Quality Rating of Learning Objects Using Bayesian Belief Networks

Candidate: Kate Han
Type: Master of Applied Science, Interactive Art (MASc-IA), School of Interactive Arts and Technology
Date: July 19, 2004
Senior Supervisor: Dr. Vive Kumar
Thesis: Download Thesis Document

Abstract

The unceasing growth of the Internet has lead to new modes of learning in which students routniely interact online with instructors, other students and more frequently, digital resources. Much recent research has focused on building infrastructure for these activities, especially to facilitate searching, filtering and recommending online resources known as learning objects [1].

Although newly defined standards for learning object metadate [2] are expected to greatly improve searching and filtering capabilities, students, teachers, and instructional developers may still be faced with choosing from many pages of result listings returned from a single learning object query. The listed objects tend to vary widely in quality. Without proper recommendation, the learning object enquirers not only grope in the dark in front of overwhelming information, but also easily fall for poorly designed and developed instructional materials, wasting time and effort. Hence, there is a clear need for quality evaluations prior to recommendation that can be communicated in a coherent, standardised format to measure the quality of learning objects. Consquently, we need certain criteria to obtain this evaluation.

In the last few years, a number of quality rating standards have been developed. As different evaluation instruments are deployed in learning object repositories serving specialised communities of users, what methods can be applied for tranlating evaluative data across instruments to allow this data to be shared among different repositories? How can the large number of possible explicit and implicit measures of preference and quality be combined to recommend objects to users?

In this research, a new way of collecting and analysing learning object quality evaluation data is explored. Bayesian Belief Networks [3], a mathematical theory, is applied to tackle problems of insufficient and incomplete reviews in learning objects repositories, as well as translating and integrating data among different quality evaluation instruments. Two Bayesian Belief Networks are constructed to probabilistically model relationships among different roles of reviewers, among various explicit and implicit ratings, and among items of different evaluation measurements. Results from simulated testing cases show that the model makes quantitatively reliable inferences about different dimensions of learning object quality.