Model-Based Diagnosis with ASP for Non-groundable Domains

Moritz Bayerkuhnlein, Diedrich Wolter

Abstract: Model-based diagnosis is a technique for identifying malfunctioning components in systems. While it has successfully been applied to systems such as digital circuits, this paper aims to extend applicability to systems such as programs that process values from large domains, for example, term structures. In these cases, especially when multiple components may be faulty, it is challenging to identify a diagnosis that provides a consistent model with respect to the specified domain. This paper presents an Answer-Set Programming (ASP) based method for computing such diagnoses. We are particularly interested in functional circuits over domains of values, such as rational numbers and inductive data types, to diagnose faults in programming assignments in order to advance intelligent tutoring systems. This article shows how a consistent diagnosis, justified by intermediate values, can be achieved efficiently using ASP. Additionally, an adaption to Constraint Answer Set Programming with s(CASP) is presented that avoids grounding, allowing domain sizes to be handled that are too large to be grounded.