An Agent-Based Framework for Simulating Institutional Lock-in Under Repressive Regimes

Abstract

Computational models of institutional lock-in, including the Constitutional Lock-in Index (CLI), have been applied primarily to democratic systems. No agent-based framework integrates cognitive conformity, material dependency, elite filtering, and coercive capacity into a unified model of authoritarian persistence after normative collapse. This paper presents Multilevel Selection of Authoritarian Regime Survival, an agent-based architecture extending CLI methodology to authoritarian persistence, in which five agent classes interact through promotion, filtering, belief updating, resource allocation, and exit/voice decisions. The framework draws on Khurshid's Tgmenks theory of the Ecological Niche of Knowledge for domain-theoretic grounding and on Extended Phenotype Theory for its replicator-centered logic. Four falsifiable hypotheses are translated into experimental protocols with explicit manipulation, control, metric, and falsification specifications. Five historical cases are coded ordinally into parameter calibration templates, and three analytical examples, derived by hand from the specified transition rules rather than executed stochastic simulation, illustrate the model's capacity to generate discriminating predictions under contrasting configurations. Comparative calibration and the worked examples are consistent with the qualitative predictions of all four hypotheses: high-lock-in configurations are predicted to absorb a 30% resource shock within roughly 10 time steps, medium-lock-in configurations to collapse by t = 40, and epistemic openness to relate nonlinearly to performance. These are model-based predictions derived analytically, not outputs of executed stochastic simulation; no agent-based code has yet been run, and full Monte Carlo validation is the paper's principal item of future work. Replication scaffold (unexecuted at submission): github.com/adrianlerer/multilevel-selection-abm.

Received: 22 April 2026 | Revised: 20 July 2026 | Accepted: 10 August 2026

Conflicts of Interest

The author declares that he has no conflicts of interest to this work.

Data Availability Statement

Data sharing is not applicable to this article as no new data were created or analyzed in this study. The comparative calibration in Section 5 reproduces publicly available secondary-source historical and political science literature.

Author Contribution Statement

Ignacio Adrian Lerer: Conceptualization, Methodology, Software, Formal analysis, Investigation, Writing – original draft, Writing – review & editing, Visualization, Project administration.