LewiSpace: An Educational Puzzle Game Combined with a Multimodal Machine Learning Environment

摘要

In this paper, we will present an educational game that we developed in order to teach a chemistry lesson, namely drawing a Lewis diagram. We also conducted an experiment to gather data about the cognitive and emotional states of the learners as well as their behaviour throughout our game by using three types of sensors (electroencephalography, eye tracking, and facial expression recognition with an optical camera). Primary results show that a machine learning model (logistic regression) can predict with some success whether the learner will give a correct or a wrong answer to a task presented in the game, and paves the way for an adaptive version of the game. This latter will challenge or assist learners based on some features extracted from our data in order to provide real-time adaptation specific to the user.