Abstract

Automated commonsense reasoning is essential for building human-like AI systems featuring, for example, explainable AI. Event Calculus (EC) is a family of formalisms that model commonsense reasoning with a sound, logical basis. Previous attempts to mechanize reasoning using EC faced difficulties in the treatment of the continuous change in dense domains (e.g., time and other physical quantities), constraints among variables, default negation, and the uniform application of different inference methods, among others. We propose the use of s(CASP), a query-driven, top-down execution model for Predicate Answer Set Program- ming with Constraints, to model and reason using EC. We show how EC scenarios can be elegantly modeled in s(CASP) and how its expressiveness makes it possible to perform deductive and abductive reasoning tasks in domains featuring, for example, constraints involving dense time and fluents.

Evaluation

We showed how event calculus can be modeled in s(CASP), a goal-directed implementation of constraint answer set programming with predicates, with much fewer limitations than other approaches. s(CASP) can capture the notion of continuous time (and, in general, fluents) in event calculus thanks to its grounding-free top-down evaluation strategy. It can also represent complex models and answer queries in a flexible manner thanks to the use of constraints. The main contribution of the paper is to show how the event calculus can be elegantly modeled using the s(CASP) predicate ASP system that seamlessly supports constraints. The modeling of the event calculus by s(CASP) is more elegant and faithful compared to other approaches such as F2LP, where time has to be discretized. While other approaches such as ASPMT do support continuous time, their reliance on SAT solvers makes their implementation really complex as associations among variables are lost during grounding. The use of s(CASP) brings other ad- vantages, for example, justification for a successful query is obtained for free: in a query-driven system, justification is nothing but a proof trace. Likewise, explanations for observations via abduction are also generated for free, thanks to goal-directed, top-down execution of s(CASP).

Benchmarks

Benchmark.