.png){kind=icon}
.png){kind=icon}
Return
Cognitive load in e-learning: how different environments shape learning efficiency
.png){kind=icon}
PRE
AI
DOI：10.1080/01587919.2026.2697165.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
This study examines how physical e-learning environments affect cognitive load (CL) and learning performance. Using multimodal physiological signals—electroencephalography (EEG), galvanic skin response (GSR), and photoplethysmography (PPG)—alongside environmental monitoring and self-reported workload, we assessed CL across four naturalistic settings: a study room, library, coffee shop, and park. Forty-eight participants watched an instructional video while physiological and environmental data were recorded in real time. CL scores were estimated via an large language model (LLM) framework using structured multimodal physiological inputs. Results show that physical surroundings significantly affect CL: the park imposed substantially higher cognitive demand than structured indoor settings, while the coffee shop showed a non-significant upward trend. Among environmental predictors, visual complexity—quantified via computer vision-based object detection—emerged as the strongest correlate (r = 0.77), followed by social density (r = 0.61) and ambient noise (r = 0.60). LLM-predicted CL correlated strongly with NASA-TLX self-reports across all conditions (r = 0.68–0.95), confirming convergent validity. This study introduces a scalable, ecologically valid LLM framework for multimodal CL estimation, with direct implications for designing adaptive, cognitively supportive digital learning environments.
Keywords:
Cognitive load
e-learning environments
physiological measurement
large language models
multimodal sensing