Method for detecting a learner's stalled situation by using the variance of operational time intervals

Automatically detecting a learner's stalled situation or deadlock in an educational support system using computers is very useful. This paper proposes a method for detecting such a deadlock on the basis of the record of a learner's operations and evaluates it by tests on human subjects. Specifically, it uses the moving variance of the operational time intervals to detect a deadlock. As a result, in about 80% of the practice problems with self‐declarations of a deadlock by the subjects, the deadlock could be detected at a point in time prior to the self‐declarations. In addition, deadlocks were detected by the proposed method in three cases in which states close to a deadlock had been determined from questions after the completion of the practice problems, even though there were no self‐declarations. © 2003 Wiley Periodicals, Inc. Syst Comp Jpn, 34(7): 63–75, 2003; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/scj.10256