Multimodal motivation modelling and computing towards motivationally intelligent E-learning systems

Abstract Motivation to engage in learning is essential for performance. Learners’ motivation traditionally assessed using self-reported data, which time-consuming, subjective, and interruptive their process. To address this issue, paper proposes a novel framework multimodal assessment of learners’ e-learning environments with the ultimate purpose supporting intelligent systems facilitate dynamic, context-aware, personalized services or interventions, thus sustaining engagement. We investigated performance machine classifier most least accurately predicted motivational factors. also contribution different electroencephalogram (EEG) eye gaze features assessment. The applicability was evaluated an empirical study we combined tracking EEG sensors produce dataset. dataset then processed used develop by predicting levels range factors, represented multiple dimensions motivation. proposed approach feature selection combining data-driven knowledge-driven methods train assessment, has been proved effective our at selecting predictors from large number extracted data. Our revealed valuable insights role played brain activities movements on Initial results logistic regression have achieved significant predictive power all factors studied, accuracy between 68.1% 92.8%. present work demonstrated will inspire future research towards motivationally systems.