator-regret, nor any simultaneous combination of the two, affords a fitting emotional response to faultless group harms. Nevertheless, the fact that each emotion aligns with NO-FAULT is encouraging. In light of this, the discussion points towards the need for a distinct variety of regret. I now turn to develop this alternative.

5. Biperspectival regret

Although agent-regret and spectator-regret appear mutually exclusive, it is not logically impossible for Avis and Ernest to experience a single distinct emotion that draws on elements of both. Occupying a biperspective, each contributor to a faultless group harm becomes a quasi-agent and a quasi-spectator simultaneously. If so, we should expect the contributor to experience an emotion situated between agent-regret and spectator-regret, partaking of each without being reducible to either.

At this point, it is helpful to revisit the emotion’s formal object. As stated in Section 3, I take this to be ‘comparative disvalue of group harm to which one faultlessly contributed’. The argument developed thus far already supports this formulation, yet, for any reader who remains unconvinced, I shall substantiate it further in what follows.

I have examined the formal objects of agent-regret and spectator-regret by modelling them on the formal object of regret in general, as each is a subspecies of general regret. The same approach must guide the emotion appropriate to faultless group harm. Thus, its formal object must accord with that general pattern. Yet further constraints apply here. Because the fitting response lies, as I argue, between agent-regret and spectator-regret, its formal object must also be compatible with both. The simplest way to articulate such a second-order property is to emphasise the nature of the harm while acknowledging that the agent merely contributed to it, rather than caused it alone. In this manner, the formulation ‘comparative disvalue of group harm to which one faultlessly contributed’ seems fitting.

Instead of simultaneously experiencing two discrete states (agent-regret and spectator-regret) as though they were merely combined, Avis and Ernest may experience a single synthesised regret that performs the function of both. But what does ‘synthesis’ amount to here? That the distinction between combination and synthesis is not merely terminological becomes apparent when we consider the constitutive thoughts of the emotions in question.

The constitutive thought of agent-regret is ‘I wish I had not done that’, whereas for spectator-regret, it is ‘I wish that had not happened to the child’. Yet Avis’s thought is of an altogether different character. She thinks ‘I contributed to the accident, as did Ernest, and I wish it had not occurred’. This thought is irreducibly singular, for to decompose it into an agential and a spectatorial component would be to lose the very recognition that distinguishes the contributor’s situation, namely that the harm arose from what they did together. A mere combination of agent-regret and spectator-regret would fail to articulate this recognition, since it would preserve both constitutive thoughts intact, side by side. By contrast, synthesis yields a single emotion whose constitutive thought reflects what neither component thought can express on its own.

This irreducible thought reveals the ‘as-if’ positions of quasi-agent and quasi-spectator merging into a structurally distinct evaluative stance. In this sense, I term the emotion that synthesises elements of agent-regret and spectator-regret ‘biperspectival regret’. This emotion is unique to this unified stance, taking as its formal object the ‘comparative disvalue of group harm to which one faultlessly contributed’.

Following the argument set out in Section 4, I assert that the coexistence of agent-regret, spectator-regret, and biperspectival regret does not bring about any evaluative conflict so long as each emotion is directed at a distinct particular object. Avis can concurrently feel agent-regret directed at her own individual action and spectator-regret directed at Ernest’s individual action. And, alongside these emotions, because biperspectival regret is defined by its own distinct formal object, she can experience it directed at the group harm itself. What biperspectival regret distinctively affords is a fitting response to this group harm considered as a whole. Within Williams’s framework, no other emotion is sufficient to fulfil this role, since no other formal object applies to such an outcome. Therefore, rather than competing with or supplanting agent-regret and spectator-regret, biperspectival regret serves as an essential complement to Williams’s taxonomy.

Biperspectival regret aligns with all the features of faultless group harm. As a form of regret, it is consistent with NO-FAULT. Because it embodies characteristics of both agent-regret and spectator-regret, it is fitting for harms exhibiting CONTRIBUTION, BIPERSPECTIVE, and MULTIPLE. Hence, whenever its formal object arises, biperspectival regret is the fitting emotion for a contributor.

However, one might justifiably ask whether biperspectival regret is a distinct emotion, or whether it is genuinely a species of regret. Although McQueen argues against regret as warranted for faultless harms, he nevertheless provides a substantive analysis of the emotion. Interestingly, despite this apparent opposition, the features he identifies as necessary for a plausible account of regret are indeed accommodated by biperspectival regret (McQueen Reference McQueen2024: 12). These features help to clarify what regret is about, how it feels, the thoughts, judgements, and beliefs it involves, and its influence on future behaviours (2024: 13). I shall now examine these features to determine whether biperspectival regret likewise possesses them.

The first feature is that regret is intentional (e.g. Bagnoli Reference Bagnoli2000; Eldridge Reference Eldridge2017). That is, unlike undirected moods, regret always has an object. Biperspectival regret uniquely targets faultless harm resulting specifically from a group action to which the agent contributed. While general regret is fitting for unfortunate incidents, biperspectival regret is specifically warranted when an incident satisfies MULTIPLE, thereby highlighting the collective dimension of the object.Footnote ⁹

Secondly, regret involves both thoughts and feelings. To regret something entails forming a belief that something regrettable occurred, thereby eliciting certain feelings (Solomon Reference Solomon2007). McQueen suggests that regret involves pain due to one’s actions (2024: 12). However, in my view, pain is more apt for remorse or guilt. Regret is comparatively milder. Thus, I do not find that ‘pain’ adequately reflects the experience of regret, which instead involves a form of moral distress that might not necessarily be painful but rather uncomfortable or unsettling. In the case of biperspectival regret, these feelings would be experienced to a lesser extent compared with regret for harm caused individually. Even so, biperspectival regret involves feelings of discomfort or disquiet.

Thirdly, regret is recalcitrant. We cannot cease regretting despite our judgements deeming it inappropriate. Likewise, if biperspectival regret is a genuine emotion, for which I argue, it operates more or less the same as other emotions in terms of recalcitrance. However, this claim may appear to be contentious. As discussed earlier, experiencing an emotion presupposes its formal object with an evaluative character. If one continues to experience regret despite ostensibly deeming it inappropriate, certain evaluative judgements must persist to render regret fitting. The confusion between these judgements resembles mistaking moral correctness for emotional fittingness (D’Arms & Jacobson Reference D’Arms and Jacobson2000). Here we confuse judgements that align with the formal object with judgements about whether one ought to feel regret. However, without the former judgements, regret would be impossible. Hence, evaluative judgements connected with the formal object precede judgements concerning the appropriateness of experiencing the emotion. Therefore, regret indeed proves recalcitrant due to these prior judgements, as does biperspectival regret, given its reliance on its formal object.

Lastly, regret motivates improved future decisions (Svoboda Reference Svoboda2020: 86–87). Even when it fails to motivate successfully, it prompts us to contemplate less regrettable future actions. Biperspectival regret also motivates, albeit less intensely. Those experiencing biperspectival regret may reasonably be expected to behave more carefully regarding outcomes beyond their individual control. For instance, we could reasonably anticipate Avis and Ernest operating the draisine even more cautiously in future. Likewise, Griffith might reasonably be expected to reconsider his commuting habits or to attend more carefully to the cumulative consequences of everyday actions.

Story (Reference Story2019) makes a similar point in his discussion of ‘normative entanglement’. He argues that agents whose moral status becomes bound up with one another’s actions in collective contexts are defeasibly entitled to demand due care from those who affect their moral standing (Story Reference Story2019: 610–611).Footnote ¹⁰ The quasi-spectatorial aspect of biperspectival regret is what in part calls for such normative demands, since the contributor recognises that others’ conduct bears on her moral situation. Yet the quasi-agential aspect extends the motivational force to the contributor herself, prompting her to attend to her own conduct as well.

Possessing these features, biperspectival regret clearly qualifies as a form of regret. Its constitutive thought, articulated in the synthesis above, renders the feeling milder than agent-regret yet more intense than spectator-regret, thereby reflecting the contributor’s intermediate position between agent and spectator. Because this phenomenology corresponds to the formal object I have identified, namely the ‘comparative disvalue of group harm to which one faultlessly contributed’, I propose that biperspectival regret is the only appropriate emotion for a contributor who acknowledges that formal object. However, some might think of other candidates.

6. Biperspectival regret and other alternatives

Recently, several philosophers have approached regret from novel angles. Three particularly notable forms are Telech’s ‘relation-regret’, Ferguson’s ‘social-regret’, and Wringe’s ‘collective moral regret’.

Relation-regret, as articulated by Telech (Reference Telech, Szigeti and Talbert2022: 236–237), is a negative, self-directed emotion experienced in response to harmful actions committed by someone with whom one shares an intimate relationship, such as a family member or loved one. Two significant features distinguish relation-regret from biperspectival regret. First, relation-regret explicitly excludes any causal involvement by the individual. By contrast, biperspectival regret essentially involves one’s causal contribution. Secondly, relation-regret arises specifically in response to morally significant wrongdoing, exemplified by cases such as Sue Klebold’s regret concerning her son’s involvement in a massacre, or Rainer Hoess’s regret for his grandfather’s complicity in the atrocities at Auschwitz (Telech Reference Telech, Szigeti and Talbert2022: 249). These examples indicate that relation-regret arises in cases of intentional wrongdoing. However, Sepinwall (Reference Sepinwall2022: 213) characterises similar phenomena as instances of shared guilt arising from shared agency between closely connected individuals. This makes one feel partly responsible for the morally wrong actions of loved ones. Similarly, Sussman (Reference Sussman2018: 797) emphasises remorse as fitting in these contexts due to the acknowledgement of moral fault. Such emotions are explicitly tied to intentional wrongdoing, and are thereby inapplicable in cases for which biperspectival regret is appropriate.

Another relevant emotion is Ferguson’s social-regret (Reference Ferguson2024: 249). He suggests that when one has a social connection (but not an agential connection) to a group responsible for wrongdoing, one may experience social-regret. Like biperspectival regret, social-regret involves group-based incidents. However, unlike biperspectival regret but like relation-regret, social-regret involves no personal causal contribution to the harm either (Ferguson Reference Ferguson2024: 253, 264–265). Ferguson also situates social-regret between agent-regret and spectator-regret and argues that spectator-regret fails to recognise one’s affiliation with the wrongdoers (Ferguson Reference Ferguson2024: 256). Nevertheless, I suggest social-regret aligns more closely with spectator-regret, since both fundamentally involve witnessing a harm without causal involvement, unlike agent-regret, which requires causal implication. Furthermore, as with relation-regret, social-regret also responds specifically to moral wrongdoing. Therefore, biperspectival regret remains distinct.

Wringe (Reference Wringe2024) has recently developed an account of ‘collective moral regret’ directed at collective omissions. In his paradigmatic cases, a harm exists or unfolds independently of the agents in question, and those agents, though collectively positioned to avert it, fail to organise an adequate response. For instance, where a group collectively fails to rescue a drowning swimmer, the constitutive thought is ‘we should have saved the swimmer, and we did not’. In such cases, the regret is directed at a gap between what was collectively required and what was collectively done, although it is felt by individuals. Nevertheless, this structure does not obtain in faultless group harms. For instance, in Draisine, the child is not at risk independently of what Avis and Ernest do. In a sense, the harm does not await a collective response that fails to materialise, but rather, the harm is produced by the group action itself. Avis cannot coherently think ‘we should have saved the child’, for the child was not endangered until their operation of the draisine created the danger. Nor can she think ‘we should not have done what we did’, for the harm was unforeseeable and operating the draisine was not in itself wrongful. Collective moral regret, as Wringe conceives it, presupposes that the group stood in a position to prevent a harm and did not. By contrast, biperspectival regret arises when the group action itself gave rise to harm that no contributor intended or foresaw. Therefore, these two forms of regret respond to fundamentally different relations between groups and harms, and neither can be subsumed under the other.

Aside from regret-based emotions, some scholars propose collective emotions explicitly directed at group wrongdoing, such as collective guilt (Gilbert Reference Gilbert2000; Reference Gilbert2002; Tollefsen Reference Tollefsen2006; Isaacs Reference Isaacs2011; Neog Reference Neog2023), collective remorse (Gilbert Reference Gilbert and Jokic2001; Wringe Reference Wringe2012), and corporate guilt and indignation (Björnsson & Hess: Reference Björnsson and Hess2016). These emotions, whether experienced by individuals or groups, presuppose intentional and voluntary wrongdoing. Indeed, guilt and remorse specifically require agents’ recognition and acknowledgement of their moral faults (Scarre Reference Scarre2017: 570; Sussman Reference Sussman2018: 795; Wojtowicz Reference Wojtowicz2022a: 73). Such recognition would be misplaced in cases of faultless group harm, such as those at issue here. Consider whether it would ever be justified to blame an individual who, despite acting faultlessly, contributed to faultless harm (e.g. Avis). Clearly, this would be inappropriate. Thus, emotions like collective guilt or remorse, traditionally associated with culpability and intention, are unsuitable responses to cases involving faultless group harms.

Another way of approaching the problem is through the concept of complicity. Kutz (Reference Kutz2000) builds his account of complicity around ‘participatory intentions’. He argues that one who intentionally participates in a collective endeavour is accountable for what the group does, even independently of whether one’s individual contribution made a causal difference to the outcome (Kutz Reference Kutz2000: 82–83, 122–123; cf. Mellema Reference Mellema2016: 2–8). Yet complicity accounts, however varied in their details, share a structural presupposition that excludes the cases with which I am concerned. They address situations in which participants intentionally engage in wrongdoing or are aware of the wrongful character of the collective activity. By contrast, in faultless group harms, the contributor neither intends nor foresees the harmful outcome, and the harm does not constitute wrongdoing on her part. Therefore, complicity accounts are not equipped to identify the fitting emotional response in such cases. Biperspectival regret speaks to precisely this space, for it concerns not the moral accountability of intentional participants in collective wrongdoing but the fitting emotional response to faultless group harm.