Unsupervised learning algorithm for adaptive group formation: Collaborative learning support in remotely accessible laboratories

Abstract

Skills and knowledge that can be gained by groups of individuals will be affected by the characteristics of those groups. Systematic formation of the groups could therefore potentially lead to significantly improved learning outcomes. This research explores a framework for group formation that continuously adapts rules used for the grouping process in order to optimize the selected performance criteria of the group. We demonstrate an implementation of this approach within the context of groups of students undertaking remote laboratory experiments. The implementation uses multiple linear regression analysis to adaptively update the rules used for creating the groups. In order to address specific learning outcomes, certain behaviors of the group might be desired to achieve this learning outcome. We can show that by using a set of individual/group characteristics and group behavior we can dynamically create rules and hence optimize the selected performance criteria. The selected performance is in reality the group behaviour, which might lead to improved learning outcomes.