A Unifying Framework for Probabilistic Argumentation

Formal argumentation has attracted significant attention in the field of Knowledge Representation and Reasoning over the past two decades. Dung’s Argumentation Framework (AF) has been extended in various directions, including approaches that incorporate quantified uncertainty regarding the existence of arguments and attacks. However, comparatively less effort has been devoted to integrating probabilistic reasoning into structured argumentation or other extensions of AF. In this paper, we introduce the Unified Probabilistic Argumentation Framework (UPAF), a general and expressive theoretical model capable of capturing a wide range of existing argumentation formalisms. UPAF can be equipped with an environmental model that assigns (not necessarily in-dependent) probabilistic events to elements of the underlying argumentation structure, thus enabling reasoning under uncertainty. We demonstrate that UPAF can encode classical and extended forms of Dung’s framework, as well as structured argumentation frameworks such as Assumption-Based Argumentation and Defeasible Logic Programming. Finally, we discuss the computational complexity of key reasoning tasks within UPAF instances.