Performance Improvement of Technical Diagnosis Provided by Human-Machine Cooperation

Abstract

The paper proposes a human-machine approach dedicated to the technical diagnosis. Both human and machine reasoning introduce a coordination problem when a shared activity has to be achieved. The ability for sharing and interpreting data is the main condition to cooperate during diagnosis. We propose a shared workspace in order to support both human and machine reasoning activities. The proposed workspace is very close to human mental models in order to reduce mental workload. We define means to coordinate reasoning and mechanisms to share data. In order to evaluate costs and benefits of a cooperative diagnosis, we have applied our concepts to the diagnosis task of a real phone network. A decision support tool dedicated to phone network diagnosis has been developed and tested in real contexts with professional operators. The results presented in the paper show the impact of the cooperative work and the increase of performance of the humanmachine teams.