A Logic Programming Framework for Reasoning about Know-How

The BDI model is a well accepted architecture for the representation of rational agents, both at theoretical and practical levels. The beliefs component in this model is focused on the representation of beliefs about the world and other agents and widely independent of an agent’s intentions. We argue that also the logical representation of an agent’s intentions and its know-how, which captures the beliefs about actions and procedures, has to be taken into account when modeling rational agents. With a declarative rather than procedural representation of know-how we obtain the possibility to reason with these procedural beliefs in the same way as with any other logical beliefs. Using the notion of know-how as introduced by Singh we formalize a usable and concrete agent architecture that benefits from this representation of procedural beliefs in multiple ways. To model both, the agent’s logical beliefs as well as its know-how, we make use of extended logic programs under the answer set semantics that are capable of handling uncertain information. This way we are lifting the limitations imposed by the use of different forms of representation. We define a general algorithm that uses this representation for means-end reasoning. Furthermore, we present diverse ways of reasoning under uncertainty about know-how that are enabled by our framework and show relations to ex-