Balancing System Expressivity and User Cognitive Load in Semantically Enhanced Policy Modelling

摘要

Policy Engineering is the process of authoring management policies, detecting and resolving policy conflicts and revising existing policies to accommodate changing resources, business goals and business processes. Policy authoring involves developing a policy rule base populated with policies that specify where actions specified on behalf of subjects may or should be performed on targets (resources). Policy engineering can be a daunting process in terms of complexity of the subjects, targets and actions used in rules, and the potential for conflicting policy rules to be committed to the rule base. Semantic modelling of rule elements can help automate the detection of such conflicts, but the additional layers of abstractions may themselves add to the complexity faced by the policy author. In this paper, we aim to assist the policy author by increasing the system expressivity with semantics, while at the same time minimizing the perceived cognitive load due to additional model complexity. We build on work with abstractions aimed at achieving this goal in the modelling of organisational grouping for subjects of policy-rules, which supports specific authority-based group abstractions to ease the maintenance of the subject model in the face of frequent organisational change. In this paper, we study this approach as used in combination with description logic based modelling of target semantics plus logic programming assertions across subject, targets and actions. This is performed through detailed analysis for policy authoring deliberations observed in a user evaluation of these modelling techniques.