Detection of mind wandering using EEG: Within and across individuals

Dong, Henry W.; Mills, Caitlin; Knight, Robert T.; Kam, Julia W. Y. · 2021 · Crossref

DOI: 10.1371/journal.pone.0251490

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 addresses the challenge of reliably detecting mind wandering, a phenomenon characterized by attention shifting from external tasks to internal thoughts, which is linked to performance errors and negative affect. Current detection methods rely heavily on self-reported thought sampling, which is intrusive and subject to bias. The authors aimed to determine if machine learning models using electroencephalography (EEG) data could accurately predict attention states both within individuals (person-dependent) and across individuals (person-independent), thereby enabling unobtrusive, real-time detection. Fourteen participants performed an auditory target detection task while scalp EEG was recorded. Participants responded to pseudorandom thought probes indicating whether they were "on task" or "mind wandering." The researchers extracted event-related potential (ERP) features, specifically the N1 and P3 components, from the ten trials preceding each probe. They computed the mean and standard deviation of these ERP amplitudes for each block. To classify attention states, the authors employed two machine learning models: a linear logistic regression and a non-linear support vector machine (SVM) with a radial basis function kernel. Data were normalized to account for individual differences, and class imbalance was addressed using Synthetic Minority Over-Sampling Technique (SMOTE). The results demonstrated that both models successfully classified attention states above chance levels. For within-subject classification, the SVM achieved an area under the curve (AUC) of 0.715, while logistic regression achieved an AUC of 0.635. Crucially, the models also generalized across subjects, with the SVM achieving an AUC of 0.613 and logistic regression an AUC of 0.609. This indicates that ERP patterns observed in a group can reliably predict the attention state of a new, unseen individual. The significance of this work lies in its demonstration that machine learning models can generalize to "never-seen-before" individuals using electrophysiological measures. This is the first study to establish such cross-individual generalizability for mind wandering detection using EEG. These findings highlight the potential for developing real-time, unobtrusive systems to monitor covert attention states, which could eventually reduce reliance on intrusive self-report methods and help mitigate the negative impacts of mind wandering in daily life.

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
• mind wandering