Intelligent Learner Modeling in Real-time

A learner model was developed for students learning using a computer-based tutorial. An initial model was populated with the results of a pretest of the subject matter to be learned, a biographical questionnaire and two learning style assessment tools, Kolb LSI and Entwistle ASI. Neural networks provided supplementary information by tracking online behaviours, such as which options were selected, in which sequence and how long they spent on each option. Options included seeing an example, reading explanatory text or taking a practice test. Online learning behaviours that led to successful and unsuccessful learning outcomes were then assessed using statistics and neural nets. Post-tests measured learning outcomes. Only the Entwistle ASI proved to be a useful a priori measure. Both statistical and neural network analyses produced a similar learner model categories, which means it is possible to perform valid real-time learner modeling. This would allow real-time adaptation of both the content and the delivery format of the instructional materials. There is an added advantage in being able to continually adapt or keep pace as learners progress and continue to learn.

Autorentext

Dr. Dalkir is a professor at McGill University, specializing in artificial intelligence, innovation and knowledge management. She has a B.Sc. and an MBA from McGill and a Ph.D. in Educational Technology from Concordia. Her research is focused on organizational learning, tacit knowledge/expertise and knowledge transfer in succession planning.