EEG complexity measures for detecting mind wandering during video-based learning

Tang, Shaohua; Li, Zheng · 2024 · Crossref

DOI: 10.1038/s41598-024-58889-9

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study investigates the efficacy of EEG complexity measures for detecting mind wandering (MW) during video-based learning, addressing the challenge of maintaining student engagement in educational settings. While mind wandering is known to detract from learning outcomes, existing detection methods often rely on linear features like band power or event-related potentials, which may lack flexibility or require controlled laboratory conditions. The authors aim to determine if nonlinear complexity metrics, which characterize the brain’s spatiotemporal dynamics, can effectively distinguish MW from non-MW states in a more naturalistic, video-learning context. The researchers collected EEG data from 28 participants viewing educational videos using a modified probe-caught method. To increase the capture rate of mind wandering episodes, an experimenter manually triggered additional probes based on observed facial expressions, supplementing random automated probes. EEG signals were pre-processed using Artifact Subspace Reconstruction (ASR) and, in some pipelines, Independent Component Analysis (ICA) to remove eye-movement artifacts. The study systematically evaluated six complexity metrics: three regularity measures (multiscale sample entropy, permutation entropy, and dispersion entropy) and three temporal dimensionality measures (Higuchi’s fractal dimension, Katz’s fractal dimension, and detrended fluctuation analysis). These were compared against traditional band power features. To handle limited sample sizes and class imbalance, the authors employed sample augmentation via overlapping windows and synthetic minority over-sampling, followed by feature selection using minimum redundancy and maximum relevance. Results indicated that traditional band power features achieved the highest performance (mean AUC 0.646) when eye-movement artifacts were removed. However, multiscale permutation entropy (MPE) demonstrated comparable performance (mean AUC 0.639) without requiring the removal of eye-movement artifacts, suggesting it is more robust to noise and easier to implement in real-world scenarios. Combining all feature types improved decoding performance to a mean AUC of 0.66. The study highlights that while complexity metrics alone may not surpass optimized linear features in clean data, their resilience to artifacts and the benefits of feature fusion make them valuable for practical applications. The findings underscore the potential of EEG complexity measures for real-time mind wandering detection in educational environments. By demonstrating that MPE can perform well without rigorous artifact rejection, the study simplifies the preprocessing pipeline, facilitating the development of adaptive educational systems. These systems could monitor student attention and respond to lapses in real-time, thereby enhancing learning outcomes. The research contributes to the field by providing a systematic comparison of complexity metrics and validating their utility in naturalistic, video-based learning contexts.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-nvidia on 2026-08-10; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• neuro workload indices

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Empirical Findings: physiological data