Affective and behavioral assessment for adaptive intelligent tutoring systems

Adaptive Intelligent Tutoring Systems (ITS) aim at helping students going through the resolution of a given problem in a principled way according to the desired outcomes, the intrinsic capabilities of the student, and the particular context in which the exercise takes place. These systems should be capable of acting according to mistakes, boredness, distractions, etc. Several works propose different models to represent the problem being solved, the student solving it and the tutor guidance to the desired solution. The system complexity requires non trivial models whose corresponding parameters need to be estimated with difierents kinds of data, usually requiring heavy and difficult sensing and recognition tasks. In this work, we present some of the work in progress in the BIG-AFF project. Between other issues, we deal with the use of low cost and low intrusive devices to gather contextual data to losely drive the actions of an ITS without constructing a fully structured student model with corresponding affective and behavioral states. The idea is to improve the students' learning outcome and satisfaction by progressively learning how to adapt the ITS in terms of the sensed data.