An Argumentation-based Approach for Explaining Goal Selection in Intelligent Agents

Agents need to explain their goal-setting decisions: The paper argues that intelligent agents must be transparent about how they decide which goals to pursue, not just their final actions. This is critical for responsible AI systems that affect human decisions or safety-critical domains.

Argumentation provides a structured explanation method: The researchers use argumentation—a logical framework showing reasons for and against different choices—to help agents explain their goal selection process in a way that mirrors human reasoning and is easier for non-experts to understand.

Conflicts in decision-making should be made visible: The paper emphasizes that agents should explain not just what goals they chose, but also what competing goals or constraints they had to resolve. This transparency about trade-offs helps users understand whether an agent's priorities align with their own values.

Two levels of explanation serve different needs: The authors propose "partial explanations" (simpler, focused) and "complete explanations" (comprehensive, showing all reasoning), suggesting that responsible AI systems should offer flexible transparency that matches what different users need to understand and trust the system.

Natural language explanations improve accessibility: By converting logical reasoning into pseudo-natural language explanations, the approach makes agent decision-making understandable to non-technical stakeholders, which is essential for meaningful human oversight of AI systems.