An intelligent framework for monitoring students Affective Trajectories using adaptive fuzzy systems

In this paper we investigate the Affective Trajectories Hypothesis in an educational context, and its possible implications on Affective Computing. Using the results from an online survey we try to explore the relationships of the Affective Trajectories basic elements, namely one´s current affective state, prediction of the future, and experienced outcomes following this prediction, with a set of education related emotions. The relations of these elements with flow, excitement, calm, boredom, stress, confusion, frustration and neutral linguistic emotional labels are presented and discussed. Their predictive power is evaluated by using these elements as inputs to different classification systems, and observing their performance in mapping different combinations of those elements to specific emotion labels. A data-driven fuzzy approach is utilized in order to linguistically model the underlying relations between the emotions, and the basic elements, by creating easily interpretable fuzzy rule bases. In our research we suggest that the basic elements are combined in a personalized way in order for an individual to choose a specific emotion label to describe his affective state. For this reason a fuzzy adaptive approach is also implemented, in order to demonstrate the importance of individual differences in this process, and the benefits of having a personalized system that can perpetuate modelling of emotional trajectories over learning tasks. Finally an overview and a basic implementation of an affective computing system which uses these elements are presented, and future research directions discussed.