2.a. Contingent features

Motivation to disburden: First, although people frequently act on behalf of another to make things easier for them, such motivation is unnecessary for on behalfness. This is clear from the fact that someone can act on behalf of a deceased person. For example, a daughter might express gratitude on behalf of her deceased mother to a benefactor her mother never had the chance to meet. This is acting on behalf, it seems natural to think, because the daughter is in some sense realizing, or attempting to realize, what she takes to be the will of her mother. The fact that she cannot bring her mother satisfaction, or improve her wellbeing in any way, is beside the point.

Benefit: This simple point helps us to reject a more general proposal as to what acting on behalf of might consist in: the idea that thus acting is essentially a matter of acting in order to benefit T or increase their welfare. Of course, in many everyday cases, an orientation to someone’s will and to their welfare will generate the same result (Pitkin, Reference Pitkin1967, 156). However, when someone asks us to do something for them, we can do it on their behalf without needing to consider whether doing so really will benefit them. Indeed, we can do something on behalf of some T whilst believing that it will not enhance their wellbeing, even believing that the deed might be detrimental to it. T might request that A brings him some homeopathic medicine, which A believes is at best useless, possibly even harmful. Nevertheless, it seems that A can still pick up the medicine (or “medicine”) on T’s behalf.

Requests: Acting on behalf of someone, we can say, involves adopting their “will” at least in the local matter at hand, and attempting to realize it. This is the sense in which A takes up the agential perspective of T relative to some state of affairs. For the purposes of this article, I will take it that the will of an agent is simply a relevant goal she possesses in that context. We will return to the idea of goal possession. Here, we can note the default everyday assumption that a person’s requests are reliable guides to their goals. Only where we have indications that she is under duress, under psychological compulsion, or massively deluded do we retreat from the assumption. Nevertheless, requests are not necessary for an agent to act on behalf of a target. There are other sources of knowledge or belief that T has a goal, which may well ground A’s action on behalf of T. The behavior of the elderly supermarket customer may suffice for Sue to draw conclusions about the customer’s goal; knowledge of the general human aversion to being attacked is likely to allow Parisa to attribute the goal of self-defence of Liam.Footnote ⁵

Institutionalization: The fact that no explicit request is needed for behalfness helps to situate another key set of cases. Certain institutional arrangements, particularly those of judicial and political representation, are set up to enable specific individuals to act on behalf of others. Lawyers act on behalf of their clients and elected representatives are supposed to act on behalf of their constituents. In spite of their differences, these cases are unified by requiring explicit delegation of the principal’s “will.” Moreover, this explicit step of transfer also involves the principal renouncing her own agency in the matter at hand, at least up to a point. However, neither of these features—delegation or investment with independent authority to act—need to be present for action to be performed on behalf of another. Their necessity is, rather, a consequence of the institutionalization of on behalfness. Such formalized features both specify legal or political criteria for on behalfness and set its limits. In such contexts, we need to know whether the for-relation is instantiated and how far it reaches; although everyday behalfness can involve such explicit delegation, its precise contours are often less tangible.Footnote ⁶

2.b. Normative on behalfness

Political and legal forms of action on behalf of others are restricted and channeled by norms. Moreover, certain everyday cases also involve important normative dimensions. Take the example of A standing up for T’s moral right not to be subject to racist abuse. Speaking up for someone’s moral claims is certainly an important example of acting on behalf of another. It has recently been proposed that we should think of all action on behalf of others as normatively constituted. More precisely, the proposal is that A’s φ-ing is an action on behalf of T iff it is constitutive of A’s φ-ing that A acts on claims, duties, or normative powers derived from the claims, duties, or normative powers of T (Edlich & Vandieken, Reference Edlich and Vandieken2022, 543). A speaks up on behalf of the racially abused T by expressing T’s right not to be thus abused. According to the proposal, this is even possible if T does not have the goal of standing up for her rights in the context (Edlich & Vandieken, Reference Edlich and Vandieken2022, 546). Presumably, it is also compatible with T not having any goal to stand up for her rights.

This view appears to be only applicable to a highly restricted set of cases. In this, it contrasts with the goal-based view. The two views are equally applicable where A acts to realise a goal of T’s and where that goal is derived from some normative status that T has. However, if the normative status does not generate a goal of T, A can still act normatively on behalf of T, but cannot do so on the goal-based view. The example of A protesting against the racial abuse of T, where T has no goal to protest against being abused, is a case in point.

Conversely, if T’s relevant goal does not derive from a normative status, the view proposed here will support the assertion that A acts on behalf of T, whereas the assertion would be false on the normative view. Indeed, the normative view appears to exclude the majority of cases we would normally classify as action on behalf of another. A can adopt T’s goal to realize p and realize p on behalf of T whether or not T has a claim that p be realized. Most such cases will presumably count as normatively permitted. But to φ where φ-ing is permitted is not to φ because φ-ing is permitted. It will be to do so for some other reason. A might, for instance, simply think that her (A’s) φ-ing would save T time and φ for that reason. Indeed, according to everyday understanding, it is not even true that what A does on behalf of T has to be morally permissible. T might have a blood feud with S and intend to murder S; knowing this, A might murder S on T’s behalf.

The normative and goal-based views seem to reconstruct distinct and apparently compatible, if overlapping relations. One might wonder whether one of the views is more basic and could therefore be used to reconstruct the other.Footnote ⁷

The normative account might appear to be the basic one if setting and pursuing goals is construed as necessarily acting on a normative power. However, this strategy raises the question as to what acting on a normative power might mean. Acting on claims or duties is naturally understood as acting for reasons provided by, or constitutive of, such normative statuses. Normative powers, in contrast, are exercised and hence do not entertain the same kind of relationship to reasons. Moreover, a strategy that emphasizes the normative powers at work in morally uninteresting everyday human behavior would presumably deny that non-human animals can set and pursue goals in the same sense that humans can. The analysis of goals I propose in the following section requires no components that are only accessible to agents involved in normatively structured interaction. I would suggest that parsimony at this point keeps us closer to our everyday understanding.

Alternatively, one might attempt to ground the normative view in the goal-based view. One could see the claims advocated for in normative on behalfness as derived from goals of the relevant normative community, which is thought of as the issuer of relevant principles or norms. These goals could be in place even if individuals protected by the norms do not share them. This strategy would, however, be incompatible with Edlich and Vandieken’s account, which grounds in a relational view of morality, according to which moral claims have their “sources” in individuals (Edlich & Vandieken, Reference Edlich and Vandieken2022, 552f; Wallace, Reference Wallace2019, 62). They emphasize this feature of their view in order to reconstruct the intuition that the relevant moral claims are advanced on behalf of particular others, rather than on behalf of moral principles, or indeed of the moral community. Hence, their view depends on the truth of a very specific moral metaphysics, a fact which naturally reduces its ecumenical appeal. A view that assigns “the moral community” relevant goals would, however, be no less rich in metaethical presuppositions.

The key case of apparent normative on behalfness that would remain outside the purview of the latter view is one in which (a) a victim would not even have a pro tanto desire for her claim’s satisfaction if she were to be presented with the key normative and empirical details of her case,Footnote ⁸ and (b) the agent’s proactive intervention nevertheless appears insufficiently described as an intervention in the name of what is morally right. Whether there are such cases would have to be elucidated by attention to specific examples. If it turns out that there are, further metaethical accounts would also be explanatory contenders, such as an Adam Smithian view that the relevant goals are those of impartial respondents. Adjudicating these important issues is clearly beyond the scope of this paper.

3.a. Goal possession

Begin with the first issue: talk of goals is ambiguous between talk of states of affairs that are aimed at and talk of aiming at such states of affairs. Simply naming a state of affairs does not specify that the attitude to it is one of aiming. “(My) owning a mansion in Hollywood” or “world peace” can, in one sense, be goals, but they can also be the objects of other attitudes, such as belief, pride, or satisfaction. We can express that a state of affairs is a goal by inserting its representation into the structure “Let it be the case that p.” An attitude that is adequately expressed in this way can be labeled “optative.”Footnote ⁹ According to this wide conceptualization, goals can be set by whimsical, evanescent desires or by the wholehearted subscription to what one takes to be non-negotiable values or principles. On behalfness can therefore be in play either in momentous or trivial matters.

It seems, though, that there are some goals of an agent one cannot adopt in order to act on their behalf. People can be alienated from goals. We can stipulate that these are goals of an agent that she does not possess. As already indicated, duress, psychological compulsion, or massive delusion can lead us to refrain from seeing a goal expressed by an agent as possessed by her. Under such circumstances, the goal is no longer a candidate to ground action on behalf of T. Conversely, there are circumstances under which T has not taken a positive optative stance on some matter, but where A is convinced that T would take such a stance if she were to attain awareness of the relevant circumstances. If A thinks T would want a ticket to some event if T knew it was imminent, A can buy the ticket on T’s behalf, although T has not formed a relevant desire because of her lack of a relevant belief. Hence, the “will” adopted may be a desiderative disposition, where the agent believes the disposition would be triggered under ameliorated cognitive conditions. The rationale for circumscribing goal possession in these ways is that, when A ascribes a goal to T as the basis for acting on T’s behalf, A is taking it that this is “where T stands” on the matter at hand.

3.b. Motivating reasons and expressive actions

The second issue concerns the way an agent’s goal-orientation relates to her possession of motivating reasons for action. The reasons, I assume, are facts. Certain facts are motivating reasons for an agent in view of their relevance for the realization of her goals, where goals can include values or principles that are optatively underwritten.Footnote ¹⁰ Thus understood, most of our motivating reasons are “instrumental”: in acting for such a reason, an agent takes some fact, in view of which her φ-ing appears necessary for, or conducive to, the realization of her goal, to be a sufficient reason for her φ-ing. The goal is no part of the reason, but is, in the background, essential to the fact’s being a motivating reason. The fact that T’s flowers need water is a reason for her to water them in view of her desire to have flourishing flowers.Footnote ¹¹ The same fact becomes a reason for A to water the flowers on behalf of T in view of A’s adoption of T’s goal.

There is, however, a class of actions the reasons for which are not, even in this sense, instrumental. Not all facts that motivate our actions do so because they appear indicative of opportunities to attain goals. This led Rosalind Hursthouse to postulate a class of “arational actions”: someone who jumps for joy or hides their face in shame is not doing so in order to attain some goal (Hursthouse, Reference Hursthouse1991, 58). Moreover, at least some such actions can also be performed on behalf of others, such as when a father jumps for joy at his daughter’s performance or a daughter hides her face in shame at her father’s behavior. Hence, if A’s possession of a goal g is a crucial background condition for her acquisition of motivating reasons to φ—whether in propria persona or on behalf of T—this condition cannot depend on A construing her φ-ing as a means to attaining g. Instead, a fact’s enabling or constraining action suitable to attaining g will have to be a specification of a more general schema for motivating reasons that also encompasses spontaneous actions which express the agent’s subscription to g.

In fact, Hursthouse had claimed that such actions are not performed for reasons at all. She based this claim on a Davidsonian belief/desire account of reasons. However, a number of authors have convincingly disputed the claim. According to Monika Betzler, reasons for such expressive actions are furnished by “challenges” to an agent’s evaluative perspective (Betzler, Reference Betzler2009, 280); according to Chris Bennett, situations an agent takes to contain a particular value or disvalue provide reasons to perform actions that “do justice” to the situation’s importance for the agent (Bennett, Reference Bennett, Abell and Smith2016, 78ff.). On both accounts, the status of facts as reasons derives from something or someone mattering to the agent in a way that is impacted by the relevant facts. In other words, the agent with motivating reasons for expressive action subscribes to some value. Such subscription, I have claimed, involves taking on an optative stance of the form “Let p be the case.”

The decisive point for our purposes is that facts provide motivating reasons for spontaneous expressive actions for A where they appear to indicate that a goal of A’s, i.e., the content of an optative stance of hers, has been impacted in some specific way. The apparent realization of a goal can be a sufficient reason to jump for joy; its frustration a reason to swear; its definitive defeat a reason to tear one’s hair. In contrast to cases of instrumental action, the action schemas realized in such spontaneous expressive actions may not figure in the content of any belief of the agent. Such schemas may be partially innate or simply overlearned reaction types.

An spontaneous expressive action φ, we can say, involves an agent treating some fact p, in view of p’s apparent relevance for the realisability of her goal, as a sufficient reason for her φ-ing. This circumscription is wider than that covering instrumental motivating reasons, which ground in a specific form of relevance: that the fact is one in light of which the action appears conducive to the goal’s realization. Hence, the characterization appropriate for expressive action turns out to be the generic characterization required to cover both variants of motivating reasons. Moreover, the characterization is transferrable to expressive action on behalf of others: a father has sufficient reason to jump for joy at his daughter’s winning the race because it impacts his adopted goal that she be successful; and a father’s spilling beer all over himself provides his daughter with sufficient reason to hide her face in shame because it impacts her adopted goal that he not make a fool of himself.

3.c. Goal adoption and reason acquisition

Returning to action on behalf: I am proposing that A’s φ-ing on behalf of T involves A adopting T’s goal and, as a result, acquiring T’s reasons for φ-ing. Now, we can say that adopting a goal is simply a matter of taking an optative stance on the matter at hand. What is then decisive is the connection between A’s stance and the (perhaps counterfactual) stance of T. We are interested in cases in which A adopts some g that she believes T possesses because she (A) believes that T possesses g. The “because” here is both causal and justificatory: the effect of A’s belief that T has goal g is A’s adoption and pursuit of g in virtue of A’s taking T’s possession of g to justify her (A’s) adopting and pursuing it herself. Importantly, the justificatory component excludes A’s belief that g is T’s goal from leading A to adopt g “behind her (A’s) own back”: she cannot just find herself with the goal as a result of the ascription. She needs to think that the fact that the goal is T’s is a reason for her (A) to adopt it and then do so for that reason. Where this is the case, I shall talk of allogenic goal adoption (AGA).Footnote ¹²

Where A takes T’s possession of g to be a reason for her (A) to adopt g, her doing so will in turn generally ground in other reasons. Those reasons will frequently be relational: that T is A’s daughter, belongs to some social group to which A belongs, or that T is someone A admires. However, the reason might be non-relational, as with the fact that T is suffering in some specific way.

In short, AGA is a move made for a reason (that g is T’s). On the one hand, this reason is likely to be backed up by some further fact, which A takes to justify seeing T’s possession of g as such a reason. On the other hand, AGA puts A in a position to acquire a new set of reasons: certain facts that are otherwise irrelevant to the agent become motivating reasons for her. To act on behalf of someone thus involves acting for the reason that p, where the status of p as a reason is grounded in an allogenically adopted goal.

3.d. The initial proposal

We now have the materials to present a first-pass analysis of acting on behalf:

(AOB/1)

A acts on behalf of T in φ-ing iff:

1. 1. A takes some fact p, in view of p’s apparent relevance for the realisability of goal g, to be a sufficient reason for her φ-ing;

2. 2. A adopts g because she believes g to be T’s goal;

3. 3. A φs for the reason that p, where that reason is grounded in the apparent realisability relevance described in (1) and the allogenic feature described in (2).

AOB/1 is just the structured presentation of the materials we have been discussing. Condition 3, however, requires a brief comment: it connects A’s action to her taking p as a reason with a particular kind of grounding. Without such an explicit connection it is possible that A might take p to be a reason grounded in more than one way and act for the reason that p, construed in some other way. Perhaps A enjoys looking after children so much that, although she takes T’s desire that the kids be looked after to ground as a sufficient reason for her to do so, she is just looking forward to looking after children again. She might then agree to do so for the latter, rather than for the former, reason. What guides her action has to be her taking of p to be a reason, where the reason is grounded in her allogenic goal adoption in conjunction with the realizability relevance of p.

4.a. Indexicality

The first worry concerns the coherence of the concept of AGA. Goals’ contents generally involve reference to their subject. For example, John’s daughter can express her goal as “that I win the race.” The self-reference by means of the first-person singular pronoun is, as Castañeda and Perry have influentially argued, irreducible to any reference mediated by reference to properties or names. Most importantly here, it picks out its referent in a way that allows the direct engagement of the referent’s action-guiding and deliberative mechanisms. Perry thought that no belief that identifies Perry via property ascription can be the “same belief” as one referring to himself by means of “I” (Perry, Reference Perry1979, 3). Should we, similarly, worry that it is impossible for A to adopt T’s goal, as the change in the way of referring to T entails that A cannot have the same goal as T? Certainly, in taking on T’s goal, A cannot represent T’s goal in such a way that the connection to T’s system of action control is upheld. Through the goal’s allogenic adoption, the connection is instead established to A’s own action control system. Indeed, that is the whole point of AGA.

It is a matter of metasemantic controversy how to characterize the difference between John’s daughter’s self-reference and John’s reference to her by means of “she.” But even if the difference is taken to be one of content, and not merely of mode of presentation, it is difficult to see why this threatens the idea of AGA. John can think about his daughter’s goal in a way that recognises her as its agent, without being able to inhabit her first-person agential perspective. Even if the goal’s content changes, it is surely close enough. Indeed, Castañeda claims that we have specific semantic means to refer to the first-person perspective of another agent, in spite of our inability to express that perspective (Castañeda, Reference Castañeda1967, 94). Third-person pronouns such as “she” and “her,” when used in certain ways in attitude ascription, are what he calls “quasi-indicators.” Their function, he claims, is precisely to enable intersubjective reference to “indicators” such as “I” and “me,” whilst taking account of the irreducibility of the first-person standpoint to which such indicators provide a direct referential route.

The importance of “quasi-indicators” in thinking about others’ goals becomes clear when we turn to a first type of case that might be thought to show that the AOB/1 conditions are insufficient:

A is clearly not siding with, but working against, T. But there may appear to be a sense in which A adopts g because she believes g to be T’s goal and acts for reasons that are indicative of g’s realisability. Nevertheless, A does not really adopt T’s own goal. This can be seen from the fact that A eradicates any reference to T’s agential perspective in the goal’s content, instead deploying her (A’s) own first-person “indicator.” Adopting T’s goal would require that A respect T’s agential perspective, picking it out by means of a third-person “quasi-indicator.” The competitor acquires the goal that she, the competitor, acquire some property, not that T do so. The same is true of what we might call “copycat action,” in which A pursues g in order to be like T. Again, A’s goal will be indexed to A herself, and not to T: because T dyes her hair blue, A dyes her, A’s, hair blue. But that A’s hair be blue is no goal of T’s. Hence, competition and copycat action are incompatible with AGA, not counterexamples.

4.b. Agent-neutral goals and extensional adequacy

However, index adjustment cannot deal with all issues of extensional adequacy. This is a result of the fact that not all an agent’s goals are indexed to the agent herself. The contents of some of our goals do not involve us: we can aim at peace between warring factions or at higher academic standards. Call these “agent-neutral goals.” Where T has such goals, their adoption by A will not involve their indexing to A.

A little reflection suffices to show that allogenic adoption of agent-neutral goals can only take place under very special circumstances. This is because agent-neutral goals essentially ground in the value they are taken to have. But where an agent takes on a goal because of its putative value, she does not take it on merely because it is the goal of some other agent. Brief consideration of the way in which children come to take on the goals of their parents may be helpful here.Footnote ¹³

Importantly, most cases of children’s adoption of their parents’ goals are not cases of AGA. First, children adopt many goals because their parents have these goals, but without seeing their parents’ possession of the goals as justifying their adoption. Such processes involve causality without any sense of justification. Second, when children do adopt their parents’ goals for what they take to be good reasons, the reasons that sway them will very often concern other matters: avoidance of disapproval, garnering of approval, or the presumed value of the goals. Goals adopted for any of these reasons are not allogenically adopted. So, even when a child does adopt an agent-neutral goal of their parents, such as tidiness, the standard cases will not involve AGE. Children are likely to begin with automatic, justification-independent goal acquisition—inculcation—before perhaps coming to appreciate what they take to be the goal’s agent-neutral value. Where this is the case, neither phase of the goal’s possession involves allogenic goal adoption.

This leaves us with very few cases. Perhaps a child might adopt the goal of a tidy dining room simply because she knows her father always aims to keep the dining room tidy—although she cannot see any value in tidiness and is unconcerned with his approval. If this is the case, when the child tidies up, she is indeed acting on her father’s behalf. She is then doing something for her father; she is doing it merely because her father wants the state of affairs that she is bringing about to be brought about.

There are, however, two types of counterexamples that AOB/1 cannot deal with. Both concern the temporal gap between the moment of adoption of a goal and the moment at which on behalfness is in play. In the first case, A at t1 adopts the agent-neutral goal of tidiness merely because it is T’s, but by t2 comes to believe that tidiness is instrumentally good. She continues to pursues it for this latter reason, even though she has in the meantime cut off all contact to T. Where this is the case, A is at t2 no longer acting on behalf of T. However, an advocate of AOB/1 is stuck with the assertion that this is a case of on behalfness, as all three conditions are fulfilled. The conditions are insufficient after all.

It turns out that they are also unnecessary. In order to see this, we can imagine an unusual example in which A acts on behalf of T by pursuing an agent-neutral goal of T’s:

Here, A’s action on behalf of T does not involve A adopting T’s agent-neutral goal of regional peace because it is T’s. A’s having the role of group spokesman entails his antecedent possession of the goal. This, let us assume, was the case because of the value A already took the goal to have. AGA, then, cannot be a necessary condition of on behalfness. Indeed, it may seem that, in this case, A cannot even act on behalf of T if on behalfness is goal-based.

4.c. From adoption to espousal: the tweak

Fortunately, we can explain where things have gone wrong. On behalfness is indeed goal-based, but the form of goal-basing needs reconceptualizing. What is required turns out not to be a genetic condition, but a guidance condition: that A’s action is guided by g, in light of relevant beliefs of A, where and because A construes g as a goal of T’s. Where a goal guides an agent’s action, I shall talk of goal espousal on the part of the agent. Where A espouses T’s goal, A’s action is responsive to features of the world that A thinks would make a difference in the way T would attempt to realize the goal. Now, where A possesses the same agent-neutral goal in proprio persona, her pursuit of the goal as T’s goal may or may not make an observable difference. However, where this is not the case, the difference can be revealed by relevant counterfactuals. These can concern, first, the action’s performance or non-performance or, second, the way or style in which the action is performed.

Beaten-up activist provides an example of the first kind. Here, the spokesman calls for exclusively non-violent responses because he takes such a call to further the goal of regional peace, doing so because the continuing pursuit of the goal is what he takes T to want. Because he already possessed the background goal, he may well have performed precisely this action anyway, without any thoughts for T’s goals. But imagine that his spokesperson role involves all sorts of other duties and that, had it not been T who had been struck down, he would have prioritized another of those duties. Such a counterfactual supports the claim that he is pursuing the goal he already had under a particular guise, that is, as the goal of T. Here, it is the spokesman’s conceptualization of the goal as the victim’s that explains its guidance of the spokesman’s behavior.

Another more prosaic case provides an example of the second kind: A’s flat has a number of plants that she wants to flourish. T, who moves in, is more strongly invested in plant flourishing than is A and takes over watering duties. As a result of her stronger attachment to the plants, T takes their flourishing more seriously than A does. When T is away for a longer period, A takes over plant watering duties again, now attempting to fulfil these in the particularly serious and detailed way that T would. A has in propria persona the goal that the plants flourish; however, were she not to conceive the goal as T’s, she would be much more lackadaisical in pursuing the goal.

We can think of A’s guidance by a goal in view of its ascription to T as involving a specific mode of representation, in part constituted by the deployment of “quasi-indicators” to make salient the first-person perspective of T. In order to capture the case in which A espouses a goal because it is T’s, I will talk of allopsychic goal espousal. This slightly heavy-handed terminology seems worthwhile, as it makes explicit the replacement of a genetic condition by a guidance condition that merely presupposes that A takes T to possess the goal, independently of how A came to have the goal herself.Footnote ¹⁴

Replacing allogenic goal adoption (AGA) with allopsychic goal espousal (AGE) solves the extensionality problems of AOB/1. An analysis that works with AGA overgenerates since it is compatible with A adopting g because it is T’s, but then pursuing g for other reasons—perhaps because A now believes g is valuable. If A’s φ-ing is under the control of allopsychic construal, this possibility is excluded.

The requirement that the response be guided by an allopsychically espoused goal also solves the problem that AOB/1 undergenerates. Whereas working with AGA excludes A acting on behalf of T if A has g independently of T, switching to AGE allows acting on behalf under such conditions. The occurrent guidance of an action by a goal seen as justified as a result of its possession by another agent is independent of who had the goal first. Hence, we can explain how, in Beaten-up activist and in the flower-watering example, A can act on behalf of T.

Replacing AGA with AGE in the analysis of acting on behalf produces corresponding modifications of condition 2 (“A espouses g” replaces “A adopts g”) and condition 3 (“the allopsychic,” rather than “the allogenic feature”) of AOB/1. At the end of the following section, I will provide a definitive analysis of generic on behalfness, which incorporates this tweak.

5.a. Analogies with action

The first thing to notice is that, in each of these cases, the agent might have responded with an action to the considerations in light of which they responded emotionally.Footnote ¹⁵ For example, if Afra is near enough to Tara, she might move between her and the dog. Conversely, an agent can take on an emotion on another’s behalf in view of the same considerations as those in view of which they can act on their behalf. It is natural to say that they can do so for the same reasons:Footnote ¹⁶ John’s daughter’s crossing the line first is a reason for his joy; that the dog is large and has sharp teeth is a reason for Afra’s fear.

Emotions on behalf of some T, then, are like actions on behalf of T in being responses to reasons the responder takes T to have.Footnote ¹⁷ Moreover, such emotional responses are similarly guided by an allopsychic goal of T: John espouses his daughter’s goal to win; Afra espouses Tara’s goal not to be harmed. This corresponds to a general dependency of emotional dispositions on agents’ goals. Consider fear: being intent on keeping a relationship intact, aiming to pass an exam, and wanting to avoid appearing to be a stereotypical blond are all goals that may ground dispositions to be afraid and hence feed into an agent’s treating certain facts as reasons for fear.

5.b. Disanalogies with action

The grounding of emotional responses, and of the reasons for those responses, in agents’ goals is familiar to everyday understanding. However, there are two important disanalogies between action and the emotion cases which can easily obscure this structural commonality.

The first is that emotional responding never has the instrumental structure characteristic of paradigm cases of acting for reasons. A’s emoting does not entail that A take her emoting to be necessary or conducive to the realization of a goal. Indeed, emoting, whether on someone else’s behalf or not, does not require thinking about the emotion at all. Rather, what is required is merely that A take some fact to have immediate goal-relative significance. It follows that, where A emotes on behalf of some T, what is required is the satisfaction of an equivalent of the first condition of AOB/1. We simply replace the action token variable φ with an emotion token variable.

The second disanalogy concerns the cognitive relation of an agent to her reasons in the action and the emotion cases. In the action case, it has generally been assumed that the reasons for which people act are the objects of their beliefs. An agent, it may seem, needs to believe that p is a—normative—reason for her to φ in order for p to play the role of a motivating reason. Whether or not this is true for action—and, again, spontaneous expressive actions are grounds for skepticism—it is certainly not true for emotions. It seems clear that emotions are often triggered by mere perceptions.Footnote ¹⁸ But if this is the case, it may seem implausible that people emote for reasons in the same way that they act for reasons.

The challenge here is to explain the basing relation, that is, the relation between the adoption of an attitude and its cognitive base. For the purposes of this paper, I shall adapt a proposal advanced by Lord and Sylvan as an analysis of the epistemic-basing relation. According to the proposal, treating a fact f as a normative reason for belief B involves being disposed to respond to f-like facts with B-like beliefs that are true, where the disposition counts as a kind of competence. More generally, treating as a normative reason for a response involves having the competence to respond in conformity with certain patterns that meet standards of success (Lord & Sylvan, Reference Lord, Sylvan, Carter and Bondy2019, 161ff.). Adapting the proposal for our purposes, we can say that facts are treated by an agent as normative reasons for particular emotions iff their responding to them in those ways is part of a competence-constitutive pattern of response which tends to generate emotions that are appropriate. John, for instance, tends to feel joy when things go as he had hoped or planned, and Afra tends to respond to dangerous situations with fear, rather than doing so with amusement.

Thus understood, treating a fact as a reason is a broader relation than that of taking or believing a fact to be a reason. Conversely, taking a fact to be a reason is an explicit way of treating it as a reason. The analysis of on behalfness should cover both belief and inexplicit treating, where the latter is plausibly at work both in emotional responding and in spontaneous expressive action. Hence, in order to generate a generic analysis of on behalfness, but also in order to get an inclusive concept of acting on behalf, AOB needs a further slight modification. In its first condition, “A takes some fact p … to be a reason” is replaced by “A treats some fact p … as a reason.”

Implementing this change, along with the substitution of AGE for AGA, leaves us with the following generic analysis of responding on behalf:

(OB)

A responds on behalf of T in ρ-ing iff:

1. 1. A treats some fact p, in view of p’s apparent relevance for the realisability of goal g, as a sufficient reason for her ρ-ing;

2. 2. A espouses g because she believes g to be T’s goal;

3. 3. A ρs for the reason that p, where that reason is grounded in the apparent realisability-relevance described in (1) and the allopsychic feature described in (2).

5.c. Emotions on behalf as ditransitive emotions

With these disanalogies out of the way, we can turn to the operation of allopsychic goal espousal in cases of emotional on behalfness. Returning to our cases, we can see the difference made by the guidance of a goal construed as the goal of the target. As in acting on behalf, this is not just a matter of A having a goal the content of which includes things going well for T. What, again, is decisive is that, in emoting, A is guided by the goal itself whilst construing it as T’s. A, we can say again, takes on T’s goal under a mode of presentation that foregrounds T’s first-person agential perspective on its realization. To get clear on the difference, compare:

• (1OB) John is happy for his daughter that she has won the race.

• (1P) Joe is happy that his daughter has won the race.

• (2OB) Afra is afraid for Tara that the dog might bite her.

• (2P) Avi is afraid that the dog might bite Tara.

We might think of 1OB and 2OB as describing ditransitive emotions, that is, as emotions with both a nominal and a propositional object. In contrast, 1P and 2P have purely propositional objects. The key difference is that the latter two emotions can be taken on for reasons for which the goals of the agents referred to in their contents are irrelevant. In 1P, Joe might, as it happens, be completely uninterested in whether his daughter wants to win the race, but be happy merely because she has outcompeted Jim’s daughter; and in 2P, Avi’s fear of the dog’s bite might derive entirely from his own squeamishness at the sight of blood. In such cases, E’s emoting is not being guided by a goal understood as a goal of T. Rather, a fact that in some way concerns another agent, and may or may not concern their goals, is taken to be relevant for some goal of the emoter and treated as a thus-grounded reason for her emotion.