The ELSIE Knowledge Based System

The ELSIE Expert System runs on a IBM-PC compatible of any reasonable size, and hence on the machine used by most surveyors. It was developed at the University of Salford in a 18-month project funded by the Alvey Programme.

ELSIE was built using the Savoir Expert System shell, which was selected for its efficiency and its inference net formalism. ELSIE's four modules (budget, time, procurement and development appraisal) are integrated around a project database so that information given in one module will be passed to others if required. All modules operate by asking the user a sequence of questions.

The knowledge base contained links that determine the dependencies among the pieces of information held, so that the answers given are propagated 'intelligently' throughout the knowledge base and may affect which questions are asked later on, or in which order. After the main question sequence, the result is declared and the user arrives at a 'What Now' point, at which explanation reports can be obtained, information can be over-ridden, and the information can be archived in the project database for later use. Not only may the answers to the question sequence be overridden but also upwards of a hundred assumptions that the module makes. In practice, the question answers would be changed to perform what-iffing, while the assumptions would be overridden to take account of special or local circumstances.

The user interface is relatively simple, comprising 'windows' composed of character graphics. To certain questions the user may answer 'unknown', in which case the module makes a knowledgeable assumption.

By 1991 over 400 copies of ELSIE had been sold to the surveying profession, and it was felt that this would provide an excellent empirical base for studying knowledge based systems in use. Consequently, a study by the Universities of Salford and Newcastle, funded by the Science and Engineering Research Council, found evidence of knowledge based systems having significant impact on the roles and tasks of their users, and from this work a three level model of usage and benefits was developed. It also investigated degrees of trust by users in knowledge based systems.

Copyright (c) Andrew Basden, 1997.