Real-Time Quantification of Cognitive Load in Learning Environments Using Multimodal AI: A Dynamic Framework for Continuous Mental Effort Estimation

Accurate measurement of cognitive load remains a central challenge in educational psychology, particularly in dynamic learning environments where mental effort fluctuates continuously. Traditional assessment methods, including self-report scales and post-task evaluations, fail to capture real-time variations and often suffer from subjectivity and temporal bias. This study proposes a novel multimodal artificial intelligence framework for continuous, real-time estimation of cognitive load by integrating physiological, behavioral, and contextual data streams. The proposed framework combines eye-tracking metrics, electroencephalography signals, facial expression dynamics, and interaction-based behavioral features within a unified deep learning architecture. A hybrid model integrating temporal convolutional networks and attention-based fusion mechanisms is developed to capture both intra-modality temporal dependencies and cross-modality interactions. Data were collected from ۱۲۰ participants engaged in structured learning tasks with varying intrinsic and extraneous cognitive load levels. Ground truth labels were established using a combination of validated subjective scales and expert annotations. The model achieved a prediction accuracy of ۹۱.۳% and a root mean square error of ۰.۱۸ in continuous cognitive load estimation, outperforming baseline unimodal and traditional machine learning approaches. Sensitivity analysis revealed that eye-tracking and EEG signals contributed most significantly to model performance, while behavioral features enhanced temporal stability. The proposed system demonstrates strong generalization across task types and learner profiles, indicating its potential for scalable deployment in adaptive learning systems. This study advances the field by introducing a real-time, data-driven approach to cognitive load measurement, enabling fine-grained monitoring of mental effort. The findings support the integration of multimodal AI systems into intelligent tutoring platforms to optimize instructional design, reduce cognitive overload, and enhance learning efficiency.

کلیدواژه ها:

Cognitive Load ، Multimodal Learning Analytics ، Artificial Intelligence in Education ، Real-Time Mental Effort Estimation ، Eye-Tracking and EEG Integration ، Adaptive Learning Systems