AI Ethics Framework Models Moral Reasoning as Distribution Over Theories
A new framework for ethical AI decision-making models moral reasoning as a distribution over normative ethical theories, moving beyond binary judgments. The approach introduces a normative ethics simplex that integrates multiple theories, supported by a benchmark of 450 cases across 15 subtheories for ensemble learning. The research, published on arXiv (2605.28707), addresses the need for explainable and accountable AI in socially consequential contexts.
Key facts
- Framework models moral reasoning as distribution over normative ethical theories
- Introduces a normative ethics simplex integrating multiple theories
- Benchmark of 450 cases across 15 fine-grained subtheories created
- Cases describe ethical dilemmas in natural language with contextual features
- Uses stacked ensemble learning for implementation
- Published on arXiv with ID 2605.28707
- Addresses limitations of scalar or binary moral judgments in AI
- Focuses on accountability and explanation in autonomous decision-making
Entities
Institutions
- arXiv