Conditional Updates of Answer Set Programming and Its Application in Explainable Planning

In explainable planning, the planning agent needs to explain its plan to a human user, especially when the plan appears infeasible or suboptimal for the user. A popular approach is called model reconciliation, where the agent reconciles the differences between its model and the model of the user such that its plan is also feasible and optimal to the user. This problem can be viewed as a more general problem as follows: Given two knowledge bases π a and π h and a query q such that π a entails q and π h does not entail q, where the notion of entailment is dependent on the logical theories underlying π a and π h , how to change π h-given π a and the support for q in π a-so that π h does entail q. In this paper, we study this problem under the context of answer set programming. To achieve this goal, we (1) define the notion of a conditional update between two logic programs π a and π h with respect to a query q; (2) define the notion of an explanation for a query q from a program π a to a program π h using conditional updates; (3) develop algorithms for computing explanations; and (4) show how the notion of explanation based on conditional updates can be used in explainable planning.