DB2 is a complex relational database application that presents difficult challenges to enterprise database administrators (DBAs). For that reason, the IBM Toronto Labs has a dedicated User Experience Team for DB2. IBM has undertaken a new thrust to incorporate autonomic computing into DB2. An acknowledged challenge in meeting this objective is engendering user trust in the automation. The purpose of this research program is to develop a design framework for appropriate operator trust in, and reliance on, automation.

The purpose of this first stage was to determine the applicability of Lee and See’s conceptual model of automation reliance to recently-developed automated functionality within DB2, specifically, to critique interfaces and documentation based on the three categories of attributional abstraction (purpose, process and performance) identified by the model.

This project was a continuation of [CAS 04] and explored the potential impact of CWA on a “policy-based” approach to automation administration. In addition, a field study of operational database administrators (DBAs) was conducted to investigate their interaction with DB2 and with automation within DB2. This field study shed light on DBAs’ reliance on procedures and practice in development environments (“sandboxes”) to ensure database availability. In this environment, it is important that automation be transparent in its processes and in reporting its performance in order to gain operators’ trust in a high-availability environment.
Additionally, a framework was developed to leverage the Abstraction Hierarchy to describe an automated tool in such a way as to identify relevant purpose-, process- and performance-related information.

This project was a continuation of [CAS 04] and [CAS 05]. The framework developed in [CAS 05] was applied to a new automated tool in DB2, to identify information from multiple levels of attributional abstraction. The results of this analysis were applied to a subsequent release of DB2, where the information identified by the analysis was added to documentation and clarified in logs generated by the automation

This project is a continuation of [CAS 04], [CAS 05] and [CAS 06]. An experiment has been designed to determine the impact of information from multiple levels of attributional abstraction on the resolution of database administrator trust in new automation. External factors, including general personal trust, are also being measured in order to determine their impact on operator trust and reliance decisions.