Exploiting Readily Available Web Data for Scrutable Student Models

This paper describes our work towards building detailed scrutable student models to support learner reflection, by exploiting diverse sources of evidence from student use of web learning resources and providing teachers and learners with control over the management of the process. We build upon our automatically generated lightweight ontologies using them to infer from the fine-grained evidence that is readily available to higher level learning goals. To do this, we have to determine how to interpret web log data for audio plus text learning materials as well as other sources, how to combine such evidence in ways that are controllable and understandable for teachers and learners, as required for scrutability, and finally, how to propagate across granularity levels, again within the philosophy of scrutability. We report evaluation of this approach. This is based on a qualitative usability study, where users demonstrated good, intuitive understanding of the student model visualisation with system inferences.