Towards Smart Educational Recommendations with Reinforcement Learning in Classroom

In this paper, we propose to construct a cyber-physical-social system that uses multiple sensors such as cameras and a quiz creator to track the learning process of the students and applies reinforcement learning techniques to provide learning guidance based on the multi-modal sensing data in smart classroom. More specifically, the smart learning recommendation system measures the heartbeats, quiz scores, blinks and facial expressions of each student to formulate the learning states and applies reinforcement learning to recommend the effective learning activities for students based on their current learning states. The interactive learning recommendation process in a smart classroom with multiple sensors can be modeled as a Markov decision process. Our simulation results have preliminarily demonstrated the effectiveness of this smart learning recommendation system. This work may provide insights into constructing a future intelligent learning environment for enriched personalized experiences.