We talk about empathy as though it were a fixed quantity. Something you either have or you don’t. A moral capacity that, once developed, operates consistently across contexts and people.
The research says otherwise.
Empathy is not a steady current. It fluctuates; with proximity, with familiarity, with perceived similarity. It spikes for people who look like us, live like us, grieve like us. It dims, quietly and often unconsciously, for people who don’t. This is not a character flaw distributed unevenly across bad people. It is a structural feature of how human empathy works, built from evolutionary architecture that was never designed for the scale and diversity of the world we now inhabit.
The empathy gap is not the absence of empathy. It is empathy working exactly as it was built to, and that is precisely what makes it so difficult to correct.
The Neuroscience of Who We Feel For
The evidence starts in the brain, which is where the comfortable story about empathy begins to unravel.
Neuroscientists studying empathic response have consistently found that our neural mirroring systems, the mechanisms that generate felt resonance with another person’s pain or pleasure, activate differently depending on perceived group membership. In studies using fMRI imaging, participants observing others in pain showed measurably stronger neural responses when the person in pain was perceived as belonging to their own racial or social group. The pain of the out-group registered, but more faintly. More abstractly. As information, rather than experience.
This is not a metaphor. It is a measurable difference in how the brain processes suffering depending on whose suffering it is.
Lasana Harris and Susan Fiske’s research on social neuroscience introduced the concept of infrahumanization, the tendency to implicitly attribute less fully human qualities, particularly emotional depth and complexity, to out-group members. This is not the crude dehumanization of propaganda. It operates subtly, below the level of conscious intention. People who would never endorse explicit prejudice nonetheless show implicit patterns suggesting they attribute less inner life to those perceived as different. Less nuance. Less pain. Less humanity, in increments too small to feel like discrimination.
The philosopher Simone Weil wrote that attention is the rarest and purest form of generosity. The neuroscience suggests that attention itself is rationed and the rationing follows the contours of perceived similarity.
The Empathy Gap in Medicine, Where the Cost Is Measured in Bodies
Nowhere are the consequences of the empathy gap more concrete, or more lethal, than in medical care.
The data on pain treatment across demographic lines is consistent and disturbing. Studies conducted across multiple healthcare systems have found that Black patients are systematically undertreated for pain compared to white patients presenting with equivalent symptoms. The gap holds across conditions: post-operative pain, cancer pain, emergency room presentations. It holds even when controlling for insurance status, severity of condition, and the patient’s own self-report. And it holds, significantly, among clinicians who score low on explicit racial bias measures; which means the gap is not primarily a story about conscious prejudice. It is a story about attunement.
Women experience a parallel version of this. Decades of research document that women’s pain is more likely to be attributed to psychological causes, more likely to be undertreated, and more likely to result in delayed diagnosis for serious conditions including cardiac events. The infamous “gender pain gap” is not a result of women reporting differently. It is a result of clinicians receiving those reports differently.
The philosopher of medicine Havi Carel has written about the epistemic injustice embedded in clinical encounters; the systematic tendency to undercount certain patients’ testimony about their own experience. What Carel identifies as an epistemic failure maps precisely onto the empathy gap: the patient’s reported experience is filtered through the clinician’s felt resonance with it, and that resonance is not neutral. It is shaped by every dimension of perceived similarity and difference that the clinical encounter carries.
People are in pain. The degree to which that pain is believed, investigated, and treated varies by who is in pain. That is not a medical outcome. It is a social one.
How the Gap Shapes Justice and Who Pays
The criminal justice system is, among other things, an empathy allocation machine.
Research on jury decision-making, sentencing patterns, and prosecutorial discretion consistently reveals that empathic identification with defendants and victims tracks demographic similarity in ways the formal language of blind justice explicitly prohibits. Studies of victim impact statements show that jurors respond more emotionally to victims who share their background. Studies of sentencing show that defendants who present as culturally legible to the decision-maker receive different outcomes than those who do not.
This is the empathy gap operating at institutional scale. Not as individual prejudice, though that exists too, but as the accumulated effect of thousands of micro-decisions, each shaped by the degree to which the person making the decision can imaginatively inhabit the experience of the person in front of them.
Jennifer Eberhardt’s research on racial bias in the criminal justice system documented something particularly stark: the degree to which defendants were perceived as stereotypically Black predicted the severity of their sentencing, independent of the facts of the case. What was being punished, partly, was the activation of social distance. The less a jury could identify with a defendant, the less constrained their judgment became.
The concept of moral circle expansion, developed by philosophers from Peter Singer to Martha Nussbaum, holds that moral progress is partly a history of extending the boundaries of who counts as a full moral subject. What the justice research makes visible is that this expansion remains profoundly incomplete. The circle is larger than it was. It is not as large as the language of universal rights suggests.
Policy, Distance, and the Statistics That Don’t Move Anyone
There is a phenomenon well documented in psychology that has no clean solution: identified victim effect.
People donate more, advocate more, and feel more urgency about a single named individual in crisis than about a thousand unnamed ones. A photograph moves more people than a casualty figure. One child with a name and a story generates more political response than a policy that would save ten thousand children without names. The empathy system is calibrated for proximity and particularity. It was not designed to process abstraction.
This creates a structural problem for policy and collective action. The crises that most require broad empathic response, distant famines, systemic poverty, climate displacement, are precisely the ones least able to activate it. They are large, statistical, and geographically removed. They involve people who are, by every measure of perceived similarity, different from the populations making the decisions that affect them.
Paul Bloom, in his critique of empathy as a moral guide, argued that empathy’s partiality makes it an unreliable engine for justice. He is right that empathy, left to its own dynamics, will always preferentially serve those closest to the empathizer. Where his argument runs thin is in the assumption that rational calculation is a straightforward substitute. Reason without felt resonance has its own failure modes, and the historical record of purely utilitarian decision-making about distant others is not reassuring.
The problem is not empathy versus reason. It is that neither, operating in its default mode, reliably crosses the empathy gap. Something more deliberate is required.
Perspective-Taking Is Not Empathy But It Might Be More Useful
One of the most practically significant distinctions in the psychology of empathy is the difference between affective empathy and cognitive empathy; between feeling what another person feels and understanding what another person feels.
Affective empathy, the felt resonance, is the version most subject to in-group bias. It is harder to generate across difference, more easily exhausted, and more vulnerable to the implicit infrahumanization that dims the signal from out-group pain. Cognitive empathy, perspective-taking, the deliberate imaginative reconstruction of another’s situation, is more trainable, more sustainable, and less automatically constrained by similarity.
Research by Adam Galinsky and colleagues has found that structured perspective-taking exercises reduce bias in ways that affective empathy induction does not. The mechanism matters: you are not trying to feel what the other person feels, which the brain resists across difference. You are trying to understand how the situation looks from inside their position, which is a more tractable cognitive task.
This distinction has practical implications. Organizations designing bias interventions that rely on emotional empathy induction are working against neural architecture. Interventions that build structured perspective-taking skills, concrete, specific, situation-grounded, are working with it.
None of this is easy. Perspective-taking across deep cultural difference requires sustained effort and genuine intellectual humility. It requires accepting that your imaginative reconstruction of someone else’s experience will be partial, probably wrong in important ways, and still worth attempting. It requires treating the gap not as a failure of character but as a feature of cognition; one that can be worked with, even if it cannot be fully overcome.
What It Would Actually Take to Close the Gap
The empathy gap will not be closed by awareness campaigns. It will not be closed by being reminded that all humans suffer equally. It will not be closed by good intentions, which are among the least reliable predictors of empathic behavior across difference.
What the research points toward, instead, is contact, narrative, and structure.
Contact, sustained, cooperative, equal-status engagement with people outside your group, remains one of the most robustly supported mechanisms for expanding felt empathy. Gordon Allport’s contact hypothesis, developed in the 1950s and refined across hundreds of subsequent studies, holds that under the right conditions, contact reduces the social distance that underlies empathic disparity. The conditions matter: equality of status within the encounter, common goals, institutional support, genuine interdependence. Without them, contact can reinforce distance rather than reduce it.
Narrative does something different. It provides particularity, the named individual, the specific situation, the interior life made legible, for people the reader would never encounter directly. Literature, documentary, testimony: these are empathy technologies. They extend felt resonance across the gap of difference by giving the other person’s experience a shape the imagination can inhabit. This is not sentiment. It is a mechanism.
Structure matters because individual empathic capacity, however developed, is insufficient to override institutional defaults without institutional design. Medical protocols that flag demographic patterns in pain treatment. Sentencing guidelines that require explicit justification for departures. Hiring processes that audit for consistency across candidate demographics. These are not substitutes for empathy. They are guardrails for the moments when empathy fails, which, across the empathy gap, will be often.
The gap exists. It has always existed. The question is not whether human beings are capable of crossing it; the evidence suggests, conditionally and effortfully, that they are. The question is whether the systems we build treat that effort as necessary, or whether they are still operating on the comfortable fiction that empathy, like justice, is blind.
It isn’t. It never was. And the people on the wrong side of that gap have known it all along.
