What Is Empathy? Three Different Things With One Name

The standard division in the literature separates the following:

• •Affective empathy, the sharing of another person's emotional state, so that seeing distress produces something of the same feeling, which is fast, automatic and present in infants
• •Cognitive empathy, also called perspective taking or theory of mind, the ability to model what another person is thinking and feeling without necessarily feeling it, which develops through childhood and is what a skilled negotiator uses
• •Empathic concern, or compassion, the motivation to act on another's behalf, which is not the same as feeling what they feel and is the component that predicts helping
• •The three dissociate: psychopathy is associated with intact cognitive empathy and impaired affective empathy, which is what makes manipulation possible, while autistic people frequently show the reverse pattern, with strong affective responses and difficulty inferring intentions
• •Personal distress is a fourth and unhelpful state, in which sharing another's suffering becomes self-focused and produces withdrawal rather than help

Imaging studies find that observing someone in pain activates parts of the same network that is active during one's own pain, particularly the anterior insula and anterior cingulate cortex, which are involved in the affective dimension of pain rather than in locating it on the body. Modelling someone's beliefs activates a different network, including the temporoparietal junction and medial prefrontal cortex, which is the system implicated in theory of mind tasks generally. The discovery of mirror neurons in macaque premotor cortex in the 1990s, which fire both when an animal performs an action and when it watches another perform it, generated an enormous literature proposing them as the basis of empathy, and the claims made were considerably stronger than the evidence, which concerns action understanding rather than emotion, and the field has since become much more cautious. Tania Singer's work distinguishing empathy training from compassion training found that they produce different patterns of activation, with compassion engaging reward-related regions rather than pain-related ones.

Empathy is not evenly distributed and its biases are systematic. It is stronger toward people who resemble us, share our group membership, or are physically present, which is the mechanism behind the observation that one identifiable individual generates more response than statistics describing thousands, an effect Paul Slovic has documented repeatedly and calls psychic numbing. It fades with numbers, so that reported willingness to help can actually decrease as the number of victims rises. It is reduced toward outgroups and, in some studies, reverses into pleasure at their misfortune. And it can be fatigued, which is a real occupational problem in medicine, social work and emergency services, where sustained exposure to suffering produces burnout, withdrawal and reduced quality of care. The professionals who last are generally those who develop compassion without absorbing the distress, which is the distinction between the second and the fourth components above.

Paul Bloom's argument, set out in a book deliberately titled against empathy, is that the affective component is a poor guide to moral decisions precisely because of those biases: it is innumerate, parochial, present-focused and easily manipulated, and it leads to decisions that feel right and help fewer people. A donor who gives to the case in the newspaper rather than to the more effective intervention is behaving exactly as affective empathy directs. His alternative is what he calls rational compassion, caring about others' welfare while making decisions with reasoning rather than with feeling. The counterargument is that compassion without any affective grounding is thin motivation, that the biases can be corrected by training and institutions, and that in close relationships and in clinical care the felt response is doing something reasoning does not. The productive part of the dispute is that both sides agree the components should be separated, since most of the disagreement comes from arguing about different processes under one word.

The evidence is better than for most interventions of this kind and more modest than the claims made for it. Compassion meditation training has produced measurable changes in behaviour and in brain response in controlled studies, with effects on helping in laboratory tasks. Perspective-taking exercises improve attitudes toward outgroups in the short term, and the durability is less clear. Reading literary fiction was reported in 2013 to improve performance on tests of inferring mental states, a result that has failed to replicate consistently. Medical schools teach communication and perspective-taking explicitly, after research found empathy scores declining across clinical training. What seems reasonably robust is that perspective taking is a skill that improves with practice and instruction, that compassion can be cultivated deliberately and is protective against burnout in a way that shared distress is not, and that the affective component is largely a disposition rather than a skill.

Empathy covers at least three processes: sharing another's feeling, modelling what they think without feeling it, and being motivated to help, and these run on partly separate brain systems and come apart in psychopathy and in autism in opposite directions. Shared feeling is biased toward the familiar, the present and the single identifiable case, and it fades as numbers rise, which is the basis of the argument that it is a poor guide to moral decisions. Compassion can be trained and protects against burnout, while absorbing distress does not.