AdABI: an Adaptive Assessment System Based on Bayesian Inference
AdABI: an Adaptive Assessment System Based on Bayesian Inference

Summary/Abstract: Student assessment holds the essential role of informing both the instructor and the learner of the extent to which they have attained the established learning objectives. As computerized assessment has lately become widespread and effective in academic environments, issues such as the optimal degree of automation and the trade-offs between the pedagogical and technological aspects in designing assessment systems require special attention. This paper introduces an adaptive web-based assessment system - AdABI, designed to address such issues by: (1) allowing teachers to define their multiple-choice tests, to set their evaluation criteria and the examination schedule, and to view the results of an assessment; (2) allowing the students to take examinations for the courses they are enrolled in and to view their personal results; (3) creating and automatically updating the student’s profile after each examination; (4) adapting the student’s future examinations based on the inferred profile. The adaptive behaviour of the testing and evaluation module is implemented through a Bayesian network by means of which the system is able to deliver personalised tests to the student. More precisely, the system infers three characteristics (tendency to guess the answers, adaptability, and fatigability) for each particular student based on the student’s answer patterns and adapts the question types and the order of the questions in future examinations so as to counterbalance the student’s characteristics. Consequently, by controlling for the student’s profile in examinations, AdABI may attenuate the bias induced by the learner’s answer patterns. This way, it may ultimately provide a premise for fairer evaluations.

• Details
• Contents