Manifold Answer-Set Programs for Meta-reasoning

Abstract

In answer-set programming (ASP), the main focus usually is on computing answer sets which correspond to solutions to the problem represented by a logic program. Simple reasoning over answer sets is sometimes supported by ASP systems (usually in the form of computing brave or cautious consequences), but slightly more involved reasoning problems require external postprocessing. Generally speaking, it is often desirable to use (a subset of) brave or cautious consequences of a program P1 as input to another program P2 in order to provide the desired solutions to the problem to be solved. In practice, the evaluation of the program P1 currently has to be decoupled from the evaluation of P2 using an intermediate step which collects the desired consequences of P1 and provides them as input to P2. In this work, we present a novel method for representing such a procedure within a single program, and thus within the realm of ASP itself. Our technique relies on rewriting P1 into a so-called manifold program, which allows for accessing all desired consequences of P1 within a single answer set. Then, this manifold program can be evaluated jointly with P2 avoiding any intermediate computation step. For determining the consequences within the manifold program we use weak constraints, which is strongly motivated by complexity considerations. As an application, we present an encoding for omputing the ideal extension of an abstract argumentation framework.