A logical analysis of normative reasoning: Agency, action, and argumentation

Abstract

In this thesis, employ logic to increase our understanding of normative reasoning. We do this by including agents in our formal analysis. Norms are inextricably linked to agents: they provide reasons to act and influence how we shape our world. Nevertheless, agentive aspects are often abstracted away, yielding oversimplified formalisms and understudied themes. Recent developments in Artificial Intelligence (AI) have created novel challenges for the logical study of normative reasoning. This thesis addresses several of these topics by assigning a pivotal position to agents. The conducted research is interdisciplinary, drawing from methods in philosophy, logic, and AI. The thesis comprises three parts: (I) agency, (II) action, and (III) argumentation. In Part I, we investigate how obligations impact the choices of agents. To do so, we adopt and extend the agency logic of ‘Seeing To It That’ (STIT). We formally investigate the limits of contrary-to-duty reasoning when reasoning about choices and obligations over time. Furthermore, we conduct a comprehensive logical study of the principle of Ought implies Can and its relation to other normative reasoning principles. In Part II, we study ways in which obligations and prohibitions promote the actions performed by agents. We formally analyze instrumentality statements, which express actions as instruments for attaining ends. Furthermore, we apply our formalism to an ancient theory in Sanskrit philosophy that reduces obligations to instrumentality statements. In Part III, we investigate explanations in the context of defeasible normative reasoning. We develop a sequent-style proof-theoretic approach tailored to generating explanatory arguments and show how these arguments can be used in formal argumentation to create explanations. Furthermore, we develop a general, nonmonotonic proof-theoretic formalism that incorporates argumentative concepts like attack and defense, and extend it to defeasible normative reasoning.