1 Introduction

A man opening the door of his fridge and grabbing a beer can on a hot summer day. A woman taking her credit card and inserting it in an ATM. These examples of behaviors are easily intelligible to us insofar as we can interpret them as intentional actions that fall under the realm of reasons. It is because the man wants to quench his thirst and believes that there is some beer in the fridge that he does so. It is because the woman wants to withdraw cash and believes that the distributor is the right place to get some that she does so. In these simple cases, the reasons that have been found and attributed take the form of beliefs and desires that these agents are supposed to possess and that explain why they behave in that way. Spelling out our understanding of such behaviors in these terms is now commonplace in philosophy. If reasons can be attributed to agents, it is because our practices of mutual understanding are based on the presupposition of a shared rationality between all human beings. Appeals to rationality—as ambiguous as this term in all its broad implications can get—have been therefore advanced in numerous philosophical discussions of mindreading in humans (Dennett 1987; Heal 2003; Millar 2004). In particular, it is because we assume that the agents we are interpreting are rational that we can attribute reasons to them. Their behaviors can be assessed according to principles of rationality insofar as they have more or less good reasons to act. These are necessary conditions for us to be able to explain and predict, and therefore, understand their behaviors in a satisfactory way.

The intersections between mindreading and rationality are fruitful and nourish a rich philosophical literature. Our interest in this paper will be narrower, but it will have general implications for the way we conceive of folk-psychological understanding. We will address the interrelation between reasons and a specific sub-category of mindreading, namely empathy. We take empathy to be essentially an act of imaginative transposition through which a subject, by putting herself in another person’s shoes, is able to represent this person’s subjective perspective in a given situation.Footnote 1

In contemporary debates, one of the most sophisticated attempts to articulate the relation between empathy and reasons has been put forward by the philosopher Karsten Stueber. His notion of “reenactive empathy” (Stueber 2006) is close to ours, since he conceives of empathy as a cognitive process that “requires us to adopt an alternative to the reality of our perspective by transporting ourselves imaginatively into the shoes of another person” (Stueber 2016, p. 327). More precisely, Stueber qualifies this process as “reenactive” because the subject performing it is supposed to reenact in her own mind some of the relevant mental states of the other person in a given situation. Stueber inscribes his account of empathy within the broader framework of folk psychology as guided by a principle of rational agency. In such a framework, reenacting the target’s mental state enables the interpreter to grasp her reasons for acting (Stueber 2017). Discussing this position, which we call an agential conception of empathy, henceforth agential empathy (AE), is the aim of this article.

This model is primarily concerned with a notion of empathy as having to do with the grasping of reasons behind an agent’s behavior. Stueber’s theory is particularly interesting because, as we will see, it is largely aligned with the traditional view according to which our folk-psychological understanding of agents essentially consists of apprehending their reasons for acting, these taking the paradigmatic form of a belief and desire couple. This paper intends to prove that AE, as Stueber articulates it, is defective, not only because it does not do justice to the epistemic richness of empathy, but also because it fails to adequately perform the epistemic function that Stueber ascribes to it.

In other words, we will show that if we want to properly comprehend empathy as capable of providing an understanding of human agency through the lens of reasons, we also need to acknowledge its capability to convey some insight into the experiential dimension of our mental states, viz. the way that they are phenomenally experienced by the agents. We believe this is an aspect that Stueber’s account of empathy overlooks at the expense of the explanatory power of his theory.Footnote 2 We will call this neglected aspect of empathy “Experiential Empathy” (EE). Acknowledging EE’s capability to provide us with an understanding of the experiential dimension of our mental states opens up the space for decoupling the epistemic function of empathy from explaining agency in terms of reasons, i.e. from AE. In fact, achieving insight on the very phenomenal quality of our mental states can constitute an end on its own. We will show that EE can either work alongside AE to provide us with a richer understanding of human agency, or it can constitute an independent epistemic goal of empathy.

Our paper is structured as follows: we will begin by presenting Stueber’s agential conception of empathy (first part), before showing that this view does not provide sufficient explanations of reenactive empathy and thus needs to be complemented (second part). We will prove that Stueber’s model is incomplete by introducing the case of emotional actions. Finally, we will develop our view on EE and  focus on some crucial implications pertaining to its epistemology (third part).

We will shape our arguments based on Stueber’s model of empathy, not only because he is one of the most influential and productive theorists of empathy in contemporary debates, but also because his position is exemplary of a widespread tendency to conceive of imaginative perspective-taking (in our terminology, what we call empathy) as the attribution of beliefs and desires within an agential framework. Indeed, many researchers have usually spelled out processes of perspective-taking in terms of the reenactment, through imagination, of the relevant beliefs and desires supposedly held by the target.Footnote 3

2 The agential conception of empathy

This section begins by introducing Stueber’s agential conception of empathy, namely, a form of empathy that is above all concerned with the understanding of human agency in terms of reasons. We then show how Stueber’s view is largely committed to a standard model of action explanation commonly known as “Humean”, a model which we will present as inadequate to account for the case of emotional actions. This is particularly evident in the case of the so-called arational actions (Hursthouse 1991).

2.1 Agential empathy: reenactive empathy is all about reasons

Stueber (2006, 2008, 2011, 2012, 2017) sees empathy as our default strategy for mindreading, thus playing a central role in our explanations in folk psychology. In this respect, Stueber restricts the scope of empathy to the realm of rational agency. Indeed, to him, “folk psychology is best understood as an explanatory practice, where we attempt to understand each other as rational agents” (Stueber 2011, p. 168). It is within this “rationality assumption” that Stueber articulates his view of empathy. It implies that we can only understand—and, in particular, empathize with—agents insofar as they are rational agents. In his definition of this term, Stueber invokes Davidson’s position in his famous paper “Action, Reasons and Causes” (1963). According to this view, a rational agent is an agent who sees herself as acting out of reasons and who can, therefore, be evaluated according to criteria of rationality.

Rational agents are not merely creatures who act because something is happening inside them. Rather, they are able to take a reflective stance towards their own agency and to take ownership of their actions in terms of their reasons for acting (Stueber 2012: 59).

They do not merely regard themselves as passive entities tossed around the world by their impulses and desires: as fully rational agents they are in the position to say something in favor of their actions (Stueber 2017, p. 140). In other words, to use Karen Jones’ terminology (Jones 2003), a rational agent is not only an agent who is able to track reasons (reason-tracker), but she must also be able to respond to them (reason-responder). Responding to reasons involves much more complex and elaborate reflexive processes—such as a sense of ownership towards one’s actions—than simply tracking reasons—something that also non-human animals can do easily. The fact that an agent can respond to reasons consequently explains the existence of intentional actions, which are precisely carried out in the name of reasons that do not only explain them, but also justify them in light of some propositional attitudes implicitly or explicitly held by the subject. Therefore, only a rational agent will be able to understand another rational agent insofar as she knows what it means to act for reasons.

Throughout his work, Stueber follows a common strategy in the empathy debate distinguishing between two kinds of empathy, basic empathy and reenactive empathy (RE). Basic empathy represents our automatic, quasi-perceptual ability to understand other individuals as minded creatures. It is taken to be mostly accomplished by built-in mechanisms of imitation and resonance such as mirror neurons (Stueber 2006). RE is taken as an imaginative enterprise where we try to reenact other agents’ reasons for acting in a certain way in a given situation. We will focus exclusively on this second, more sophisticated and cognitively demanding form of empathy, because it is only thanks to it that the empathizers can interpret their targets not only as tracking reasons, but also, as responding to them. That is, only this form of empathy can play a fundamental role in enabling us to understand the reasons that motivate or even justify the actions of agents. Indeed, Stueber claims that RE’s epistemic achievement consists in the grasping of reasons for acting. In our terminology, this means that only RE can be associated with AE.

2.2 The belief-desire model of action explanation (and beyond)

2.2.1 The Humean view of action explanation

At first sight, Stueber holds quite a restrictive view of practical reasons. The great majority of the examples he gives is mostly concerned with reasons as being constituted by a belief-desire couple:

To be a responsible agent who can be seen as the author of his actions it is thus required not only that the action is caused by a particular set of beliefs and desires. […] Instead it is necessary that the agent sees these beliefs and desires in some sense as appropriately related to what he is doing and that he acknowledges them as his reasons for his action. (2006: 42)Footnote 4

In this sense, Stueber’s account is largely in line with the standard model of the explanation of action, which is usually referred to as “Humean”, or as the belief-desire model of action. This view was defended, among others, by Smith (1987, 1998) and Davidson (1963).

According to the Humean model of action explanation, intentional actions are caused and explained in terms of reasons taking the form of a belief-desire couple.Footnote 5 More formally:

The reason motivating the intentional action A is that the agent desires D and believes that A is the best way to accomplish D.

That is, an action is necessarily explained in terms of a desire and a corresponding instrumental belief. The reasons provided by this model owe their motivational component to the mere presence of the desire. Humeans explain this resorting to the notion of “direction of fit” (Searle 1983). Indeed, while beliefs possess a mind-to-world direction of fit insofar as they aim at representing the world, desires possess a world-to-mind direction of fit for they aim at changing it so that it can fit them. Their content represents the way according to which this must be done, which explains how they can be motivating, at the difference of beliefs.Footnote 6

Stueber’s account of empathy works smoothly in cases where all we have to do to understand an agent is to fill up our imagination with the relevant pretend-belief and pretend-desire and run the simulation routine. Suppose we run into a man who, crouched under his car, is stretching his arm to grab something under the vehicle. We can easily make sense of his behavior if we know that he wants to get the keys of his car back and that he also believes that the keys fell under it. By reenacting these thoughts in our own mind, that is, by personally engaging in a mental simulation where we pretend to have the same beliefs and desires as him, we are able to grasp them as providing reasons for the man to behave as he does. Even though we would not act in the same way—maybe because our cranky knees could not allow us to crouch as he does—we would still be able to see these thoughts as speaking for his behavior.

As we will see though, this model cannot adequately account for other forms of agency that, while being intelligible to empathizers as stemming from motivating mental states, are nevertheless irreducible to a belief-desire couple.

2.2.2 The irreducibility of arational actions explanation

The Humean model has been particularly criticized for proposing an inadequate view of some particular kinds of intentional actions. We are sympathetic to such criticism and believe that this has important consequences on how we conceive of empathy. What we want to show is that it is possible to find paradigmatic cases for which, in order to understandFootnote 7 an agent’s behavior, the Humean view is unsatisfying.

A major difficulty for this model is the case of arational actions, which was introduced by Hursthouse (1991). Such actions are for instance jumping or clapping one’s hands out of joy, kissing a photograph out of love, covering one’s face out of shame or tearing up a photo of a hated one out of jealousy. Hursthouse called such actions arationalFootnote 8 because they cannot be explained in terms of reasons, if the latter are understood through the lens of the Humean view. Her strategy is to prove that these actions can only be explained referring to the corresponding emotion: “in the grip of the relevant emotion, the agent just felt like doing them” (1991, p. 61).

Before we can move further in our discussion, a point needs to be clarified. Arational actions are just one kind of actions performed in the grip of an emotion. Emotional actions can sometimes happen to be rational as well. For instance, running out of the house out of fear because the house is on fire is a rational action. Indeed, it can be evaluated as such because it relies on a justified reason (in this case, the fact that the house is really on fire). Following Döring (2003), we can call these actions rational emotional actions. Hursthouse mentions these cases but does not tackle them in detail. She prefers to focus on the case of arational actions, which are more enigmatic since they do not seem to obey to any principle of rationality. We will focus on them in this section and tackle the case of rational emotional actions in the next one.

Hursthouse gives three conditions to define arational actions (1991, p. 59):

  1. (1)

    that the action be intentional;

  2. (2)

    that the agent does not do it for a reason expressed in the form of a belief-desire couple. That is, the agent does not accomplish the action holding the instrumental belief that it will satisfy the associated desire;

  3. (3)

    that the action would not have been done if the agent had not been in the grip of the emotion that caused it.

To prove her point, Hursthouse presents different possible explanatory desires and shows that they cannot be associated with any adequate belief. Let us take one of her most quoted examples: “Jane (…), in a wave of hatred for Joan, tears at Joan’s photo with her nails, and gouges holes in the eyes” (Hursthouse 1991, p. 59). Humeans can try to explain this action saying that Jane truly desires to harm Joan and that she believes that tearing a photo of Joan will enable her to do so. This explanation has to be rejected for it postulates the existence of an absurd instrumental belief, namely the belief that scratching the photo will effectively harm Joan.

Hursthouse gives two other possible belief-desire couples to show that both fail to explain these actions. The first one is that Jane desires to express her emotion and believes that tearing Joan’s photo is expressing it. Hursthouse shows that attributing such a belief does not work, for it implies that the action can be submitted to an evaluation in terms of correction. As we cannot say that the agent correctly (or wrongly) expresses her emotion in this case, this belief does not apply here.Footnote 9,Footnote 10 The second one is that Jane desires pleasure and believes that tearing Joan’s photo will be pleasurable. Such a belief is also implausible because it is hard to see why tearing a photo would provide any pleasure. If such an action may provide some relief, this is different from saying that it is intrinsically pleasurable. By comparison, eating sweets can be a pleasurable experience, though it is not obvious that the satisfaction of such a desire would always be pleasurable, as in the case of eating sweets compulsively.

Therefore, the only sensible solution is to concede that Jane’s action is expressive of a certain desire, which is wanting to perform the action in the grip of the emotion. It would be mistaken to add any further purpose to this action.

2.2.3 More sophisticated responses

At this point, Humeans may want to play their last card. Goldie (2000) argues that arational actions can fit the belief-desire model and proposes to this end a more refined explanation. To him, attributing to Jane the desire to tear Joan’s photo out of hatred, as well as the belief that the photo represents Joan, is not intelligible in the light of the emotion of hatred alone. In his terminology, this desire is not “primitively intelligible”. A further explanation must be introduced to complete it: Jane must also have the desire to harm Joan.

To explain how such a desire can lead one to feel relieved by tearing a photo, Goldie argues that this desire must take the form of a wish. He takes it to be something that one desires, and whose satisfaction involves an act of imagination. “In the sense in which I will use the term, when I wish for something, I desire that thing, and I also imagine, or I am disposed to imagine, the desire to be satisfied” (2000, p. 129).

Therefore, Goldie attributes to Jane a wish that takes the form of a desire to harm Joan coupled with the imagining that tearing Joan’s photo will enable her to satisfy it.

To connect both parts of the explanation and make both desires (harming Joan and tearing the photo) intelligible, Goldie brings the explanation to a symbolic level. Knowing that harming Joan personally is not possible due to social constraints, Jane would satisfy her desire symbolically by tearing a photo of Joan. The symbolic relation is guaranteed by the fact that the photograph truly represents Joan.

Two worries can be raised here. First, we agree with Kovach and De Lancey (2005) that this view is quite incomplete. Indeed, it is not clear at all that the wish that Goldie postulates should be motivating and thus explain the occurrence of such an action. We are therefore left in need of further explanation.Footnote 11 Second, since this explanation involves two kinds of desires, we cannot see how it would fit the procrustean bed of the Humean model, which only postulates a unique desire associated with an instrumental belief.

2.2.4 Is Humeanism all there is to say about Stueber’s agential empathy?

Before proposing our solution to the issues raised above, an important clarification is needed. It would be uncharitable to attribute to Stueber a reductionist view where reasons are nothing but the result of the relevant belief and desire couple.Footnote 12 Indeed, while most of his arguments are focused primarily on the recreation of “Humean” reasons, in some passages of his work he acknowledges that it is possible to recreate through empathy mental states other than beliefs and desires, such as emotions (2006, p. 161). This is due to the fact that emotions can contribute to our rational agency and may thus be evaluated in terms of rationality. Stueber agrees in this sense with De Sousa (1987) and Damasio (1994). To him, RE plays a crucial role in enabling us to grasp the reasons that may explain them. For instance, it is by putting myself in the shoes of a person who has just missed her train that I will be able to understand why she is so angry, and why this anger is much more important than in the case of a person who has missed it being late for a few hours (Stueber 2006, p. 161; Goldman 1995). Imagining myself in the other person’s shoes allows me to understand what aspects of the situation are likely to trigger this anger, that is, what features of the target’s perspective provide reasons for her anger:

As Aristotle correctly and repeatedly pointed out, the ability to respond in a specific situation in such a manner presupposes that one is emotionally attuned to the world in the right manner. […] We ordinarily judge an emotional response to be appropriate or inappropriate; we recognize, as Aristotle put it, that it is appropriate to feel angry at the right time, the right place, and toward the right person. We indeed take it to be a good reason for you to be angry if somebody intentionally steps on your foot to hurt you. But we would not find this reaction appropriate in the case where somebody steps on your foot by accident. (2006: 160–161)

In this passage Stueber states that our emotional states can also be evaluated according to principles of rationality as far as we can see their occurrences as instantiated by some state of affairs that—at least partly—motivate them. This is certainly true, but we do believe there is still a story to be told about how emotional states shape and motivate our behavior and how empathizers are able to re-enact these mental states to adequately understand their efficacy in other people’s behaviors and mental life more generally.

Stueber’s model works smoothly in cases that nicely fit the Humean framework. When it comes to cases that do not easily fit it, such as arational actions, it reveals itself as defective. So, if we want to include emotions within the scope of RE, a few considerations need to be introduced. The next section will develop them.

3 Towards an experiential conception of empathy

We now want to prove that Stueber’s model is not sufficient to explain how empathizers are able to reenact other agents’ perspectives in the case of emotional actions, and, therefore, empathically understand the behaviors they instantiate. Indeed, the only way to empathize with agents performing these actions is to achieve insight into the phenomenological dimension of the corresponding emotions, a point that Stueber largely neglects. What is needed to make sense of emotional actions is not only to understand why an agent acted the way she did but also to understand how she was motivated to do so. To this end, we will introduce some considerations in emotion theory, which will enable us to explain the specific role of emotions in RE and more precisely, to account for the role played by their phenomenal component. Starting from the case of arational actions, we will generalize these conclusions to the case of emotional actions.

3.1 From arational actions to experiential empathy

3.1.1 Emotions as affective perceptions

The first comprehensive attempt to understand the role of emotions in explaining arational actions is found in Döring (2003, 2007). Following Hursthouse’s view and against the Humean model, she proposes that emotions be conceptually primitive: arational actions explanations, as we saw, are irreducible to the ascription of belief-desire couples. They can only be explained by referring to the corresponding emotions. Examining how she proves her point will enable us to highlight the fundamental role of the phenomenal dimension of emotions in this respect.

Döring counts among the proponents of a perceptual model of emotion according to which emotions can be regarded, more or less literally, as a specific kind of perception.Footnote 13 This affective perception provides us with a sui generis access to evaluative properties, or values, just like perceptions provide us with access to objective properties such as colors or spatial properties.Footnote 14 The emotional representational content is therefore revealed non-inferentially through the emotion’s conscious subjective character, i.e. through its phenomenal dimension (2003, p. 226). Let us continue with the case of Jane tearing Joan’s photo out of hatred. Döring’s view amounts to saying that Jane’s emotion consists of the felt representation of Joan as an obnoxious person. This representational content takes a non-judgmental form.Footnote 15 Bringing us into contact with an aspect of the outside world that is revealed to us non-inferentially, emotions can justify evaluative judgments, just as perceptions are able to inform our empirical knowledge. The intelligibility of emotions is thus not primitively based nor constituted by the presence of any evaluative judgment.Footnote 16

Now, how can an emotion explain the corresponding arational action? Döring proposes that “explaining an action by emotion means specifying how the emotion’s affect relates to its representational content in causing the action” (2003, pp 223–224). That is, in Jane’s case, one has to specify how the fact that she perceives Joan as obnoxious is related to the phenomenal aspect of her emotion of hatred that causes her to tear Joan’s photograph.

Explaining this point comes back to investigating the nature of the motivational dimension of emotions. The phenomenal aspect of emotions plays a crucial role here. “Emotions are capable of motivating because their representational content is at the same time felt, i.e., because they are affective perceptions” (2003, p. 224). Said otherwise, the very felt phenomenology of emotional states compels us to feel motivated to act out of it. For instance, the distinctive phenomenal character of fear most of the time impels us to move away. Feelings are an essential component of emotions, they constitute “the emotion’s conscious subjective character, i.e., from what it is like to experience, or rather to feel, the emotion” (2003, p. 226).

Perceiving affectively the dangerousness of a situation necessarily comes with a tendency to act, differently from the case of a cold and affectless perception of this situation. This motivational push is independent of any desire. The existence of arational actions provides a decisive argument to prove it. Arational actions are characterized by the fact that they do not motivate the subject to act in accordance with the representational content of the emotion. In this case, the fact that Jane, through her emotion, perceives Joan as obnoxious, should motivate her to harm Jane directly. However, she limits herself to tear Joan’s photo. That is, her emotion fails to provide an end to the arational action. From this, Döring concludes that emotions cannot be composed of any desire. Her argument relies on the fact that in the Humean view, desires are conceived as having a world-to-mind direction of fit, which explains how they can be motivating. From the fact that emotions constitute a homogeneous category, she deduces that this conclusion must extend to all emotions. That is, the motivational component of emotions is sui generis and cannot be equated with a Humean view of motivation. Emotions motivate us to act independently of any desire that we may hold.

If we accept that arational actions cannot be explained in terms of desires providing an end to action, then Döring’s argument against the Humean view of actions is valid. It makes indeed no sense to introduce a fragmentation within the emotion category based on their capacity to be rationalized, since the same emotion, such as anger, for instance, can lead to arational actions (kicking against a wall) as well as to rational actions (fighting with someone).Footnote 17

3.1.2 Introducing experiential empathy and phenomenal insight

The next step consists in examining the consequences of these theoretical considerations on the model of RE. In what sense are emotions involved in RE? Can the phenomenal dimension of these experiences also play a role when we seek to recreate through our imagination the mental states that accompany arational actions?

From what we said above, the agent has to recreate the corresponding emotion to be able to empathize with an agent performing an arational action. More precisely, it is by recreating in her mind its phenomenal component and imagining how it can compel her to act that she can understand why someone in the grip of an emotion behaved as she did. That aspect is what we call “phenomenal insight” to account for the capacity of empathy to convey understanding about the experienced dimension of mental states. It is because we are aware of what it feels like to be in the grip of an emotion that our empathizing will be successful.

An empathizer who sees Jane tearing Joan’s photo but who has never been herself in the grip of an emotion of anger cannot understand Jane doing this apparently purposeless action. It is because the interpreter can imagine what it feels like to be in a state of anger that she is able to understand why Jane tore Joan’s photo. Empathizing with Jane provides her with a phenomenal insight into the way anger can motivate one to act arationally. To quote Johnston (2001), in the case of emotional actions, the presence of an affective state can make the occurrence of an “action especially intelligible to the agent himself” (189)—and, we would add, to the empathizer as well. This is what constitutes according to him the “authority of affects”.

More generally, we call experiential empathy the very dimension of empathy in virtue of which the empathizer is led to pay attention to the phenomenal aspect that accompanies the mental states of the agent with whom she empathizes. In this sense, empathy does not only aim at providing us with an insight into the reasons that motivate the behavior of the agent being interpreted. It also aims at providing us with a dive into the very felt experience that accompanies them. Experiential and agential empathy cooperate in the case of arational actions to provide us with an exhaustive view of the mental states of the target. Moreover, we maintain that experiential empathy is a requisite for agential empathy to be epistemically successful. Missing that aspect, that is, the authority exerted by the felt dimension of affective states, would deeply impair the accuracy of our understanding the other agent’s reasons for acting in such a way.

3.2 Is experiential empathy needed to explain rational emotional actions?

To conclude this section, let us show that our position extends more generally to the case of rational emotional actions, which constitute the other kind of emotional actions. This point is important because, as we will show below, rational emotional actions can apparently be re-enacted independently of experiential empathy. Developing this point will enable us to stress on the necessity to articulate AE and EE when rational agency is at stake.

Prima facie, the reenactment of these actions seems to be marked by the seal of indeterminacy. Indeed, it is not clear (1) whether we have to resort to emotions or to a Humean view to explain them, and (2) whether the experiential component of empathy should be introduced for AE to succeed.

As regards point (1), the reasons which supposedly explain such actions can indeed take the form of a belief-desire couple. For instance, let us take the case of Tom, who runs out of his burning house after seeing some heavy smoke in his kitchen. Within the Humean agential framework, we will understand him by reenacting his desire to protect himself as well as his belief that running away from the house is the best means to this end. Since desires are traditionally viewed as motivating states, there does not seem to be any principled opposition for holding such a position.

We believe, however, that this explanation is unsatisfying in such cases. Indeed, it does not adequately explain how the agent is motivated to act and, furthermore, it does not do justice to the processes involved in RE either. Indeed, nothing enables us to distinguish the reenactment of this situation from the cases when we try to understand an agent reproducing the same actions but who acts coldly—maybe walking out of her house, instead of running—, or from the cases where the interpreted agent holds the same beliefs and desires, yet without being motivated to act.

Now, the proponents of the Humean model may reply by constructing a more sophisticated desire, such as saying that Tom desired to protect himself from what he believed to be the start of a fire and hence a dangerous situation.Footnote 18 Even though more informative, this explanation is not sufficient either. What is missing again is that this explanation does not enable us to grasp how attributing this desire to the interpreted agent can be enough to account for his motivation to act. Moreover, it would be strange to say that an external agent interpreting Tom would mainly reenact the beliefs and desires that would explain the action, substantially omitting the relevant emotional states.

In other words, it is not because I am able to simply recognize that Tom is in a dangerous situation that I will be able to understand why he was motivated to act in the way he did. Indeed, we often make evaluative judgments without being inclined to act accordingly. For instance, let us imagine that I am on a hiking trail with a friend. In front of us is a rickety bridge. While recognizing that the situation is dangerous, I resolutely begin to walk on the bridge. Surprisingly, my friend decides to retrace her steps. We can say that fear leads her to act so as to avoid a situation that she perceives not only as dangerous but also, to be avoided. Said otherwise, the fact that my friend’s perception of the situation is affectively charged motivates her to act accordingly. Appealing to my friend’s desire not to be harmed by such a dangerous situation would not do the work either, because we can reasonably ascribe this desire to me as well, without this motivating me to avoid the bridge.

At the same time, this last point also enables us to understand why the agential dimension of empathy alone is not sufficient in cases of emotional actions. Indeed, the mere fact that I am able, as an empathizer, to recognize reasons, does not entail that I will find them motivating. This is also the case when I reenact the corresponding emotion, as Stueber is inclined to think. Reenacting an emotion within an agential framework means above all that we ascribe to emotions the cognitive role of guiding our reasons attribution. This view raises two concerns. First, in the light of the explanations given on the role of the phenomenal component of emotions, the agential dimension of empathy needs to be complemented by the experiential one to account for the way in which emotions provide us with reasons for acting. Second, merely knowing those reasons, without being able to ground them in the emotional experience they stem from, is not enough to explain how an agent may be motivated to act accordingly, as showed just above.

Putting the pieces of the puzzle together leads us to conclude that reenacting emotions and, more precisely, reenacting their phenomenological component, is the only way to make the corresponding actions fully intelligible.

Reenacting a mental state such as an emotion to understand how it can motivate an agent to act goes beyond reenacting it as if it were a propositional attitude that could plainly be decomposed in a belief and a desire. Stueber problematically flattens the significant phenomenological differences between those mental states, without doing justice to their specificity. In his model, recreating an emotion in our mind and attributing it to others seems to take the exact same form as recreating the corresponding beliefs and desires that can explain it. As we saw before, though, beliefs and desires on one side, and emotions on the other, do not play the same role in explaining emotional actions.

All in all, we can conclude that agential and experiential empathy work together within our practice of RE. The weight that is granted to each of them depends on the kind of mental state that we re-enact. Stueber insists that AE is the default view of RE. This can be accepted in the case where we empathize with an agent performing an action that is rationally evaluable. The epistemic goal of empathy is here turned towards reasons. When the action is an emotional one, we have shown that AE may still be a successful folk-psychological strategy, but only if properly integrated with EE. In fact, in the case of emotional actions, to be motivated to act is above all to feel motivated to act. That is, we have to reenact the phenomenal dimension of emotions in this case. Stueber’s view is therefore insufficient (for it does not truly account for the motivational push of emotions) and sometimes fallacious (because it mitigates and misconstrues the role of emotions in RE).

4 Experiential empathy beyond the scope of agential empathy

As already anticipated, Stueber restricts the scope of empathy to the realm of agency. In this last part, we want to go beyond the scope of AE to propose a second criticism of Stueber’s model. Empathy’s capability to provide us with phenomenal insight into other people’s mind (EE) is indeed not only necessary to properly account for the empathic understanding of emotional actions but can constitute an epistemic goal on its own. In what follows we will show that there are cases where empathy’s main epistemic goal is simply a matter of achieving phenomenal insight.

4.1 Phenomenal insight without agency

As largely acknowledged until this point, it is certainly true that in some sense there are always reasons for us to act. Stueber, in particular, takes motivating reasons to be the necessary ingredient of our folk-psychological (i.e. empathetic) understanding.Footnote 19 These reasons can be reenacted through empathy. In the absence of such reasons, it follows, there would be no space for empathy.

Such a view seems to be perfectly plausible indeed. Imagine we happen to see our friend suddenly starting to run towards the bus stop in the middle of the conversation. After a brief moment of disorientation, we realize that the last bus of the night which is supposed to bring him home is inexorably approaching the stop. We can easily understand his behavior: he believes that this is the bus he needs to take to go home, and he wants to go home. We can comfortably reenact these thoughts in our own mind and therefore grasp the reasons why he started to run in the middle of a conversation. By reenacting our friend’s thoughts, we do not only see his beliefs and desires as causally explaining his behavior, but we are able to represent the “first-personal push” these mental states produce on a subject in shaping his subsequent actions. This case perfectly fits the Humean view on agency, where it is not even necessary to resort to any particular emotion to satisfactorily explain and understand the action. We already discussed it in the previous sections. What is important to state once again is that, according to Stueber’s view, we need to be able to detect and reenact the target’s relevant motivating reason in order for empathy to be successful.

Imagine now a different scenario where, in the middle of a conversation, during a nice walk on a Sunday morning around the city, the same friend suddenly starts to run without any apparent reason to do so. He did not express any desire to go home: there is no bus stop he could be trying to reach, no eventual threat in the surroundings, etc. In this case, in virtue of missing our friend’s reasons for acting, we will undoubtedly fail at empathizing with him and therefore at understanding his behavior. We will lack any supporting evidence in light of which we can elucidate his motivating reason for starting to run.

With a little help from our imagination, we might even conceive a possible world in which our friend starts to run because of the spell of some evil witch. Suppose we knew that every Sunday morning, the spell causes him to start running against his will. In such a case, we cannot understand him in terms of any motivating reason, but we can find an explanatory reason. His legs still started to move for a reason. According to Stueber’s view, although we are in the position to satisfactorily explain our friend’s behavior, empathy would still be an unavailable strategy, since it is lacking a suitable motivating reason to help us reenact our target’s perspective. It follows that when confronted with cases of unintentional behaviors, the empathetic perspective-taking will simply be an unusable strategy.

One might have some intuitions that resist such conclusions. If we want to be faithful to an idea of empathy as imaginative perspective-taking, then it is easy to see that in someone’s subjective perspective there is much more than motivating reason and that, furthermore, empathy’s scope goes far beyond the mere understanding of agency. That is, we have to move beyond the agential conception of empathy, towards what we have called an experiential one, namely, one which is concerned with the experiential dimension of other agents’ perspectives.

There is indeed, as the Nagelian philosophical mantra goes, something that it is like to have an experience. More specifically, there is something that it feels like in many unintentional behaviors. We are sure that, in the imaginary world, there must be something that our poor friend would feel when he shockingly realizes that his legs are moving without any conscious deliberation of his. There is a proprioceptive dimension in the very movement of the legs in the run and there is a subjective way to relate to those experienced bodily states.

We can take a less fanciful (and hopefully more meaningful) example and imagine the case in which we see our friend being pushed by a bully with a bad attitude. Of course, the push feels in some way to our friend and we can imagine and to some extent empathize with that sensation, that is, we imagine such an experience from his perspective even though it is clearly not a case of intentional action.

We can have some grasp of these experienced features of our lives. Part of the work of empathy is precisely to contribute to this peculiar epistemic achievement that we have called phenomenal insight. Phenomenal insight seems to be an epistemic achievement of empathic perspective-taking that can in principle be available independently of reasons. When trying to achieve phenomenal insight, the empathizer is not concerned with the reenactment of the cognitive chain that led to the performance of an action, but in the very subjective dimension of the experience as it is lived by the target. She is not simply interested in agency and deliberation but also in modeling and representing, through imagination, the subjective state in which eventually—but not necessarily—agency and deliberation have their roots. To use an expression which had some fortune among researchers on empathy and mindreading, to empathize with someone is to imagine his or her experience from the inside (Heal 1998; Coplan 2011; Goldie 2000, 2011; Paul 2017).

Given the above considerations, it is easy to realize how phenomenal insight is not only limited to unintentional behaviors. There is indeed something it feels like to undergo most of our experiences. Phenomenal insight is usually an overlooked aspect of perspective-taking in general. Phenomenology is nonetheless an essential component of our experiences, and, surprisingly, numerous instances of perspective-taking precisely aim at phenomenal insight. It is mostly what happens in our experience of reading fiction. When enjoying a novel, we are not only interested in the entrenchment of thoughts and feelings of the protagonists but also, we are trying to recreate their experiences in our imagination from a first-person perspective. Take for example this passage:

These nightmares recurred again and again, till he was afraid to go to sleep. He would lie stretched on his bed, sometimes the victim of obstinate fits of insomnia and feverish restlessness, at others of abominable dreams only interrupted by the spasmodic awakening of a man losing foothold, pitching from top to bottom of a staircase, plunging into the depths of an abyss, without power to stop himself.

For several days, the exhausting nervous disturbance gained the upper hand again, showing itself more violent and more obstinate than ever, though under new forms.

Now the bedclothes were a weight not to be borne; he felt stifled under the sheets, while his whole body was tormented with tinglings; his thighs burned, his legs itched. To these symptoms were soon added a dull aching of the jaws and a sensation as if his temples were confined within a [vise].

His distress of mind grew more and more acute, but unfortunately the proper means of mastering the merciless complaint were lacking. (Huysmans, Against Nature 1884/1931: 106)

When reading this description, we are not simply interested in knowing that the main character is having trouble sleeping. What we do is imagine these states through a first-person perspective so that we somehow achieve an experiential understanding of his experience. But this is not limited to the enjoyment of literature. In our everyday life, we are sometimes interested in this phenomenal insight into other people’s mental life.

We can take a quite common real-life experience which certainly has a distinctively salient subjective dimension: pain. This example will also show us how phenomenal insight can be achieved quite independently from reasons explanations. When we see a friend screaming in pain while holding her hand, we usually do not empathize to reenact her reasons for screaming and holding her hand. It is usually pretty obvious that the feeling of pain was the reason behind that.Footnote 20 What we might want to empathize with is rather the very felt experience of pain from our friend’s point of view. Interestingly enough, while there might not even be any reason for our friend to feel pain in her left hand, we could in principle—assuming that she is not lying—still reenact the sense of physical discomfort and psychological disarray that accompanies every true experience of pain. By stating this we do not want to claim that reason-explanations are of no-use to reach phenomenal insight. Being acquainted with the reasons behind a certain kind of subjective experience might still endow the empathizer with a richer insight into the perspective of the other person. Coming back to the experience of pain, knowing the reason behind an experience of pain can contribute to imagining more accurately the felt dimension of that experience. There is in fact a difference between knowing whether our friend’s hand was hurt by being accidentally hit by a hammer or cut by a knife.

Even more, when dealing with phenomenal insight, the whole talk about reasons can sometimes be misleading as it may skew our intuitions in an epistemically unproductive way. Indeed, it can imply that the empathic process aims at reasons reenactment, which would be misconstruing the empathic process. Moreover, it would mean that reasons are a necessary condition to empathize, as well as it suggests that elucidating them would always be epistemically relevant. We may doubt that reasons should always possess a positive epistemic role in the empathic process, as they can sometimes orientate the interpretation in a way that is not true to the phenomenal insight. Therefore, one might prefer to talk about getting the relevant contextual cues shaping and informing the particular experience one is trying to empathize with. For example, knowing why I am angry (I have been pushed by a stranger or by a close friend) and which person I am angry with (a stranger in the metro, my close friend, etc.) can enhance my imaginative power as an empathizer in better representing the target’s experience. Getting to know the relevant information about the experience we are trying to reenact in our imagination can contribute in carefully shaping with more accuracy an empathic reproduction of the other person’s experience.Footnote 21

4.2 The epistemic reliability of phenomenal insight

At this point, we would like to conclude our essay by briefly addressing an urgent epistemological concern about EE. The considerations that follow should by no means be taken as exhaustive, but as a tentative conceptual framework for the epistemic reliability of EE that could be enriched by further work on the topic. The epistemological concern we have in mind is that of the accuracy conditions of phenomenal insight: what are the prerequisites, all things being equal, for an accurate representation of someone else’s subjective experience? For instance, how are we able to represent someone’s pain as it phenomenally appears to a conscious subject?

The epistemic ground that allows empathizers to bridge the gap between them and their targets is represented by what we might call their experiential competence. By this formula we mean that one’s own set of past experiences can fill our imagination with phenomenal properties of a specific experiential occurrence. If we have once experienced the taste of coconut, we are endowed with the representational capabilities that allow us, among other things, to imagine what it feels like to taste coconut.Footnote 22 Of course, the necessary precondition for such an experience to work as experiential competence is that we retain a representation of it in our memory. It is rather obvious that if we had a specific experience we did not retain any traces of in our memory, we will not be able to add that experience in our set of experiential competence.

An example here might help: suppose I happened to have chickenpox when I was 3 years old. Chickenpox then might surely figure as an experience that I personally underwent. As an undeniable proof stands the fact that I am now immune to chickenpox. And yet, since I was so young, I cannot recollect any aspect of that experience. As a result, I am not in the position to accurately represent what it feels like to have chickenpox.

Suppose now I contracted chickenpox at the ripe old age of twenty-seven. Given the absence of any pathological or other contextual impairments to my memory system, I retain at present very vivid memories of that disease. This endows me with the fine-grained epistemic capacities to accurately represent what chickenpox feels like: the unbearable itchiness, the maddening frustration of not being allowed to scratch yourself, etc. This retained information can now serve as experiential competence and enrich my empathetic capabilities.

On the basis of such an example we can draw the provisional conclusion that, in order to gain phenomenal insight about another person’s experience, we need to have the relevant experiential competence, viz. we need to have undergone experiences that are comparable to the one our target is undergoing, and to be able to summon them up. The similarity between our target’s experience and our past experience of the same type allows our imagination to be filled with the adequate phenomenology of a certain experiential occurrence.

As intuitive as these considerations may sound, they are still far from being uncontroversial. One might in fact raise the issue as to how similar the target’s and the empathizer’s experiences need to be in order for the latter to legitimately aim at empathic accuracy: is it necessary for the empathizer to have undergone exactly the same kind of experience? Coming back to our chickenpox example: does the empathizer really need to have contracted the same disease or can she work out what it feels like to have chickenpox without having contracted it herself? Indeed, she might surely have experienced some occurrences of itchiness and some others of frustration even though they maybe did not happen both at the same time and for the same reason as in the case of chickenpox. In this latter case, it seems that there would still be space for the empathizer to understand what it feels like to have such a disease even though she did not experience it as such.

Of course, that would be cognitively costlier because instead of recollecting the experience as such from our memory, we would have to somehow “re-assemble” it in our imagination, so to speak, from scratch. Nevertheless, this would amount to a mere pragmatic difficulty, rather than constitute a principled impossibility. The above considerations are far from being purely speculative but are in complete harmony with some recent findings in the neuroscience of memory. In fact, our capacity for episodic memoryFootnote 23 seems to be subserved by a larger system in the brain that is also recruited for a variety of purposes such as imagination, dreaming and social cognition (Schacter et al. 2007; Hassabis and Maguire 2007, 2009; Spreng 2013; Mahr and Csibra 2018). The contemporary literature in philosophy of mind and cognitive science is immensely enriched by a lively and intriguing debate about the nature of episodic memory, e.g. what distinguishes it from other forms of remembering, how we should conceive of the relationship between memory—more specifically, episodic memory—and imagination, just to mention a few issues.Footnote 24

Without committing ourselves to any strong account of episodic memory, we can simply draw a lesson from this debate: information stored in our memory (i.e. memory trace) can be flexibly recruited for a variety of purposes, one of which—not the exclusive one—is episodic memory. That implies that the same contents that figure in episodic memory can be recruited and flexibly recombined in imagination, and, more specifically, in empathic imagination. Being the result of past experiences that we personally underwent, memory traces can vehicle information about the experiential features of a specific event, i.e. they can convey information about what it feels like to undergo a certain kind of experience. Memory traces of our past experiences can therefore be recruited in order to serve as “raw material” on the basis of which we can calibrate our imaginative simulation of the other person’s perspective.

These brief considerations seem to suggest that the success conditions of empathy heavily rely on the capability for the empathizer to resort to the relevant memory trace that can “match” the perceived specificity of the other’s perspective. It is precisely this match between the empathizer’s experiential competence and the target’s actual experience that allows phenomenal insight to hit its target.

What we have said so far, though, should by no means be taken as if empathic understanding amounted to nothing but the mere recombination of our previous experiences projected onto our target. What we are defending instead is the much plainer and more intuitive claim that imagining other people’s experiences is enriched and partly constrained by our own past experiences and, more precisely, by what we call experiential competence. Indeed, the claim that our empathic capabilities are constrained by what we have personally experienced plus our ability to retrieve it, and the claim that our empathic understanding reduces to what we have experienced in the past, should carefully be distinguished and kept aside in order to avoid conceptual confusion. The flexibility with which stored information of our past experiences can be manipulated and recombined allows us to produce empathic outputs that could not be plainly reduced to the experiential competence we possessed before engaging in the empathic process. That is what ultimately grounds what we might call empathy’s epistemical generativity, viz. empathy’s capability to produce some new knowledge about other people’s experiences we did not personally undergo. Said more clearly: modeling our empathic imaginings alongside our target’s experiences is in fact what allows our representational capability to stretch and recombine our experiential competence so that we can achieve some new knowledge. Acquaintance with our targets’ experiences can enable us to represent experiential features for the first time, in an inedited fashion.

A concrete example might help us illustrate this point more clearly. Suppose that I never felt a joy so intense to drive me to tears. That implies that I lack the specific experience of crying out of joy. Still, I experienced joy and I cried sometimes. My acquaintance with cases of people crying out of joy provides me with some evidence that allows me to notice in a new fashion a personally unexperienced feature of joy. This external acquaintance enables me to understand how it is possible to experience a joy that is so intense to lead to tears and what that experience might feel like: I can flexibly recruit my experiential competence in order to produce an empathic output that enables me to combine together the powerful impact of joy and the personal unsettlement of the experience of crying.

Of course, as already stated, the fact that we can flexibly manipulate our experiential competence does not mean that it is completely unbounded. The complete lack of the relevant type of experience would imply a total empathic blindness: to resort to a classic example in the philosophical literature, it is impossible for Mary to imagine what seeing red feels like, if she has never seen red—or any other color—in her life (Jackson 1982; Paul 2014, 2017).

Still, most of the time, neurotypical adults are provided with a broad experiential competence so that in most cases, there will be space for empathy.Footnote 25 As we saw, flexible reuse of our memory traces and acquaintance with the specificity of our target’s perspective can in fact enable us to push our empathic capacities beyond what we already know about ourselves.

5 Conclusion

In the present essay, we hope to have shown a unique and often neglected epistemic contribution of empathy, namely its capability providing subjects with phenomenal insight into other people’s mental lives. This aspect, besides being an indispensable ingredient of our folk-psychological understanding of emotional actions, can constitute in some occasions the main epistemic goal of empathy.