Why do legal institutions persist even when evidence overwhelmingly favors reform? The standard explanations, from veto players (Tsebelis, 2002) to path dependence (Pierson, 2000), describe outcomes without explaining their cognitive mechanics. Recent experimental evidence offers a candidate mechanism: Confer et al. (2025) demonstrate that children as young as four alter their evidentiary standards as a function of group membership, sampling less evidence before adopting group beliefs (χ²(1) = 78.15, p = 0.012), inflating confidence in group-supporting evidence by approximately 39% (χ²(1) = 128.03, p < 0.001), and revising beliefs at half the rate when group identity is salient (χ²(1) = 7.46, p = 0.006). If this cognitive architecture, which operates before metacognitive control develops, persists into adulthood and scales through professional socialization, it could supply the missing microfoundation for institutional persistence. I propose Heteronomous Bayesian Updating (HBU) as a formalization of this scaling mechanism and organize its institutional effects through a multilevel game-theoretic typology. The framework generates falsable predictions that I test illustratively against the Constitutional Lock-in Index (CLI) developed in prior work. Argentina's Article 14bis, with a CLI of 0.87 and 0/23 successful reform attempts between 1991 and 2025, serves as the primary case. The core claim is theoretical: legal institutions calcify not because of conspiracy or path dependence alone, but because the human cognitive apparatus for evaluating evidence is systematically distorted by group affiliation, and legal professions constitute precisely the kind of identity-conferring groups that trigger this distortion.