Are People Basically Good or Bad?

The honest answer

The question is probably the wrong shape. The research does not support a clean verdict that people are basically good or basically bad; it supports the view that humans are unusually cooperative and prosocial by default, while also being capable of serious harm under particular conditions. Which side shows up depends heavily on situation, incentives, and social context — not on a fixed inner nature. The honest version is that the everyday baseline is more cooperative than the cynical story assumes. People help strangers, follow norms without being watched, and contribute to shared goods far more than a pure self-interest model predicts. At the same time, the capacity for cruelty is real and should not be waved away. The accurate picture is a cooperative species with conditional capacities, rather than a moral scoreboard with one winning answer.

What the data actually shows

A large body of cooperation research finds that humans are markedly prosocial compared with what narrow self-interest would predict. In economic games, people routinely share, reciprocate, and punish unfairness at a cost to themselves; everyday life is full of low-level helping — directions, held doors, small favours — that happens without reward or surveillance. The journalist Rutger Bregman's book Humankind synthesises much of this into the argument that most people, most of the time, default toward decency.

The famous studies often cited for human darkness have been substantially reinterpreted. Milgram's obedience experiments are real and unsettling, but later analysis suggests participants' behaviour depended heavily on how authority was framed and on their belief they were serving a worthy scientific cause, rather than showing blind cruelty. The Stanford prison experiment, long used as proof that situations turn people brutal, has faced serious methodological criticism — including evidence that the experimenter encouraged abusive behaviour — so it is now treated with far more caution than its popular reputation implies.

What the more careful reading supports is conditionality. People are far more likely to harm when authority sanctions it, when responsibility is diffused, when a group is dehumanised, or when norms shift — and far more likely to help when the situation makes the right action visible and easy. The behaviour tracks conditions and context, which is a different claim from people being fundamentally good or bad.

Why this feels different from how it actually is

It often feels like people are basically bad because harm is vivid and memorable while routine decency is invisible. A single act of cruelty or a shocking news story leaves a strong impression, whereas the thousands of small cooperative acts that fill an ordinary day register as unremarkable background. Our sense of human nature gets weighted toward the dramatic exceptions.

Media and stories amplify this. Conflict, betrayal, and threat are more attention-grabbing than cooperation, so the information environment over-represents the worst behaviour. Research on news consumption suggests heavier exposure to bad news is associated with a more pessimistic read of other people, even though it is not a representative sample of how people actually behave.

There is also a self-protective logic to assuming the worst: expecting bad behaviour feels safer than being caught off guard. But the research suggests this default is miscalibrated — most everyday interactions are cooperative, and the cynical prior leads people to systematically underestimate how trustworthy and helpful others actually are.

The accurate picture is a cooperative species with conditional capacities, rather than a moral scoreboard with one winning answer.

On the shape of the question

What the research says to do about it

Because behaviour tracks conditions, the research implies that shaping situations matters more than sorting people into good and bad. Cooperation tends to flourish where the helpful action is visible and easy, where contributions are noticed, and where norms make decency the expected path — and harm tends to be enabled where authority sanctions it and responsibility is diffused. Designing for the former is more effective than trying to identify and exclude 'bad people.'

On the individual level, the evidence supports a moderately trusting default. Studies suggest people consistently underestimate how willing others are to help and how positively others regard them, so extending reasonable trust is often better calibrated to reality than assuming the worst. This is a probabilistic stance, not naivety — the capacity for harm is real and worth taking seriously where the conditions for it exist.

The research on everyday helping also suggests that small, concrete acts of cooperation are common and self-reinforcing. Prosocial behaviour tends to be reciprocated and to spread, so acting on the cooperative default — rather than the cynical one — is both well-supported by the baseline data and tends to elicit the same in return.

What the research says does not help

Relying on the old textbook versions of the Milgram and Stanford studies as proof that people are fundamentally cruel does not hold up. Both have been substantially reinterpreted or criticised, and treating them as settled demonstrations of human darkness overstates what the evidence actually shows.

Adopting a blanket cynical view of human nature is also poorly calibrated. The research consistently finds that people underestimate others' trustworthiness and helpfulness, so assuming the worst tends to misread the cooperative majority of interactions and can become self-fulfilling by eroding the trust that cooperation depends on.

Flipping to naive optimism is equally unsupported. The capacity for harm is real and emerges reliably under certain conditions — sanctioned authority, diffused responsibility, dehumanisation — so pretending those risks do not exist ignores a genuine part of the evidence. The accurate stance is conditional, not a simple verdict in either direction.

Harm is vivid and memorable while routine decency is invisible.

What this looks like in real life

Illustrative

The unremarkable held door

Directions given to a stranger, a held door, a small favour returned — thousands of these fill an ordinary day and register as unremarkable background. They are exactly the cooperative baseline the research highlights and the reason a pure self-interest model underpredicts real behaviour. It is the rare act of cruelty, not the routine decency, that gets remembered.

Illustrative

Assuming the worst, and being wrong

Expecting bad behaviour can feel safer than being caught off guard, but studies suggest this default is miscalibrated: people consistently underestimate how willing others are to help and how positively others regard them. A moderately trusting stance is often better matched to reality — while still taking seriously the conditions under which harm does emerge.

Real numbers in context

This is a question better answered with patterns than with a single number, and inventing a precise statistic would be dishonest. The robust findings are directional: across cooperation experiments and observations of everyday life, humans behave far more prosocially than narrow self-interest predicts, and the classic 'proof of human evil' studies have been reinterpreted toward situational and conditional explanations rather than fixed badness.

It is worth holding the limits clearly. The Stanford prison experiment in particular faces serious methodological criticism and should not be cited as settled science; the Milgram results are real but more about framing and authority than blind cruelty. The honest summary is that people are strongly cooperative by default with a real, conditional capacity for harm — a picture the evidence supports far better than either 'basically good' or 'basically bad.'

Frequently asked questions

So are people good or bad, in the end?

The research doesn't support a single verdict. It points to humans being strongly cooperative and helpful by default, while retaining a real capacity for harm that emerges under specific conditions. Behaviour tracks situation, incentives, and context more than a fixed inner nature, so 'cooperative with conditional capacities' fits the evidence better than 'good' or 'bad.'

Didn't the Stanford prison experiment prove people turn cruel?

That study has faced serious methodological criticism, including evidence that the experimenter encouraged abusive behaviour, so it is now treated with far more caution than its popular reputation suggests. It should not be cited as settled proof that ordinary people automatically become brutal in the wrong situation.

What about the Milgram obedience experiments?

Those results are real and genuinely unsettling, but later analysis suggests behaviour depended heavily on how authority was framed and on participants' belief they were serving a worthy scientific cause — not on blind cruelty. The honest reading is that they show how authority and framing shape behaviour, not that people are fundamentally cruel.

Why do I feel like people are mostly selfish?

Partly because harm is vivid and memorable while routine decency is invisible, and partly because media over-represents conflict and bad news. Research suggests people systematically underestimate how trustworthy and helpful others are, so the cynical impression is usually worse than the cooperative reality.

Does this mean I should just trust everyone?

Not blindly. The evidence supports a moderately trusting default, because people tend to underestimate others' willingness to help — but it also shows a real capacity for harm under certain conditions. A calibrated stance is to extend reasonable trust while staying aware that situations involving sanctioned authority or diffused responsibility can enable harm.

Data sources

Bregman, R. (2020). Humankind: A Hopeful History. Little, Brown — synthesis of cooperation and prosociality research.
Milgram, S. (1963). Behavioral Study of Obedience. Journal of Abnormal and Social Psychology, 67(4), 371–378 — and later reinterpretations (e.g., Haslam & Reicher).
Le Texier, T. (2019). Debunking the Stanford Prison Experiment. American Psychologist, 74(7), 823–839.
Henrich, J., and colleagues — cross-cultural research on cooperation and fairness in economic games.

Figures are drawn from publicly available research, government statistics, and peer-reviewed studies. Where research is mixed or evolving, we say so. See our research methodology for how sources are selected and reviewed.

Last reviewed: June 2026