In-process feedback by detecting deadlock based on EEG data in exercise of learning by problem- posing and its evaluation

Feedback on learning activities is one of the most important issues in achieving adaptive learning. In this study, we propose a mechanism for solving this problem by detecting the deadlock state of a learner during a learning activity and providing feedback to eliminate such a state. Feedback on the products of learning activities (we call it “after-process feedback”) has been implemented in numerous interactive and adaptive learning environments. However, feedback during an activity (we call it “in-process feedback”) has rarely been implemented. In-process feedback is considered to be much better than after-process feedback when learners have difficulty or become frustrated with the learning material during the learning process. The difficulty in implementing in-process feedback lies in the timing and content of the feedback. It has been pointed out that the detection of a deadlock must be achieved as early as possible; otherwise, it reduces the learning motivation of the learner. Therefore, we focused on electroencephalograph (EEG) data, which are difficult to cheat and can clearly detect the state of the learner. By combining EEG data with machine learning, we developed a model for detecting when a learner is stuck, allowing us to detect the timing. After that, we generate the proper feedback by estimating the knowledge state of the learner based on the knowledge structure and task response status. We implemented and evaluated the in-process feedback approach in a learning environment posing arithmetic word problems.