Controle Inteligente de Tempo Livre em Tutoria Multissessão

• Autor(a) principal: GOMES, Viviane Margarida
• Data de Publicação: 2009
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Repositório Institucional da UFG
• Texto Completo: http://repositorio.bc.ufg.br/tede/handle/tde/990
• topic: Sistemas Tutores Inteligentes, Aprendizagem por Reforço, Tempo Livre; Intelligent Tutoring Systems, Reinforcement Learning, Free Time; CNPQ::ENGENHARIAS

Intelligent Tutoring Systems are softwares to provide customized instruction by using techniques of Computational Intelligence. This research proposes the intelligent control of free time (break interval) in multi-session tutoring. The teaching strategy employs tutoring modules with the following steps: 1) video class, 2) exercise, 3) practical suggestion, 4) free time, and 5) revision exercise. As part of the learning environment, free time (step 4) can contribute to increase the knowledge retention. Based on the student performance in exercises, the proposed system uses Reinforcement Learning to control free time durations. The intelligent agent decides according to the policy that has been indicated by the Softmax method. Among the relevant points of this algorithm, it can be highlighted the optimistic initial values, the incremental implementation and the temperature adjustment (Gibbs distribution parameter) to the selection of action. Two student groups have participated of data collection. The experimental group (with intelligent control) has been compared to the control group (where decisions belong to the student). In the groups, the intelligent agent or the student determines the action that will be followed or, in more detail, if free time will be shorter, longer or maintained. In comparison, statistical data analysis have shown significant and equivalent gains in knowledge retention. However, students from experimental group have realized more accurately the role of free time as a component of the teaching strategy