Evaluating the developer's interface to a kbs shell: some reflections

In this paper, the evaluation of the knowledge-based system shell, NEXPERT, is reported. It has been argued that AI tools of this kind have not had the success anticipated because of problems with the developer's interface. It would appear that software of this type does not help users develop a good conceptual model. An evaluation of NEXPERT has been conducted using various approaches. A group of students who had experience with the tool were asked to fill in the Software Usability Measurement Inventory (SUMI) 50 item questionnaire. The data was analysed and various aspects of usability measured: learnability, helpfulness, affect, control and efficiency. An expert user and a group of students were also interviewed to find out what they believed were the best and worst aspects of the software. The overall results are discussed and suggestions made for improving the interface of NEXPERT.