Defeasible Knowledge and Argumentative Reasoning for 3APL Agent Programming.

SEBASTIÁN GOTTIFREDI; ALEJANDRO J. GARCÍA; GUILLERMO R. SIMARI

Sydney, Australia

Workshop; Twelfth International Workshop on Non-Monotonic Reasoning (NMR 2008); 2008

In this work we propose to integrate a defeasible argumentationlogic programming formalism for representing beliefsand reasoning into 3APL thus extending its representationalcapabilities. Using this formalism the agent can represent tentativeinformation in the form of weak rules. Since strongnegation is allowed in the head of these rules, contradictoryknowledge can be represented. The formalism allows theidentification of the pieces of knowledge that are in contradictionand a dialectical argumentation process is used for decidingwhich information prevails. In particular, the argumentationbased definition of the inference relation makes it possibleto incorporate a treatment of preferences in an elegantway. This integration increases the capabilities of the componentsfor knowledge representation and reasoning, aimingto the implementation of more sophisticated autonomousagents.