Um estudo no uso de feedbacks para mudanças de estados afetivos do estudante usando eletroencefalograma e expressões faciais

Abstract

The affective state is one of the most important factors when it comes to student motivation. Studies show that positive emotions contribute to learning, while negative emotions make it difficult. The role of feedback is important, especially when it is directed and adapted to the student, and can be post-process feedback, which is provided after the answer, or in-process feedback, which is presented before the finalization of the question. However, in the educational context, the affective factor is rarely used due to several factors, such as the disproportionality between the number of students and the number of teachers, making it difficult to identify the individual affective state, or the fact that it is difficult to adapt the content and feedbacks according to each individual emotion variation. A possibility that can help in these tasks is to use the computational context to record and recognize the student’s affective states according to their facial expressions and also the neural activity via Electroencephalogram (EEG) and with that, adapt feedbacks using Intelligent Tutoring Systems (ITS) according to these changes affective. This work presents a study on the use of process feedbacks to analyze affective changes. For this, a modular approach called NeurAFFeed was elaborated, performing the collection, analysis and storage of the student’s affective states, while solving programming language issues in a Virtual Learning Environment (VLE), through face data using a webcam and EEG data through a low-cost Brain-Computer interface (BCI) called OpenBCI, so that an affective module for ITS observes such states and decides whether or not to display feedbacks in process, which will be presented to the student in the form of notifications on a VLE. The experiment used the NeurAFFeed approach in a group of twenty students, half being the Control Group (GC), without feedback, and the other half being the Experimental Group (EG), with feedback. Data analysis showed that there was a significant difference between the CG and the EG, with an increase of 14.9% in changes from affective states to preferable quadrants in face data and an increase of 12.5% for EEG data. The possibility of developing rankings of the best feedbacks according to face and EEG data was also evidenced. The analysis of the results indicates that the NeurAFFeed approach allows the realization of studies that use in-process feedbacks to verify affective changes with face and EEG data collected with non-clinical and low-cost devices such as webcam and OpenBCI, respectively. This work contributes with the possibility of using his approach for new studies in the areas of affective ITS, in-process feedbacks and accessible analysis of neural activity.