Patterns in Poor Learning Engagement in Students While They Are Solving Mathematics Exercises in an Affective Tutoring System Related to Frustration

摘要

Nowadays, detection of learner's affective state is required for adaptive learning technologies that aim to support and regulate them, due emotions are important during learning process. An affective tutoring system (ATS) was developed, with capability to detect frustration and confusion mainly, because they are associated with low and high learning outcomes. In previous experiments with students while they were solving mathematics exercises using ATS, almost all of them got a low score. Therefore, it seems to be necessary to set up user profiles in order to improve learning, in those that usually show poor motivation and engagement, to design better learning environments and virtual helper assistant to attract and identify them for extra activities. A cluster analysis was applied, and it found a correlation between frustration, low scores and clicks on help. Then, a multilayer perceptron classified different examples getting a considerable percentage of accuracy.