ARTFEED — Contemporary Art Intelligence

Strategic Perspective Activation in Context-Dependent Argumentation

ai-technology · 2026-06-01

A new paper introduces context-dependent argumentation frameworks (CDAFs), extending Dung's theory to model how an agent can strategically activate different perspectives to influence argument evaluation. The defeat function varies per context, and a perspective-labeled specialization derives it from a relevance set and priority. A worked example shows an agent's target argument rejected under full-relevance injective priorities but accepted under partial activations, a result unattainable by value-based argumentation frameworks (VAFs). The decision problem ACTIVATION-MANIPULATION is defined with baseline complexity bounds; tight bounds and multi-agent variants remain open.

Key facts

  • CDAFs extend Dung's argumentation theory
  • Defeat function determines which attacks succeed per context
  • Perspective-labeled specialization uses relevance set ρ and priority π
  • Agent's target argument rejected under full-relevance injective priorities
  • Target accepted under partial activations
  • No VAF audience can mirror the partial activation result
  • ACTIVATION-MANIPULATION decision problem defined
  • Baseline complexity bounds recorded; tight bounds and multi-agent variants open

Entities

Institutions

  • arXiv

Sources