Cognitive Monoculture and the Individual-Team Reliance Paradox: A Formal Model of Shared AI Dependence in Organizational Decision Making

Abstract

Generative artificial intelligence can improve the quality and speed of individual knowledge work while creating an overlooked organizational risk: when many employees rely on the same model, their judgments cease to be statistically independent. This paper develops a formal theory of cognitive monoculture as common-mode dependence in human-artificial intelligence decision systems. An analytical model represents each employee's judgment as a weighted combination of an idiosyncratic human estimate and advice from one of a set of artificial intelligence sources. The model derives the team-level error variance, an individual-optimal reliance level, a team-optimal reliance level, a common-model dominance threshold, and an effective independent judgment count. The central result is an individual-team reliance paradox: a reliance level that minimizes each person's expected error can increase the error of an aggregated team decision because common model error does not diversify away. Under an illustrative baseline, a twenty-person team using a single shared model more than doubles its team-level error when all members adopt the individually optimal reliance level, while its nominal twenty judgments collapse to approximately 1.31 effective independent judgments. Diversifying model sources mitigates the effect only when cross-model errors are imperfectly correlated. The framework therefore shifts artificial intelligence governance from counting human reviewers to measuring the independence architecture of their judgments.