One Adjudicator, Many Substrates: A Negligence Microfoundation of the Reach-Contingent Blame Dierential

Across a sequence of models of oversight — a vertical cascade, an observation portfolio, a dynamic rotation process — the blame a monitor bears for a surfaced failure rises with his own observability, a reach-contingent differential that drives strategic inattention, low-observability traps, and relative-blame distortions. In each model the differential is a reduced-form primitive. This paper shows it to be the equilibrium of a single object: a monitor held to a capacity (“should-have-known”) standard, and judged by an adjudicator who infers culpability from noisy evidence, bears expected blame equal to the posterior that the surfaced failure was within his reach — increasing in observability, and reproducing each model's reduced-form blame map: the posterior itself, at its own negligence rate, for the models that already use it, and its $q \to 1$ limit — the linear gate $s = \lambda$ — for the rest. The result is invariant across the substrates: the same adjudicator, applied to a hierarchy's link, a portfolio's task, or a process's period, reconstructs the corresponding reduced form, and in the portfolio it derives the cross-task blame covariance that one model imposes by hand. The microfoundation carries an institutional reading. The non-reduction principle that auditing standards assert — responsibility undiminished by reliance on, or inability to observe, the work of others — cannot be implemented by any adjudicator that must infer culpability, since a failure beyond reach carries a posterior of essentially zero. The gap between the standards' de jure non-reduction and the de facto negligence differential that enforcement operates is thus structural rather than a lapse of enforcement, and is the same on every substrate; the lever that narrows it is a safe harbour for documented diligence, which a larger monitoring budget cannot supply.