Simplifying Argumentation Frameworks by Clustering Structural Patterns.

Argumentation frameworks (AFs) provide a central formalism for abstract argumentation, representing arguments and their interactions as attack graphs and serving as reasoning engines across diverse domains. One of their main strengths lies in their ability to support explanations. However, their usefulness can be limited by size and structural complexity, which can hinder understanding. To address this challenge, we investigate simplification techniques for AFs, with particular emphasis on clustering-based abstraction. By grouping arguments into clusters and interpreting attacks accordingly, clustered AFs provide reduced representations that preserve essential argumentative properties while hiding redundant details. We study formal underpinnings for clustering on AFs in particular combined with other simplification techniques, and analyze their effects on main argumentation semantics, and explore conditions under which simplifications remain faithful and avoid spurious results. Our approach positions clustering as a principled method of abstraction and highlights its potential for enhancing explainability in argumentation.