Enhancing Information Systems with Co-operative Explanatory Capabilities

by Simon Lambert

The Knowledge Engineering Group of the Department for Computation and Information at CLRC has long experience in the area of co-operative explanatory capabilities for infor-mation systems. Originating in early work on expert systems in the 1970s, these capabilities have become more advanced with the recognition that the goal is not simply to provide an explanation of a system's 'chain of reasoning', but to improve the effectiveness of its use in a whole variety of ways. The Knowledge Engineering Group is now working on the TAP-EXTRA project, partly funded by the European Commission as ESPRIT project 20719, to show the benefits that can result.

The type of industrial situations where there is a case for enhancing information systems with these co-operative explana-tory capabilities include the following:

• reacting quickly and reliably in a control room to events that may be rare but pose a great risk
• enabling staff with less experience to make decisions
• increasing the confidence of staff in their decision support aids in the control room.

TAP-EXTRA is a Trial Application, whose purpose is to apply a previously developed technology to demonstrate its benefits and show how the risks can be cushioned. The technology comes from a previous ESPRIT project, I-SEE (Interactive Self-Explanation Engine) in which CLRC was also a leading partner. As well as three demonstrator applications, I-SEE also developed a generic toolkit for enhancing information systems with capabilities such as:

• justification in domain terms of the application's working and conclusions
• tailoring of the explanations to the tasks and competence of the users
• intelligent help facilities
• co-operative problem solving between user and system.

The I-SEE toolkit consists of a number of components, including:

• generic parts which will normally be used without change, such as the Explanation Planner
• repertories of components available for use, though possibly needing extending in new domains or application types, such as the set of Primitive Explanation Techniques
• partial models of domains and systems (eg diagnostic systems) abstracted from the developments undertaken in the I-SEE project.

In TAP-EXTRA, the application, provided by the large French company Lyonnaise des Eaux, is a system for assisting with flood control in the city of Bordeaux. In times of heavy rainfall, large amounts of data arrive at the control room.

The existing application filters this data and presents messages to the user, but there is a need for a higher level explanatory assessment of the situation to assist in reliable decision making. Currently, the users' needs are being studied in detail and a number of scenarios are being developed, while in parallel the appropriate explanation techniques are being designed.

Please contact:
Simon Lambert - CLRC
Tel: +44 1235 445716
E-mail: S.Lambert@rl.ac.uk