Towards a Domain-Agnostic Computable Policy Tool

Mitchell Falkow, Henrique Santos, Deborah L. McGuinness

Introduction

The problem with these frameworks:

"Domain-specificity" is problematic

  • Reworking/extending existing frameworks becomes more complicated
  • Can introduce errors during the process

Addressing the problems of domain-specificity

  • Propose the use of domain-specific knowledge expressed in machine-readable formats (i.e., knowledge graphs) incorporated inside CPs
  • Benefits:
    • Ensures machine-readability
    • Mitigates inconsistencies
    • Enables use of domain knowledge during evaluation
Problem:
  • Few frameworks use this approach
  • Of the frameworks that use this approach, very few provide tools to work in those frameworks

Our Approach

We propose a A Domain-Agnostic Policy Tool — ADAPT.

  • Enables construction, visualization, and management of policies
  • Combines W3C recommended standards for ontologies (OWL) and provenance (PROV) to represent policies in the form of knowledge graphs
      This enables the use of OWL reasoners to perform evaluation
  • Configured using domain-specific knowledge in knowledge graphs and domain-derived ontologies

ADAPT's Architecture

ADAPT consists of 4 services, with specific roles:

  1. the UI: handles user interaction
  2. the API: processes attribute and policy data;
  3. the evaluation engine: evaluates requests created by the user; and
  4. the knowledge store: maintains the domain-specific knowledge as well as the policies

Demo of ADAPT

ADAPT can be used in any domain as long as the user provides the necessary information to operate within that domain. Below are screenshots of a demonstration of ADAPT functioning in the healthcare domain; Even though this is specific to the healthcare domain, if the user provides an adequate set of ontologies then ADAPT will be able to function in the domain(s) of those ontologies as well.



Creating a policy with ADAPT


Visualizing a policy