Leveraging Conceptual Models of Trust in Automated Systems to Promote `Appropriate Trust' in Autonomic Systems

Abstract

There are subtle differences between automated and autonomic software systems with respect to the observability of information that supports trust between the system and its operators (designers, administrators, and users). Applying conceptual models of trust developed for human relationships with automated systems directly to autonomic systems is simply inadequate. An autonomic system can, however, be modeled as a composition of sub-level automation systems. In this manner, designers can engender 'appropriate trust' in autonomic systems using conceptual models developed for automated systems, but must consider how administrators and users will observe information required for trust, specifically: purpose, process, and performance. It is possible to provide designers, administrators and users with information to engender trust by using an abstraction hierarchy modeling framework (for example: ecological interface design). In this paper, we propose the use of goal models during early requirements engineering to describe the purpose information of a system in an abstraction hierarchy.