AI 中文总结
该研究探讨策略性利他主义的贝格均衡能否作为高效制度设计的行为基础，发现其虽可解决囚徒困境，但无法实现弱帕累托最优规则，且贝格可实施性要求高于纳什可实施性，存在策略性利他主义的悖论。
AI 中文摘要
利他主义能否作为高效制度设计的行为基础？我们研究贝格均衡（Berge equilibrium，形式化策略性利他主义的解概念：每个代理人的对手会集体最大化其收益）中的完全实施。尽管贝格均衡可解决囚徒困境，还能消除社会困境中的合作失败，但我们表明这种乐观态度无法从个体行为过渡到制度设计层面。首先，我们证明弱帕累托最优规则无法在贝格均衡中被实施；其次，在严格偏好的无限制域上，任何既弱帕累托有效又可在贝格均衡中实施的社会选择规则必然是独裁的；最后，我们证明贝格可实施性严格比纳什可实施性要求更高：策略性利他主义下可实现的每个社会目标，在自利行为下也可实现，但反之则不然。贝格均衡在行为层面的吸引力特征——对手保护每个代理人的收益，恰恰是在设计层面排除高效非独裁制度的原因。
英文摘要
Can altruism serve as the behavioral foundation for efficient institutional design? We study full implementation in Berge equilibrium, the solution concept that formalizes strategic altruism: each agent's opponents collectively maximize her payoff. While Berge equilibrium resolves the Prisoner's Dilemma and can eliminate cooperation failures in social dilemmas, we show that this optimism does not survive the move from individual behavior to institutional design. First, we show that the weak Pareto optimal rule is not implementable in Berge equilibrium. Second, on the unrestricted domain of strict preferences, any social choice rule that is both weakly Pareto efficient and implementable in Berge equilibrium must be dictatorial. Finally, we show that Berge implementability is strictly more demanding than Nash implementability: every social goal achievable under strategic altruism is also achievable under self-interest, but not vice versa. The very feature that makes Berge equilibrium appealing at the behavioral level, that is, opponents protecting each agent's payoff, is precisely what forecloses efficient, non-dictatorial institutions at the design level.
