Enablers and inhibitors in causal justifications of logic programs

Self Cite

Artificial intelligence (AI) approaches to problem-solving and decision-making are becoming more and more complex, leading to a decrease in the understandability of solutions. The European Union’s new General Data Protection Regulation tries to tackle this problem by stipulating a “right to explanation” for decisions made by AI systems. One of the AI paradigms that may be affected by this new regulation is answer set programming (ASP). Thanks to the emergence of efficient solvers, ASP has recently been used for problem-solving in a variety of domains, including medicine, cryptography, and biology. To ensure the successful application of ASP as a problem-solving paradigm in the future, explanations of ASP solutions are crucial. In this survey, we give an overview of approaches that provide an answer to the question ofwhyan answer set is a solution to a given problem, notably off-line justifications, causal graphs, argumentative explanations, and why-not provenance, and highlight their similarities and differences. Moreover, we review methods explaining why a set of literals isnotan answer set or why no solution exists at all.

Section: Explaining Negative Literals In Causal Justificationsmentioning

confidence: 99%

Answering the “why” in answer set programming – A survey of explanation approaches

Fandinno¹,

Schulz²

2019

Self Cite

“…Among these, Why-not Provenance Justifications (WnP) (Damásio et al 2013) share with our approach a semantic definition in terms of algebraic operations. A formal comparison was done in (Cabalar and Fandinno 2016a). With respect to (Pontelli et al 2009), a formal relation has not been established yet.…”

Section: Theoremmentioning

confidence: 99%

Justifications for programs with disjunctive and causal-choice rules

Cabalar

Fandinno

2016

Self Cite

In this paper, we study an extension of the stable model semantics for disjunctive logic programs where each true atom in a model is associated with an algebraic expression (in terms of rule labels) that represents its justifications. As in our previous work for non-disjunctive programs, these justifications are obtained in a purely semantic way, by algebraic operations (product, addition and application) on a lattice of causal values. Our new definition extends the concept of causal stable model to disjunctive logic programs and satisfies that each (standard) stable model corresponds to a disjoint class of causal stable models sharing the same truth assignments, but possibly varying the obtained explanations. We provide a pair of illustrative examples showing the behaviour of the new semantics and discuss the need of introducing a new type of rule, which we call causal-choice. This type of rule intuitively captures the idea of "A may cause B" and, when causal information is disregarded, amounts to a usual choice rule under the standard stable model semantics.

“…From a technical point of view, we have shown that our semantics is a conservative extension of the stable model semantics and that satisfy the usual desired properties for an LP semantics (casual stable models are supported models, minimal models in case of normal programs and can be iteratively computed by split table programs). It worth to mention that, besides the syntactic approaches to justifications in LP, the more related approach to our semantics is (Damásio et al 2013), for which a formal comparative can be found in (Cabalar and Fandinno 2016a) and that (Pontelli et al 2009) allows a Prolog system to reason about justifications of an ASP program, but justifications cannot be inspected inside the ASP program.…”

Section: Conclusion Related Work and Open Issuesmentioning

confidence: 99%

Deriving conclusions from non-monotonic cause-effect relations

Fandinno¹

2016

Self Cite

We present an extension of Logic Programming (under stable models semantics) that, not only allows concluding whether a true atom is a cause of another atom, but also deriving new conclusions from these causal-effect relations. This is expressive enough to capture informal rules like "if some agent's actions A have been necessary to cause an event E then conclude atom caused(A, E)," something that, to the best of our knowledge, had not been formalised in the literature. To this aim, we start from a first attempt that proposed extending the syntax of logic programs with so-called causal literals. These causal literals are expressions that can be used in rule bodies and allow inspecting the derivation of some atom A in the program with respect to some query function ψ. Depending on how these query functions are defined, we can model different types of causal relations such as sufficient, necessary or contributory causes, for instance. The initial approach was specifically focused on monotonic query functions. This was enough to cover sufficient cause-effect relations but, unfortunately, necessary and contributory are essentially non-monotonic. In this work, we define a semantics for non-monotonic causal literals showing that, not only extends the stable model semantics for normal logic programs, but also preserves many of its usual desirable properties for the extended syntax. Using this new semantics, we provide precise definitions of necessary and contributory causal relations and briefly explain their behaviour on a pair of typical examples from the Knowledge Representation literature. (Under consideration for publication in Theory and Practice of Logic Programming)