A novel approach to interface design for a neural network expert system

A novel approach to the interface design for a neural network expert system is presented. The approach is based on G.A. Kelly's (1955) personal-construct theory, modified by repertory gridlike techniques, which is able to handle information in two or more dimensions. Experiments have so far been carried out on an example of gunnery experts from the Royal School of Artillery, Larkhill, UK. These show that there was almost complete agreement between all experts and they were satisfied with the interpretation of the data. Further it was found that precautions were necessary when using Kelly's approach to knowledge elicitation.<<ETX>>