Research on Student Modeling

What can we learn from observing students interacting with educational software? Wellinstrumented educational software can record students’ interactions over a period of several months—considerably longer than most psychology lab studies. Although long-term longitudinal studies exist, computers can record student-tutor interactions at much finer grain size; each student response, mouse click, and even spoken word can be recorded for later analysis. Thus, computer tutors occupy a “sweet spot” in collecting educational data; the data are detailed enough to detect fine-grained learning, and over a sufficient time frame to investigate longer-term learning. My research is on educational data mining, the study of answering educational and psychological questions from this rich data stream generated by computer tutors. Can we use these data to revise our theories about how students learn the domain material? To assess a student’s progress at learning? To determine which types of instruction benefit which students?