Abduction with Negation as Failure for Active and Reactive Rules

Abstract

Recent work has suggested abductive logic programming as a suitable formalism to represent active databases and intelligent agents. In particular, abducibles in abductive logic programs can be used to represent actions, and integrity constaints in abductive logic programs can be used to represent active rules of the kind encountered in active databases and reactive rules incorporating reactive behaviour in agents. One would expect that, in this approach, abductive proof procedures could provide the engine underlying active database management systems and the behaviour of agents. We analyse existing abductive proof procedures and argue that they are inadequate in handling these applications. The inadequacy is due to the inappropriate treatment of negative literals in integrity constraints. We propose a new abductive proof procedure and give examples of how this proof procedure can be used to achieve active behaviour in (deductive) databases and reactivity in agents. Finally, we prove some soundness and completeness results for the new proof procedure.