Integrating Reflective Equilibrium and Structured Argumentation: A Logical Approach to Norm Identification

Tianwen Xu, Jieting Luo

Keywords: norm identification, structured argumentation, theory construction

TL;DR: This paper propose a logical approach to the fundamental challenge of norm identification in open multi-agent systems, by integrating reflective equilibrium and structured argumentation.

Abstract: In open multi-agent systems without centralized authority, agents cannot rely on predefined norms and must instead learn them through decentralized norm identification. Existing data-driven and machine-learning methods are vulnerable to data quality, lack explainability, and often require labeled datasets, limiting their applicability in open environments. To address these issues, this paper proposes a logic-based approach to norm identification by integrating the method of reflective equilibrium and structured argumentation. Identifying norms is then conceived as constructing a consistent moral theory, and a moral theory is represented as an argumentation theory. This allows us to inductively construct a consistent moral theory from moral judgments reflecting prevalent morality in a community, and any potential conflict within the theory are resolvable by its priority structure comprised of specificity and firmness orderings. The theory can also develop arguments that justify an agent’s decision by such theory. We prove that, whenever the firm part of the constructed theory applies, the agent can always derive a definite moral answer, fulfilling the dual demands of consistency and justification. Moreover, when presented with a set of prima facie norms as inputs, our method ensures the path-independent identification of any firm core that exists.

Submission Number: 87