Automatic Mind Wandering Detection in Educational Settings: A Systematic Review and Multimodal Benchmarking A Preprint Journal: Computers & Education
Anna Bodonhelyi, Augustin Curinier, Babette Bühler, Gerrit Anders, Lisa Rausch, Markus Huff, Ulrich Trautwein, Ralph Ewerth, Peter Gerjets, Enkelejda Kasneci · 2026
Read Automatic Mind Wandering Detection in Educational Settings: A Systematic Review and Multimodal Benchmarking A Preprint Journal: Computers & Education by Anna Bodonhelyi, Augustin Curinier, Babette Bühler, Gerrit Anders, Lisa Rausch, Markus Huff, Ulrich Trautwein, Ralph Ewerth, Peter Gerjets, Enkelejda Kasneci (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-09569.txt.
Introduction
Bodonhelyi and nine colleagues review and benchmark mind-wandering detection in online education, where minds drift about 30 percent of the time. Progress has lagged amid mixed methods, varied data prep, and inconsistent metrics. The team builds one shared pipeline tuned to each signal type. It spans EEG, facial video, eye tracking, and body signals across 14 datasets. Thirteen traditional and neural models, including federated learning, are tested on every dataset. A post-probe test explores how learners re-engage after drifting, through re-reading or re-watching.
Catalog record
| CREATOR | Anna Bodonhelyi, Augustin Curinier, Babette Bühler, Gerrit Anders, Lisa Rausch, Markus Huff, Ulrich Trautwein, Ralph Ewerth, Peter Gerjets, Enkelejda Kasneci |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 89044 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.09569
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Automatic Mind Wandering Detection in Educational Settings: A Systematic Review and Multimodal Benchmarking A Preprint Journal: Computers & Education Anna Bodonhelyi, Augustin Curinier, Babette Bühler, Gerrit Anders, Lisa Rausch, Markus Huff, Ulrich Trautwein, Ralph Ewerth, Peter Gerjets, Enkelejda Kasneci Abstract Detecting mind wandering is crucial in online education, and it occurs 30% of the time, as it directly impacts learners’ retention, comprehension, and overall success in self-directed learning environments. Integrating automated detection algorithms enables the deployment of targeted interventions within adaptive learning environments, paving the way for more responsive and personalized educational systems. However, progress is hampered by a lack of coherent frameworks for identifying mind wandering in online environments. This work presents a comprehensive systematic review and benchmark of mind wandering detection across 14 datasets covering EEG, facial video, eye tracking, and physiological signals in educational settings, motivated by the challenges in achieving reliable detection and the inconsistency of results across studies caused by variations in models, preprocessing approaches, and evaluation metrics. We implemented a generalizable preprocessing and feature extraction pipeline tailored to each modality, ensuring fair comparison across diverse experimental paradigms. 13 traditional machine learning and neural network models, including federated learning approaches, were evaluated on each dataset. In a novel ablation study, we explored mind wandering detection from post-probe data, motivated by findings that learners often re-engage with material after mind wandering episodes through re-reading or re-watching. Results highlight the potential and limitations of different modalities and classifiers for mind wandering detection, and point to new opportunities for supporting online learning. All code and preprocessing scripts are made openly available to support reproducibility and future research. Preprint submitted to Computers & Education 1 Introduction Self-directed online learning, facilitated by platforms such as MOOCs and institutional environments ( 49 ; 84 ) , increasingly takes place in informal and distributed settings where sustaining attention is particularly challenging ( 43 ; 103 ; 29 ) . In the absence of direct instructor support, learners are especially susceptible to distractions and mind wandering ( 59 ; 121 ) , a cognitive state in which attention shifts from the primary task to internal, task-unrelated thoughts ( 105 ) . This phenomenon is highly prevalent: people spend nearly half of their waking hours mind wandering (46.9%), with task-unrelated thoughts occurring in at least 30% of sampled moments across everyday activities ( 73 ) , while in learning contexts, learners mind wander during roughly 30% of study time regardless of the learning situation ( 125 ) . Similar patterns emerge in other attention-critical settings, with mind wandering reported during approximately 39–42% of driving time ( 126 ) . Recent findings indicate that nearly 50% of adolescents experience such attentional shifts frequently in daily life ( 129 ; 93 ) . These episodes are often driven by environmental stressors and misalignment between cognitive demands and context, with factors such as childhood adversity or peer conflict triggering nonlinear increases in task-unrelated thoughts ( 129 ; 2 ) . Overall, mind wandering accounts for approximately 7% of the variance in learning outcomes and is consistently associated with poorer performance across subjects and age groups. Because mind wandering substantially impairs knowledge retention ( 125 ) , there is growing interest in applying Machine Learning (ML) to detect it in real time, enabling adaptive learner support and data-driven insights for improving instructional design and digital learning outcomes. However, progress toward such applications is limited by the scarcity of publicly available datasets that are both multimodal and suitable for real-world deployment. Many existing datasets depend on costly and intrusive sensors, such as EEG headsets ( 113 ; 67 ) or dedicated eye-tracking systems ( 13 ; 32 ) , which require controlled environments and hinder scalability. In contrast, webcam-based approaches ( 20 ; 79 ) offer a more accessible and practical alternative for large-scale or at-home use, albeit typically at the cost of lower detection accuracy. Mind wandering research has advanced along two major disciplinary trajectories. In educational psychology , it is typically examined through subjective measures such as self-reports collected via experience sampling methods (e.g., probe-caught or self-caught reports) or retrospective questionnaires. In contrast, computer science approaches emphasize the automated, non-intrusive detection of mind wandering through behavioral and physiological signals, often coupled with ML techniques. These computational methods aim to enable large-scale, unobtrusive assessment, using self-reports primarily during the initial stages to establish ground truth labels for model training. Research gap There is a critical lack of systematic benchmarking across accessible sensing modalities and real-world learning contexts, which limits our understanding of which detection methods are most effective and scalable for mind wandering interventions. Bridging these fields requires a solid methodological foundation that integrates computational rigor with educational relevance. this work makes the following key contributions: 1. An up-to-date systematic review of automatic mind wandering detection research within educational contexts. 2. Introduction of a new eye-tracking dataset (available on Zenodo ). 3. A comprehensive benchmarking study with a published benchmarking framework (available on GitLab ) of widely used machine learning algorithms across 14 datasets. 4. Evidence-based recommendations identifying the most effective machine learning methods for specific data modalities used in mind wandering detection. 5. An ablation study introducing and evaluating a novel sampling approach for data collection , highlighting its impact on model performance compared to existing dataset conventions. 2 Related Work 2.1 Existing Reviews on Understanding Mind Wandering Mind wandering has been examined from multiple disciplinary perspectives, and several reviews have summarized prior work in this area. Among these, the review by 77 is the most relevant to the present study, as it focuses explicitly on automated, machine learning–based mind wandering detection. Synthesizing 42 studies, it introduces a structured framework encompassing six key dimensions: task type, ground truth collection, modality, machine learning approach, validation strategy, and evaluation metrics. However, the review is no longer fully up to date, includes studies from task domains beyond educational contexts, and reflects considerable heterogeneity in evaluation and validation practices, which limits cross-study comparability. While their review offers a multidisciplinary overview of existing systematic reviews on mind wandering, in the following we synthesize the most pertinent literature from the perspectives of educational psychology and computer science. Education More recent work has shifted toward understanding mind wandering in educational settings. 41 conducted an interdisciplinary review highlighting how mind wandering impairs reading comprehension. A complementary meta-analysis by 87 focused on eye-movement correlates of mind wandering in 16 studies, revealing behavioral patterns. 124 expanded this scope through a meta-analysis of 130 samples, demonstrating that task-unrelated thoughts are prevalent in learning contexts and consistently associated with poorer outcomes. These reviews highlight the significance of mind wandering, but largely concentrate on its consequences rather than its detection. Cognitive Science The study of mind wandering has long attracted attention in cognitive science, with early reviews focusing on its theoretical foundations and underlying neural mechanisms. For example, 107 emphasized the role of decoupled processing during reading, where external attention fades in favor of internal thought. 50 conducted a meta-analysis of 24 neuroimaging studies, finding frequent activation in the default mode network during spontaneous thought, though other brain regions also contributed. Further exploring the content of mind wandering, 10 re-analyzed three earlier studies ( 9 ; 11 ; 8 )