How to Be More Empathetic (Emotional Empathy Is Systematically Biased — Cognitive Empathy Can Be Trained)

By Gwyndalyn Henderson

In 2016, Paul Bloom, a psychologist at Yale University, published Against Empathy — a book whose title was designed to provoke and whose argument was more precise than the title suggests. Bloom's case was not against caring about other people. It was against emotional empathy specifically: the experience of feeling what another person feels, of taking on their emotional state as one's own. His argument, built on a substantial research base, was that emotional empathy is systematically biased in ways that make it a poor guide to moral behavior and prosocial action. It defaults toward people who are similar to us, physically nearby, visually salient, and emotionally legible. It underweights people who are distant, numerous, statistically described, or different from us in background and experience. It produces compassion fatigue and burnout when it is practiced as emotional absorption rather than perspective-taking. And it correlates poorly with fair, consistent, or effective prosocial behavior compared to its cognitive counterpart. The research Bloom synthesizes points toward a different target: cognitive empathy — the trained capacity to understand another person's perspective, to accurately model their experience and reasoning, to see situations from their vantage point — which is more durable, more equitable, and more trainable than the feeling of emotional resonance that most empathy advice is trying to cultivate.

Bloom: The Systematic Biases of Emotional Empathy

Bloom's critique of emotional empathy centers on its predictable failures as a guide to action. The emotional empathy response — the felt experience of resonating with another person's distress — is not distributed randomly or proportionally to need. It activates most strongly for individuals who are similar to us in background, appearance, and experience; who are physically proximate or visually present; who are individually identified rather than part of a statistical group; and whose emotional states are dramatically legible. It activates least strongly for people who are distant, different, statistically described, or whose suffering is structural rather than acute.

The statistical lives problem Bloom describes illustrates this directly. Donations to identified individuals — a single child, a specific person with a named face and a story — dramatically exceed donations to equivalent or larger causes described statistically. Knowing that 30,000 people are dying of preventable disease per day produces less emotional empathy activation than seeing a photograph of one child who represents the same crisis. This is not a failure of caring. It is the architecture of emotional empathy: it responds to proximate, individual, legible distress and discounts aggregate, distant, or statistically represented suffering. For most of human evolutionary history, this architecture was adaptive: the people whose welfare was relevant to your survival were the ones nearby and similar to you. The architecture has not updated to match a world where the most consequential actions often concern people who are distant and different.

The similarity bias is equally well-documented. Research on prosocial behavior consistently shows that emotional empathy is stronger toward in-group members — people perceived as similar in ethnicity, nationality, political affiliation, age, or background — than toward out-group members, even when the out-group member's need is identical or greater. This is not a moral failing of empathic individuals. It is a predictable feature of how emotional empathy is generated: through automatic, pre-reflective resonance with people whose experience we can readily imagine as our own. People whose experience is less familiar generate less automatic resonance, which produces less emotional empathy, which produces less prosocial behavior — regardless of the actual level of need.

Two Types of Empathy: What the Research Distinguishes

The research on empathy distinguishes two meaningfully different processes that are frequently conflated under the single word. Affective empathy (also called emotional empathy) is the automatic resonance with another person's emotional state — feeling sad when someone else is sad, distressed when someone else is distressed, joyful when someone else is joyful. It is largely automatic, activates quickly, and depends on the emotional legibility and proximity of the other person. Cognitive empathy (also called perspective-taking) is the deliberate effort to understand another person's viewpoint, reasoning, experience, and situation — not to feel what they feel, but to accurately model how the world looks from where they stand. It is more effortful, more trainable, and less dependent on similarity and proximity.

The distinction matters practically for several reasons. First, cognitive empathy is less subject to the biases Bloom documents: because it involves deliberate perspective-taking rather than automatic resonance, it can be applied to people who are dissimilar, distant, or emotionally unfamiliar — the people for whom emotional empathy most predictably fails. Second, cognitive empathy is the component of empathy that predicts effective prosocial behavior, constructive conflict resolution, and relationship quality in longitudinal research, while affective empathy predicts empathic distress and burnout under high exposure conditions. Third, and most relevant for everyday application: cognitive empathy is a learnable skill with specific trainable components, while affective empathy is more dependent on individual differences in baseline emotional reactivity that are substantially heritable and less subject to deliberate cultivation.

Singer and Klimecki: Empathic Distress vs. Compassion

Tania Singer and Olga Klimecki at the Max Planck Institute for Human Cognitive and Brain Sciences conducted a series of studies directly comparing the neural and behavioral effects of empathy training versus compassion training. Their central finding challenges the default assumption that more empathy is simply better: empathy training — training people to resonate more fully with others' suffering — increased empathic distress and negative affect, and was associated with activation of pain circuitry in the brain. Compassion training — training people to combine perspective-taking with warmth and the motivation to help — did not produce empathic distress. It produced increased positive affect, prosocial motivation, and neural patterns associated with reward and affiliation rather than pain resonance.

Singer and Klimecki's research distinguishes empathy (sharing another's emotional state) from compassion (caring about another's wellbeing and being motivated to help without absorbing their state). The distinction is not semantic. It has measurable neural correlates: empathy activates anterior insula and anterior cingulate cortex — brain regions associated with pain processing. Compassion activates medial orbitofrontal cortex and ventral striatum — regions associated with reward and positive motivation. The person who resonates fully with others' distress and absorbs it as their own is drawing on pain circuitry, depleting their own regulatory resources, and generating the burnout pattern Singer and Klimecki observed in their empathy-trained participants. The person who maintains the perspective-taking component while coupling it with warmth and motivation to help is generating a neurologically different state that does not produce the same depletion.

This research has direct implications for anyone who has been told they need more empathy and has tried to develop it through fuller emotional attunement — only to find themselves more depleted, more distressed, and ultimately less able to be present for others than they were before. The target is not more emotional absorption. It is cognitive empathy coupled with compassionate motivation. These are trainable, neurologically distinct, and produce sustainable rather than depleting prosocial capacity.

Training Cognitive Empathy: What the Evidence Shows

The evidence that cognitive empathy can be trained through specific practices is well-established. Perspective-taking interventions — structured exercises in which participants deliberately adopt another person's viewpoint, identify their likely reasoning and concerns, and generate the strongest possible version of that person's position — reliably increase accuracy of prediction about others' mental states, reduce in-group bias in prosocial behavior, and improve conflict resolution outcomes. The mechanisms are specific: perspective-taking activates prefrontal cortex regions associated with mentalizing (modeling others' mental states) and temporarily suppresses the automatic in-group favoritism that emotional empathy produces.

Research on active listening as a cognitive empathy practice similarly shows trainable effects: the specific behaviors of reflecting back what has been heard, asking clarifying questions before evaluating, and restating the other person's position before responding — none of which require feeling the other person's emotions — produce measurable improvements in relational quality, conflict de-escalation, and the other person's experience of being understood. The experience of being understood does not require the listener to feel what the speaker feels. It requires the listener to demonstrate accurate understanding of what the speaker means. Cognitive empathy, trained through deliberate perspective-taking practice, produces that demonstration more reliably than emotional resonance, which generates shared feeling but not necessarily accurate understanding.

Quick Win — The Steel-Man Protocol

This protocol activates cognitive empathy on demand in the context of disagreement — the situation where the impulse to evaluate, defend, or counter is strongest and the capacity for genuine perspective-taking is most constrained. It takes less than five minutes and requires no special conditions. It can be applied in a live conversation, in preparation for a difficult exchange, or as a reflective practice after a conflict that went badly.

1. Identify the position you are preparing to disagree with, challenge, or dismiss. Write it down as you currently understand it — the version of the other person's position that is in your head before you have done any deliberate perspective-taking. This is the straw-man version: typically simplified, partially heard, framed through your own concerns rather than theirs.
2. Construct the strongest possible version of the position. Steel-manning is the opposite of straw-manning: rather than identifying the weakest, most dismissible version of a position, you identify the strongest, most reasonable, most internally coherent version. Ask: what evidence would a thoughtful, reasonable person cite in support of this position? What values or concerns does it reflect? What problem is it trying to solve? What would someone who holds this position say in response to the strongest objections to it? You are not trying to agree with the position. You are trying to understand it well enough to represent it fairly.
3. State the steel-manned version to the other person and ask if you got it right. "My understanding of your position is [strongest version]. Is that accurate? Is there anything I'm missing or misrepresenting?" This step is cognitively demanding and often surprising: the other person frequently corrects the steel-man, providing additional information that makes the position still more coherent. This correction is the perspective-taking update. What you learn from the correction is the cognitive empathy — the accurate model of how the situation looks from their side — that you could not have produced through emotional resonance alone.
4. Only respond after confirmation. After the other person has confirmed or corrected your steel-man, respond to the confirmed version of their position rather than the one you had before. The quality of the resulting exchange — in terms of mutual understanding, conflict de-escalation, and productive outcome — is consistently better when both parties are arguing about what the other person actually thinks rather than what each assumes the other thinks.

The steel-man protocol does not require you to feel what the other person feels. It requires you to model accurately how the world looks from their position. That is cognitive empathy, activated on demand. If you want the attention and cognitive flexibility practices that make this protocol a reflex rather than an effort, The Focused Mind gives you the framework for exactly that. Bloom and Singer showed what the target is. The Focused Mind gives you the daily practice that gets you there.

See also: How to Improve Your Communication Skills for Gottman's Four Horsemen and the softened startup protocol, and How to Master Your Emotions for the Gross cognitive reappraisal research and the Lieberman affect labeling mechanism.