Beyond Homo Economicus: Computational Instantiation and Empirical Analysis of Amartya Sen's Meta-Ranking Theory in Multi-Agent Social Dilemmas

This study formalizes Amartya Sen's Meta-Ranking theory within a Multi-Agent Reinforcement Learning (MARL) framework, demonstrating that Bounded Commitment—dynamic moral commitment preserving moral residue even under resource crisis—is the key mechanism for resolving social dilemmas. v4 Updates (2026-02-22): Phase W critique defense: (W1) Explicit SUTVA analysis via exposure mapping with ATE decomposition into direct and spillover effects across 3 environments; (W2) Bounded Commitment spectrum — 4-model comparison with Moran ESS and ε-sensitivity Pareto analysis; (W3) Integrity-constrained U_meta variants resisting reward hacking under 50% adversarial populations; (W4) Cross-domain behavioral fingerprint transfer protocol across 4 environment pairs. Total 42 analysis modules generating 88 figures from a single command. Code: https://github.com/Yesol-Pilot/EthicaAI