A Computational Model of Learners Achievement Emotions Using Control-Value Theory

Game-based Learning (GBL) environments make instruction flexible and interactive. Positive experiences depend on personalization. Student modelling has focused on affect. Three methods are used: (1) recognizing the physiological effects of emotion, (2) reasoning about emotion from its origin and (3) an approach combining 1 and 2. These have proven successful only in labs, or use theories of emotion not associated with an educational setting. The Control-value theory of achievement emotions holds that appraisals of control and value are most meaningful when determining emotion. This paper focuses on the design and evaluation of an emotional student model of Control-value theory applied to online GBL environments using Approach 2. This model is implemented using a dynamic sequence of Bayesian Networks (BNs). PlayPhysics - an emotional GBL environment for teaching Physics - was designed, implemented and evaluated with 118 students at ITESMCCM. To evaluate our model, we employed cross-validation and Cohen's Kappa. Our model achieved a fair to moderate accuracy of classification, but the results are promising. Future work will focus on identifying other variables that can improve classification.