From Actions to Obligations: A Deontic Action Model Logic

Abstract

We introduce the Deontic Action Model Logic (DAML), a dynamic modal framework for reasoning about obligations over actions in multi-agent systems. DAML extends the epistemic Action Model Logic by incorporating deontic evaluation mechanisms that assess agents' actions in terms of both the desirability and the likelihood of their outcomes. Obligations arise for those actions that maximize expected deontic value among an agent's available alternatives at a given decision point, yielding a formal account for reasoning about conditional and context-sensitive obligations in settings involving strategic interaction and incomplete information. DAML supports principled action selection in norm-governed multi-agent systems, and is the first such framework to derive these obligations using the action model logic machinery. We provide an axiomatization of the logic and prove soundness and completeness with respect to its semantics. Finally, we demonstrate the expressive power of our framework through applications to the Miners' Puzzle and other multi-agent deontic scenarios.