Atypical Failure, Maximal Consequence: Bayesian Character Inference, Threshold Outcome Luck, and Multiplicative Punishment Cascades

Le résumé fourni par la source

This conceptual article develops a formal theory of atypical human failure in which severe life-course consequences can emerge from a conjunction of adverse conditions, a low-frequency behavioural lapse, stochastic outcome mechanics, categorical legal or institutional thresholds, and dynamically propagating collateral sanctions. The framework integrates moral luck, outcome bias, person-situation theory, human-error research, Bayesian character inference, and the literature on criminal-record stigma and digital punishment. It makes five principal contributions. First, a contextual hazard model distinguishes low baseline propensity from elevated conditional risk under unusual combinations of stressors and weak safeguards. Second, the article formalises threshold-amplified outcome luck, showing how relatively small differences in realised harm can produce discontinuous changes in formal burden. Third, it introduces Bayesian Predictive Atypicality, measured through posterior predictive probability and event surprisal, and then models biographical compression as salience-weighted updating in which one extreme negative observation receives excessive likelihood weight relative to a longer behavioural history; Salience-Prior Distortion is defined using Kullback-Leibler divergence from a full-history posterior. Fourth, a nonnegative punishment-propagation matrix captures spillovers among legal, reputational, occupational, social, and digital burdens, yielding a closed-form expression for discounted lifetime burden under stability conditions and a Collateral Punishment Multiplier. Fifth, the paper specifies a five-study empirical programme designed to distinguish explanation from exculpation, identify outcome-luck effects, test threshold discontinuities, and measure post-sanction persistence. The theory does not imply that rare misconduct is uninformative or that harmful outcomes should be ignored. Its normative claim is narrower: proportional accountability requires distinguishing stable propensity from situationally amplified tail failure, culpable action from stochastic outcome, and finite formal punishment from collateral processes that can become self-reinforcing and temporally unbounded.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.