The quality of human-automation cooperation in human-system interface for nuclear power plants

Abstract

The use of automation within high-risk industrial production systems has increased markedly during the last 50 years. Automatic systems have gained in autonomy and authority, whereby the activity of the systems has become less dependent on operator interventions. This has brought forward the suggestion that human-automation transactions should be conceptualized within the framework of cooperation, and consequently that automatic systems should be designed to be cooperative. The question is then how design can promote human-automation cooperation, and how the quality of cooperation can be assessed. The OECD Halden Reactor Project performed two closely related experiments, which allowed assessments of whether the quality of human-automation cooperation would be promoted by a human–machine interface designed to increase the observability of the automatic system's activity using graphical and verbal feedback, as compared to a conventional human–machine interface. The experiments were performed in a full-scale nuclear power plant simulator, using licensed operators as subjects, and applied a 2×2 within-subject design. The quality of human-automation cooperation was assessed from subjective operator judgements. The experiments demonstrated a clear improvement in human-automation cooperation quality when the observability of the automatic system's activity was increased. The relationship between human-automation cooperation quality and the effectiveness of the joint human–machine system's performance was furthermore explored, but no clear results were found. As the trend in automation design seems to imply an increase in system autonomy and authority, the issue of human-automation cooperation can be expected to further gain in importance in the future settings.