Simulating Institutional Dynamics: A Multi-Agent Framework for Predicting Legal Reform Outcomes Using Extended Phenotype Theory

Jensen Huang observed in March 2026 that biology is transitioning from science to engineering through computation. This paper argues that the same transition applies to law. Legal systems, like biological organisms, exhibit phenotypic expression (institutional dynamics) driven by an underlying genotype (constitutional structure). Drawing on Extended Phenotype Theory (EPT) as formalized for legal-institutional dynamics in prior work (Lerer, 2026a), this paper treats legal norms as cultural replicators whose fitness determines survival in an institutional ecosystem. Building on the Constitutional Lock-in Index (CLI) developed in prior work (Lerer, 2026a), I validate this four-component metric of structural institutional rigidity against four jurisdictions: Argentina (0.89), Spain (0.51), Brazil (0.40), and Chile (0.24). I implement a multi-agent simulation engine adapting the OASIS framework (Yang et al., 2024) for institutional dynamics, with seven agent types, nine legal actions, and two novel theoretical mechanisms: Heteronomous Bayesian Updating (HBU) and the Coefficient of Resistance to Institutional Change (CRI). Monte Carlo validation against 23 Argentine labour reforms (1974-2024) produces emergent CLI of 0.92 (target 0.89, error 0.033) with 93.8% reform-blocking in adversarial scenarios. Leave-one-out cross-validation across 60 cases in four countries yields 56.7% accuracy (95% CI: 44.1-68.4%), above a 43.3% baseline. I submit the programme to Mario Bunge's demarcation criteria and identify both strengths and unresolved weaknesses, particularly regarding the measurability of utilities in game-theoretic components.