Why good people follow bad orders 

Milgram's own explanation was the agentic state. On this account a person entering a hierarchy stops seeing himself as the author of his actions and begins to see himself as an instrument carrying out another's wishes. Responsibility is transferred upward, conscience is not silenced but redirected towards doing the assigned job well, and the resulting behaviour is not cruelty but a kind of dutiful abdication. Participants who continued frequently checked who would be answerable, and were reassured by being told that the experimenter accepted responsibility.

The context in which this landed was inescapable. Milgram began the work in 1961, the year of Adolf Eichmann's trial in Jerusalem, and Hannah Arendt's account of Eichmann as a functionary rather than a monster appeared in 1963. Milgram made the link explicit himself, and the phrase banality of evil has been attached to his findings ever since. That pairing gave the study enormous cultural reach and also distorted it, because a one-hour laboratory procedure at a university is not a genocide, and Milgram was careful in his 1974 book about the limits of the analogy even as his framing invited the comparison.

The most substantial modern challenge does not dispute the data. Alexander Haslam and Stephen Reicher, working partly from Milgram's own archive, argue for what they call engaged followership. On their reading participants were not passive automatons in an agentic trance. They were actively identifying with the experimenter's project, science and the advancement of knowledge, and were being recruited into it. The evidence they point to includes the pattern of the prods: appeals to what the experiment requires were far more effective than the bare command of the fourth prod, which is the least authoritarian sentence in the script and the one participants most often resisted. If people were simply obeying orders, the direct order should have worked best. It worked worst.

That distinction matters outside the laboratory. An agentic state account suggests the antidote is reminding people that they are responsible. A followership account suggests the more important question is who has managed to make a harmful course of action look like a worthy shared cause, and that resistance depends on refusing that identification rather than on individual assertiveness.

Both accounts agree on the point that survives every criticism of the methodology. Ordinary situational architecture, applied to ordinary people, produces behaviour that those same people would confidently predict they were incapable of.

Establishing that came at a price paid by the participants, and the bill arrived quickly.