Understanding others may not be something you reason out — it may be something you simulate. The brain uses the same circuitry both to perform an action and to understand that same action in someone else. Empathy may be rooted in a bodily "inner imitation," not in cool logical inference.
In the premotor cortex of macaques, researchers found a class of neurons that fire both when the monkey grasps food itself and when it watches someone else perform the same grasp — "mirror neurons." Humans lack direct single-cell evidence, but imaging shows that observing, imagining, and physically executing an action share parts of their neural representation. In other words, you may grasp another's intention because your brain quietly "runs" that action through your own motor system.
Watch someone's hand get pricked by a needle and your own pain-related regions partly activate; watch someone grimace at a disgusting taste and your insula — where your own disgust lives — lights up too. When you "feel with" someone, you aren't abstractly imagining; you're locally reusing the very circuits you'd use to live through that state yourself. A caveat: in popular media the mirror system is often mythologized — it is better seen as one cornerstone of empathy than the whole answer.
In AI, it maps to imitation learning and embodied intelligence — a robot learning from demonstration is essentially building an internal representation of action, translating "seeing" into "being able to do." In linguistics, it supports the gestural-origin theory of language, where symbolic communication may have grown out of an action-imitation system. In evolution, high-fidelity imitation is the precondition for cumulative culture — without it, every generation reinvents the wheel and "civilization" never accumulates.
This explains a hidden cost of remote work: pure text severs most mirror channels — tone, expression, pauses are all filtered out, so emotional sync and trust within a team build more slowly. Video and voice have "higher bandwidth" not just in raw information, but because they reconnect the simulation circuit. In designing human-AI collaboration, making the AI's process simulatable — showing reasoning steps rather than only conclusions — likewise earns more human trust.
Recall a misunderstanding that arose in a text-only exchange. Had it been face-to-face or on video, which filtered-out "simulation signals" would have changed the outcome?
Ordinary people will harm others under an authority structure — often not because they are "bad," but because the power of the situation far exceeds our inflated estimate of our own morality. What you do is often governed more by "what situation you're in" than by "what kind of person you are."
In the early-1960s Yale shock experiments, subjects were asked to deliver escalating shocks to a "learner" (actually an actor) for wrong answers, up to 450 volts marked "Danger." The key mechanism is the "agentic state": once a person hands responsibility over to authority, the self switches from "autonomous agent" to "executor of orders," and the moral burden transfers to the one giving commands. A single line — "The experiment requires you to continue; I take responsibility" — was enough to make most people press the button, and many did so trembling, sweating, laughing nervously, yet continuing.
Obedience isn't set by a person's goodness or badness — it swings wildly with situational variables. As the victim moved from behind a wall to the same room, obedience dropped from 65% to 40%; when subjects had to press the victim's hand onto the shock plate, it fell to 30%; when the authority left the room and gave orders by phone, it sank to about 21%. The same kind of people, merely by changing "distance" and "presence of authority," behaved utterly differently — proof that the problem lies in the structure, not in character.
In organizational behavior, it explains the diffusion of responsibility in corporate scandals — everyone executes, no one is accountable. In history and philosophy, it echoes the "banality of evil": enormous harm can be produced by ordinary people each just "doing their job." In AI alignment, it is the mirror problem — we want AI to obey instructions yet refuse them when they're harmful, which is precisely the exam humanity itself failed.
As a tech leader, the real leverage isn't repeating "have moral courage" — it's designing structures that make saying "no" easy: anonymous escalation, a rotating "red-team dissenter," dissent baked into process rather than left to individual nerve. The same logic applies to AI agents — don't only train them to obey; preserve an explicit exit to "refuse and explain," or a sufficiently "obedient" system becomes the most dangerous one.
In your team, how many barriers must a junior member cross to say "no" to an obviously wrong decision from above? Can you remove even one of them?
The "ache" of being rejected or excluded is not a literary metaphor — the brain processes social exclusion using circuitry that partly overlaps with physical pain. For humans, social connection is a survival need on the same order as food and safety, not an optional "soft" nicety.
In the classic Cyberball experiment, subjects believe they're tossing a ball online with two other players — until, mid-game, they get "ignored" and never receive the ball again. Imaging shows this experience of exclusion activates the dorsal anterior cingulate cortex (dACC) and anterior insula — exactly where the "distress" (not the location) component of physical pain lives. The evolutionary logic is blunt: for our ancestors, being cast out of the group nearly meant death, so the brain piggybacked the "social rejection" alarm onto the already-existing pain warning system.
If the two pains share circuitry, can a drug for physical pain treat "heartache"? In one study, a group took a common analgesic (acetaminophen, the active ingredient in Tylenol) for several weeks; they reported markedly lower daily "social pain," and imaging showed a weakened dACC response to rejection. A headache pill blunting the sting of being snubbed is the most counterintuitive — and most compelling — evidence that social pain is real pain.
In management, it explains why being marginalized or suddenly dropped from a key chat hurts so deeply — that's real pain, not a "thin skin." In product design, social apps' "seen but no reply" and like counts precisely pluck this circuit, a source of both stickiness and harm. In public health, loneliness has been found to damage lifespan on par with smoking and obesity, because chronic social pain is a chronic stressor that keeps inflammation and cortisol elevated, slowly eroding the body.
In parenting, a child's meltdown over peer rejection isn't "overreacting" — its neural reality is as genuine as a scrape, and deserves to be taken seriously rather than waved off with "don't mind them." In leading remote teams, belonging can't ride on slogans; it needs structure: stable small groups, contributions named and seen, clear signals that "you belong here" — all of which actively manage the team's social-pain threshold.
In the past month, which "overreaction" — yours or someone near you — was actually social pain sounding an alarm? If you treated it as real pain, how would you respond?
Moral judgment is mostly intuition — fast and emotional; reasoning is largely a "press secretary" that justifies the intuition after the fact. And human morality is not one ruler but several independent "foundations," like taste receptors, blended in different proportions by different cultures and groups. This is why cross-tribal moral arguments persuade almost no one.
A common metaphor is the "elephant and the rider": the emotional intuition (the elephant) instantly sets a direction, and reason (the rider) only afterward invents the justification. Morality's "taste buds" are usually grouped as: care/harm, fairness/cheating, loyalty/betrayal, authority/subversion, and sanctity/degradation (with liberty/oppression added later). Each person and culture has a different sensitivity mix across these, brewing its own moral flavor.
| Foundation | More liberal | More conservative |
|---|---|---|
| Care / Harm | ●●● | ●● |
| Fairness / Cheating | ●●● | ●● |
| Loyalty / Betrayal | ● | ●●● |
| Authority / Subversion | ● | ●●● |
| Sanctity / Degradation | ● | ●●● |
"Moral dumbfounding": tell people a carefully designed story that breaks a taboo yet harms no one, and most will flatly declare "that's wrong" — but when pressed for why, they can't produce a single reason, while still refusing to budge. This powerfully shows that judgment precedes reasons — we first have the gut "it's just wrong," then go back to hunt for arguments, not the other way around.
In evolution, these foundations are seen as adaptations to different cooperation problems — care protects the young, fairness deters free-riders, loyalty binds the group. In AI value alignment, it's a warning: human values are multidimensional and often mutually conflicting, and can't be compressed into a single utility function to maximize. In political communication and marketing, the skilled appeal to the foundation the audience actually cares about, not the one they personally favor.
When leading cross-cultural teams or globalizing a product, an "obviously right" design often steps on a landmine — because people elsewhere judge it with different moral taste buds. The key to persuasion and consensus is first identifying which foundation the other side is truly sensitive to (fairness? loyalty? authority?), then speaking in that flavor — rather than repeatedly upping the dose of your own. With people whose values differ sharply, assume "different taste" rather than "unreasonable," and your success rate climbs.
Recall an argument where you were sure the other side was "beyond reason." Under the assumption "we're tasting with different moral receptors," does their position suddenly become understandable?