Interactive Critiquing as a Form of Decision Support: An Empirical Evaluation

Abstract

This research focused on the design of a decision-support system to assist blood bankers in identifying alloantibodies in patients' blood. It was hypothesized that critiquing, a technique in which a computer monitors human performance for errors, would be an effective role for such a decision-support system if the error monitoring was unobtrusive and if the critiquing was in response to both intermediate and final conclusions made by the user. A prototype critiquing system monitored medical technologists for (a) errors of commission and errors of omission, b) failure to follow a complete protocol, (c) answers inconsistent with the data collected, and (d) answers inconsistent with prior probability information. Participants using the critiquing system had significantly better performance (completely eliminating misdiagnosis rates for 3 out of 4 test cases) than a comparable control group. Detailed analysis of the behavioral protocols provided insights into how specific design features influenced performance. Practical applications of this research include its use (after refinements) as a tool for routine antibody identification in blood banks.