Predictors for Human Performance in Information Seeking, Information Integration, and Overall Process in Diagnostic Tasks

This is the authors' abstract. We don't add key points for this paper.

Diagnostic accuracy can be life-critical or safety-critical, and it is of preventive interests to predict human performance on diagnostic tasks beforehand. This study conducted an experiment to select competent predictors to account for between-individual, between-task, and between-trial differences of various performance facets of human diagnostic processes. In the experiment, participants were required to diagnose a novel accident in a simulated nuclear power plant. Eight dependent variables were used to describe their information seeking, information integration, and overall process of diagnosis. The results identified the closeness of an individual’s mental model to that of an expert and the steps that an optimal algorithm took to solve a diagnostic task as individual- and task-predictors, respectively. Trial number was found to account for the practice effects. Competency and incompetency of various predictors were discussed, and consequential recommendations were provided.