Towards Computable and Explainable Policies Using Semantic Web Standards

Henrique Santos, Jamie McCusker, John S. Erickson, Alice M. Mulvehill, Oshani Seneviratne, Deborah McGuinness

Keywords: computable policy, ontology design pattern

TL;DR: An ontology design pattern for representing computable policies.

Abstract: Policies or guidelines are defined to support decision-making in a wide range of domains. These statements describe what actions are allowed or recommended under certain conditions. The policies typically present rules that may be evaluated to generate those allowances or recommendations when encoded in machine-actionable terms. As many fields attempt to generate more computable guidelines, there is an increasing need to automatically evaluate these policies and explain results. This paper presents a novel ontology design pattern for representing policies using the OWL and PROV semantic web standards. It can be used to extend domain knowledge graphs to include representations needed to support domain-specific decision making. The encoding of policy rules using OWL restrictions over PROV entities enables the representation of common policy constructs, including subjects, actions, objects, and their attributes. This modeling can be successfully applied in a number of domains to increase inferential power and to provide better support for explaining the reasons for a given evaluation result. This is demonstrated by applying the approach with web-based tools developed for two scenarios: radio spectrum access policies and health guidelines.

Submission Number: 5