Context-Dependent Defeat in Abstract Argumentation: Reduction to Value-Based Frameworks
A recent study available on arXiv (2608.15536) explores the possibility of simplifying context-dependent argumentation frameworks (CDAFs) into value-based argumentation frameworks (VAFs). In VAFs, the hierarchy of values held by an audience influences which attacks are perceived as defeats. Conversely, in various real-world scenarios, context—such as regulatory conditions or procedural stages—plays a more significant role than the audience itself. The authors introduce CDAFs, characterized by a single set of arguments and an attack relation, where the defeat function varies based on context, transforming each scenario into a standard Dung framework. The paper outlines a polynomial-time decision method to ascertain if a CDAF can be represented by a VAF with context-specific value rankings and addresses more complex related issues, with complexity bounds ranging from NP to Σ^p_3. Additionally, the authors offer a validated reference implementation along with performance metrics. This research enhances the theoretical underpinnings of argumentation in AI, with implications for legal reasoning and multi-agent systems.
Key facts
- Paper arXiv:2608.15536 introduces context-dependent argumentation frameworks (CDAFs).
- CDAFs model defeat functions that vary by context, unlike value-based argumentation frameworks (VAFs).
- The paper presents a polynomial-time decision procedure for reducing CDAFs to VAFs.
- Harder neighboring problems have upper bounds from NP to Σ^p_3.
- A validated reference implementation is provided.
- The work is relevant to AI argumentation, legal reasoning, and multi-agent systems.
- The paper is announced as new on arXiv.
- The abstract discusses the reduction question: whether context is genuinely new or can be collapsed into VAFs.
Entities
Institutions
- arXiv