Output details
11 - Computer Science and Informatics
University of Edinburgh
Agreeing on plans through iterated disputes
<22> Originality: First paper to address planning under disagreement about world state and available actions (a largely overlooked problem) with a focus on scalability.
Significance: We bridge the areas of AI planning and argumentation, enabling a "softer" alternative to "conformant planning". Rather than computing plans acceptable for all agents, our approach enables solutions that can be defended against others' potential objections.
Rigour: No other work achieves tractable argumentation in domains with thousands of actions. Our methodology is based on a new formalism for mapping expressive action theories to deductive argumentation. Published at the top international conference in the area (AAMAS).