Identification and Visualization of Students' Weaknesses by Analyzing Interactions between Students and Virtual Patients
Problem based learning (PBL) is an effective strategy for medical students to learn clinical diagnostic skills. With predesigned cases, PBL aims at training the students to face clinical problems in a teamwork manner, with proactive and collaborative exploration to derive well-justified conclusion. In the course of PBL, the tutor plays an active role in directing the students to conduct discussion and exploration, and accordingly identifies the diagnostic weaknesses of the students. PBL thus incurs a heavy burden to the tutors, and hence many previous studies employed information technology to support PBL. However, the previous studies mainly focused on providing information support to PBL, instead of analyzing diagnostic weaknesses of the students. Therefore, in this project we plan to develop and integrate suitable information techniques to support the systematic creation of PBL cases that may be used to identify the diagnostic weaknesses of students. The PBL case is the basis on which a virtual patient interacts with the students, and by analyzing the dialog between the students and the virtual patient, the system may pinpoint two kinds of weaknesses of students: reasoning sequence weaknesses and reasoning completeness weaknesses. To identify reasoning sequence weaknesses, a canonical state dependency graph (SDG) is constructed for each PBL case to govern the proper sequence of dialog between the students and the patient. To identify reasoning completeness weaknesses, a diagnostic knowledge map (DMap) is constructed to outline the relationships between risk factors, diseases, and symptoms so that the system may figure out whether the students' clinical reasoning has covered enough relationships. The study may contribute to the fields of medical education and information technology, producing practical and theoretical impacts to both fields.
Keywords: clinical diagnosis, problem based learning, virtual patient, identification of diagnostic weaknesses, state dependency graph, diagnostic knowledge map