Identifying user strategies in exploratory learning with evolving task modelling

In this paper we present work on adaptive identification of learners' strategies, gradually developing a higher level of adaptation based on evolving models of mathematical generalisation tasks in an Exploratory Learning Environment. A similarity-based classification approach is defined for the identification of strategies, using an initially small number of classes (i.e. strategies). A strategy is composed of several patterns with relations between them. An evolution monitor component observes changes in the environment and triggers a mechanism that builds-up the task model. The task model evolves when new relevant information becomes available by adding a new strategy (class) or a new inefficient pattern, i.e. patterns that make it difficult for the learner to generalise.