Morality: a battle royale—the performance of evaluative procedures in symmetric interactions

Abstract This paper presents a large-scale simulation study of how different evaluative procedures perform in strategic interaction. We model five canonical evaluative perspectives, namely outcome aggregation, universalization, self-regard, worst-case sensitivity, and other-regarding concern, as transformations of agents’ payoff functions, and we consider agents who may select among these procedures when choosing how to act. This generates a family of 31 evaluative procedures, including both foundational forms and hybrid procedures that permit conditional rule selection. We evaluate these procedures across the full space of symmetric two-player, two-strategy normal-form games using Monte Carlo methods. For each pairing of procedures, we construct the induced strategic interaction, compute its Nash equilibria, and evaluate the resulting material outcomes. This yields two comparative benchmarks: performance under universal adoption and performance in mixed populations containing self-regarding agents. The results reveal several robust structural patterns. Procedures incorporating universalization-based reasoning consistently achieve high levels of coordination and perform well across environments. Hybrid procedures that permit conditional rule selection frequently outperform fixed procedures, particularly in heterogeneous populations. We also identify an equivalence between self-regarding and other-regarding evaluation under symmetric interaction and show that certain procedures maintain a stable presence in mixed populations despite being vulnerable in direct pairwise encounters. We argue that these findings are best understood not as establishing the correctness of any moral theory but as identifying structural features of evaluative reasoning that support coordination and strategic robustness. In particular, the results suggest that universalization and conditional rule selection play a central role in promoting successful interaction across a broad class of strategic environments. We also discuss the extent to which these results may have broader ethical and evolutionary implications. The paper contributes to a growing literature that studies morality through the lenses of game theory, evolutionary robustness, and strategic interaction.