Robotic - Cognitive Adaptive System for Teaching and Learning (R-CASTLE)

• Ano de defesa: 2020
• Autor(a) principal: Tozadore, Daniel Carnieto
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Biblioteca Digitais de Teses e Dissertações da USP
• Palavras-chave em Português: Adaptive systems; Artificial intelligence for education; Human-robot interaction; Inteligência artificial para educação; Interação humano-robô; Robótica social; Sistemas adaptativos; Social robotics
• Link de acesso: https://www.teses.usp.br/teses/disponiveis/55/55134/tde-31082020-093935/

Artificial Intelligence (AI) has taken an important role in peoples routine. Mainly because it enables the automation of repetitive tasks and the customization of services for each user. Both of these resources are made possible by the knowledge that is created from data generated by past experiences. Especially in the educational field, AI can help teachers to optimize their working time in recurring actions of planning, executing and evaluating their activities. For students, AI can enhance the learning experience through interactive devices that, at first, increase students interest and motivation for being a novelty and then try to continue producing these effects in long-term interactions through techniques of adaptation and customization. However, one of the biggest problems is the lack of naturalness to use these techniques as allies. Based on the needs of teachers and students presented in literature, this project sought a way to meet these needs in a unique approach, proposing a computational architecture that communicates in an intuitive way with teachers through a graphical interface and with students through a social robot. The result is a Cognitive Adaptive System for Teaching and Learning (R-CASTLE). This system aims to enable AI algorithms as tools to assist teachers in planning, executing and evaluating their educational activities without having previously presented technical knowledge of these algorithms. At the same time, R-CASTLE offers the students a technological and challenging way to carry out practical exercises, at a level of difficulty corresponding to that presented by each of them. AI algorithms allow the robot to use visual and verbal communication to collect indicative values in students responses and body expressions to assess their attention, communication and learning skills. Further, it allows also to use this data in adapting and customizing the system in order to maintain students engaged for longer period of time in the activities. The graphical interface also provides easy ways to manipulate data generated from past activities to be modified and optimized for future activities. Although it is difficult to statistically evaluate the efficiency of this project as a whole due to the large amount of specialized data for this type of solution, studies with analyzes of isolated modules and initial tests of the complete system have pointed out optimistic indications about the potential of this tool to collaborate in a practical and intuitive way with students and teachers of elementary schools and also for those interested in using R-CASTLE in other tasks of Human-Robot Interaction.