The Odd Man and the Gaussian: Why Collective Intelligence Requires Structural Dissent

Every collective cognitive system — scientific, legal, clinical, political — faces a specific risk: converging on an incorrect answer with growing confidence. This article argues that structural dissent — the institutionalised presence of at least one heterodox perspective in any decisionmaking process — is not a dysfunction but a necessary epistemic condition for collective systems operating under correlated-error risk to approach truth. The argument converges from three independent directions. Statistically, the wisdom of crowds requires that observers be independent and diverse: the operative principle is the decorrelation of errors, not the shape of the distribution — collective wisdom is not a property of crowds, but of crowds of odd men. Legally, the Talmudic Sanhedrin’s unanimous-acquittal rule institutionalised mandatory dissent as a guarantee of the system’s cognitive integrity, not of individual justice. Philosophically, from Mill to Feyerabend, disagreement is the mechanism through which collective knowledge self-corrects. The article then analyses the epistemic risk introduced by large language models: systems that compress the distribution of perspectives, correlate individual errors, and — as adoption concentrates around a small number of dominant models — risk transforming a crowd of odd men into a homogenised system, blind to its own systematic errors.