The Transfusion Medicine Tutor: A Model for the Design and Use of Expert Systems for Teaching

Abstract

There is a long research history addressing the design of knowledge-based systems for tutoring. There exists, however, very little rigorous empirical data describing the impact of such systems on users. This paper presents the results of two studies that were conducted to investigate the effectiveness of an expert system for teaching medical technology students a complex abduction task, the identification of alloantibodies in a patient's blood. The first study, involving 30 students, showed that those students who used the system with all intelligent tutoring functions turned on and with instructor assistance available went from 0% correct on a pre-test case to 80% − 93% correct on post-tests (p < 0.001). Students in a control group, using a passive version of the system with the tutoring functions turned off, went from 7% correct on a pre-test case to 27% correct on post-tests (non significant; p > 0.05). The results of the second study found that 36 students who used the system with all intelligent tutoring functions enabled but without instructor assistance went from 19% correct on a pre-test to 83% − 86% correct on the post-tests (p < 0.001). In order to better understand the roles of the computer and the teacher in these studies, computer logs and videos collected for each student were analyzed and coded for certain key behaviors.