Confidence Ratings in Nuclear Process Control: A Pilot Study with a Simplified Task

In this paper we describe a pilot study where we explore the potential of self-rated confidence judgments in the context of control room human performance evaluation. Response time, accuracy, and confidence data were obtained in a simplified task in a simulator environment of a nuclear control room. Licensed operators were presented with blocks of questions regarding specific images portraying recognizable displays with relevant process information -- after answering the primary task they were requested to evaluate their confidence in the response. The images included either benchmark (e.g. numerical values) or innovative (e.g. numerical values plus graphs and figures) design features. Overall results show that: a) there is not a complete correspondence between response accuracy and confidence ratings; b) participants took longer to reply to the confidence ratings when answers to the primary task were incorrect; c) level of confidence seems to be discriminative, with the innovative displays showcasing higher levels of certainty.