The Nash counterfactual considers the question: what would happen were I to change my behaviour assuming no one else does. By contrast, the Kantian counterfactual considers the question: what would happen were everyone to deviate from some behaviour. We present a model that endogenizes the decision to engage in this type of Kantian reasoning. Autonomous agents using this moral framework receive psychic payoffs equivalent to the cooperate-cooperate payoff in Prisoner’s Dilemma regardless of the other player’s action. Moreover, if both interacting agents play Prisoner’s Dilemma using this moral framework, their material outcomes are a Pareto improvement over the Nash equilibrium.

纳什反事实考虑了一个问题：假设没有其他人改变我的行为，我会发生什么。相比之下，康德的反事实考虑了这样一个问题：如果每个人都偏离某种行为会发生什么。我们提出了一个模型，该模型将参与这种类型的康德推理的决定内生化。无论其他参与者的行为如何，使用这种道德框架的自主代理都会获得相当于囚徒困境中合作-合作回报的心理回报。此外，如果两个交互代理都使用这个道德框架来玩囚徒困境，那么他们的物质结果是对纳什均衡的帕累托改进。