Researchers propose Bounded Morality framework to formalize moral computation under constraints

Researchers Max Kanwal, Caryn Tran, and Patrick Mineault introduce Bounded Morality, a formal framework that extends Herbert Simon’s bounded rationality to moral cognition. The paper, published on arXiv under the cs.AI category, argues that traditional models of moral cognition—such as deontology, consequentialism, and virtue ethics—have been treated as static rules or value functions, which obscures the computational demands of real-world moral problems.

The framework formalizes moral situations along two orthogonal dimensions: moral breadth, defined as the scope of entities treated as morally relevant, and moral depth, defined as the inferential integration required to evaluate their interactions. Under limited resources, these dimensions impose an unavoidable tradeoff, defining a feasible space of moral computation. Within this space, ethical theories are recast as locally efficient strategies adapted to different demand regimes, rather than competing accounts of moral truth.

The authors derive a formal notion of moral regret and moral progress under constraint, providing a quantitative lens to evaluate tradeoffs in moral decision-making. They argue that moral alignment in artificial systems should focus on the scaling and allocation of moral reasoning capacity, rather than on direct imitation of human judgments.

The paper is structured as a 24-page manuscript with two figures and is listed as a submission to the AAAI-26 Workshop on Machine Ethics. The authors acknowledge unspecified funding sources and provide a DOI link to the arXiv entry.