pith. sign in

arxiv: 2604.09569 · v1 · submitted 2026-02-20 · 💻 cs.HC

Automatic Mind Wandering Detection in Educational Settings: A Systematic Review and Multimodal Benchmarking

Anna Bodonhelyi , Augustin Curinier , Babette B\"uhler , Gerrit Anders , Lisa Rausch , Markus Huff , Ulrich Trautwein , Ralph Ewerth
show 2 more authors
Peter Gerjets Enkelejda Kasneci
This is my paper

Pith reviewed 2026-05-15 21:00 UTC · model grok-4.3

classification 💻 cs.HC
keywords mind wandering detectioneducational technologymultimodal signalsmachine learning benchmarkEEGeye trackingsystematic review
0
0 comments X

The pith

A standardized pipeline across 14 datasets reveals the comparative performance of EEG, eye tracking, and other signals for detecting mind wandering during online learning.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper carries out a systematic review and multimodal benchmark to create a consistent way of comparing mind wandering detection methods in educational contexts. It applies one shared preprocessing and feature extraction pipeline to datasets covering EEG, facial video, eye tracking, and physiological signals, then evaluates 13 machine learning and neural network models including federated approaches. A separate ablation examines detection from post-probe data after learners re-engage with material. The results map out which signal types and classifiers show promise while exposing their current limits for building adaptive online education tools.

Core claim

By enforcing a single generalizable preprocessing pipeline on 14 datasets and testing the same set of 13 models on each, the study produces directly comparable performance figures across modalities and shows that certain signals support reliable detection while others remain limited, with federated learning offering a privacy-preserving option and post-probe analysis adding insight into learner re-engagement patterns.

What carries the argument

A single generalizable preprocessing and feature extraction pipeline applied uniformly to each modality before model evaluation.

If this is right

  • Standardized evaluation makes it possible to identify which modalities and classifiers are ready for deployment in adaptive learning systems.
  • Federated learning models allow detection without requiring raw data to leave local devices.
  • Post-probe data analysis shows detection can track both initial lapses and subsequent re-engagement.
  • Open release of code and scripts enables direct replication and extension to new datasets.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The benchmark could be used to select signal combinations that improve accuracy in noisy real-time classroom recordings.
  • Limitations observed across current datasets point to the value of collecting more varied data from different age groups and learning platforms.
  • Integrating the top-performing detectors into learning platforms could trigger automatic content pauses or summaries when mind wandering is flagged.

Load-bearing premise

The 14 selected datasets plus one shared preprocessing pipeline together capture enough real-world variation in educational settings without introducing systematic biases across signal types.

What would settle it

A fresh dataset recorded from live online lectures with a broad student population that produces detection accuracies well below the levels reported in the benchmark when the same pipeline and models are applied.

Figures

Figures reproduced from arXiv: 2604.09569 by Anna Bodonhelyi, Augustin Curinier, Babette B\"uhler, Enkelejda Kasneci, Gerrit Anders, Lisa Rausch, Markus Huff, Peter Gerjets, Ralph Ewerth, Ulrich Trautwein.

Figure 1
Figure 1. Figure 1: PRISMA (Moher et al., 2009) diagram for systematic search. [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Summary of our findings in the systematic review with research gaps that created the need for our modular benchmarking framework. [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Review summary in three categories. Skin-based signals refer to physiological measurements such as EDA, skin conductance, and [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Data preprocessing and model architecture. [PITH_FULL_IMAGE:figures/full_fig_p010_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Example sample extraction in our main experiments and ablation study. [PITH_FULL_IMAGE:figures/full_fig_p017_5.png] view at source ↗
read the original 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.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 2 minor

Summary. The manuscript presents a systematic review and multimodal benchmarking study of automatic mind wandering detection in educational settings. It covers 14 datasets spanning EEG, facial video, eye tracking, and physiological signals; implements a single generalizable preprocessing and feature extraction pipeline (tailored per modality); evaluates 13 traditional machine learning and neural network models (including federated learning approaches) on each dataset; and conducts a novel ablation study using post-probe data. The work aims to address inconsistencies in prior results caused by varying models, preprocessing, and metrics, with all code and scripts released openly. Results are presented as highlighting the potential and limitations of modalities and classifiers for supporting online learning.

Significance. If the benchmarking results hold under the chosen pipeline, the paper supplies a coherent, reproducible baseline and framework that can reduce fragmentation in mind wandering detection research. The open release of code and preprocessing scripts is a clear strength for verifiability and extension, directly supporting the claim of enabling future work on adaptive educational interventions.

major comments (1)
  1. [§3 (Preprocessing and Feature Extraction)] §3 (Preprocessing and Feature Extraction): The central claim that the results reliably highlight modality and classifier differences rests on the single generalizable pipeline producing fair cross-dataset comparisons. Because the 14 datasets differ substantially in experimental paradigms, probe timings, and signal characteristics, the absence of any validation (e.g., comparison against modality-optimized alternatives) leaves open the possibility that observed performance gaps partly reflect pipeline artifacts rather than intrinsic capabilities. This assumption is load-bearing for the benchmarking conclusions.
minor comments (2)
  1. [Table 1] Table 1 or equivalent dataset summary: explicitly state the number of samples and class balance per modality to allow readers to assess whether aggregated metrics are dominated by particular datasets.
  2. [Results] Results section: the post-probe ablation is described as novel, but a brief comparison to the main probe-based results (e.g., average F1 delta) would strengthen the interpretation of re-engagement effects.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We address the major comment below, providing a point-by-point response while maintaining the integrity of our benchmarking design.

read point-by-point responses

  1. Referee: The central claim that the results reliably highlight modality and classifier differences rests on the single generalizable pipeline producing fair cross-dataset comparisons. Because the 14 datasets differ substantially in experimental paradigms, probe timings, and signal characteristics, the absence of any validation (e.g., comparison against modality-optimized alternatives) leaves open the possibility that observed performance gaps partly reflect pipeline artifacts rather than intrinsic capabilities. This assumption is load-bearing for the benchmarking conclusions.

    Authors: We appreciate this observation on the foundational role of our preprocessing pipeline. The study's primary aim is to establish a coherent, reproducible baseline by applying one generalizable pipeline (tailored per modality) across all 14 datasets. This ensures that performance differences arise from modalities and models rather than methodological variations, directly addressing the fragmentation noted in prior literature. Introducing dataset-specific optimizations would confound cross-dataset comparisons and undermine the goal of a standardized framework. We will revise the manuscript to expand the justification for this design in §3 and add an explicit discussion of its implications and limitations. revision: partial

Circularity Check

0 steps flagged

Empirical benchmarking study with no circular derivations

full rationale

This paper is a systematic review and empirical benchmarking study that evaluates 13 ML models across 14 datasets using a fixed preprocessing pipeline per modality. All reported results on modality potentials, limitations, and classifier performance derive directly from data-driven evaluation and ablation experiments rather than any mathematical derivation, parameter fitting, or self-citation chain that reduces the central claims to the inputs by construction. No load-bearing steps match the enumerated circularity patterns.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The paper is an empirical review and benchmark that relies on existing public datasets and standard machine-learning evaluation practices without introducing new free parameters, theoretical axioms, or invented entities.

axioms (1)
  • domain assumption Standard machine-learning assumptions such as representative train-test splits and i.i.d. sampling hold for the chosen datasets.
    Implicit in the benchmarking protocol described in the abstract.

pith-pipeline@v0.9.0 · 5588 in / 1187 out tokens · 37480 ms · 2026-05-15T21:00:33.169136+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

72 extracted references · 72 canonical work pages

  1. [1]

    URL:https:// www.sciencedirect.com/science/article/pii/S0165027017300997, doi:10.1016/j.jneumeth.2017

    Classification of EEG signals to identify variations in attention during motor task execution 284, 27–34. URL:https:// www.sciencedirect.com/science/article/pii/S0165027017300997, doi:10.1016/j.jneumeth.2017. 04.008. Andrewes, D.G., Jenkins, L.M.,

    work page doi:10.1016/j.jneumeth.2017 2017

  2. [2]

    Neuropsychology review 29, 220–243

    The role of the amygdala and the ventromedial prefrontal cortex in emotional regulation: implications for post-traumatic stress disorder. Neuropsychology review 29, 220–243. doi:https: //doi.org/10.1007/s11065-019-09398-4. Andrews-Hanna, J., Irving, Z.C., Fox, K., Spreng, R.N., Christoff, K., . The neuroscience of spontaneous thought: An evolving, interdi...

    work page doi:10.1007/s11065-019-09398-4

  3. [3]

    URL:https://www.sciencedirect.com/science/ article/pii/S0097849322001856, doi:10.1016/j.cag.2022.10.007

    Detecting distracted students in educational VR environments using machine learning on eye gaze data 109, 75–87. URL:https://www.sciencedirect.com/science/ article/pii/S0097849322001856, doi:10.1016/j.cag.2022.10.007. Asish, S.M., Kulshreshth, A.K., Borst, C.W., Sutradhar, S.,

    work page doi:10.1016/j.cag.2022.10.007 2022

  4. [4]

    Precise tool to target positioning widgets (totta) in spatial environments: A systematic review,

    Classification of internal and external distractions in an educational VR environment using multimodal features 30, 7332–7342. URL:https://ieeexplore. ieee.org/document/10670586, doi:10.1109/TVCG.2024.3456207. conference Name: IEEE Transactions on Visualization and Computer Graphics. Baldi, P., Brunak, S., Frasconi, P., Soda, G., Pollastri, G.,

    work page doi:10.1109/tvcg.2024.3456207 2024

  5. [5]

    Bioinformatics 15, 937–946

    Exploiting the past and the future in protein secondary structure prediction. Bioinformatics 15, 937–946. doi:https://doi.org/10.1093/bioinformatics/15.11

    work page doi:10.1093/bioinformatics/15.11

  6. [6]

    Openface 2.0: Facial behavior analysis toolkit, in: 2018 13th IEEE international conference on automatic face & gesture recognition (FG 2018), IEEE. pp. 59–66. doi:10.1109/FG.2018.00019. Banks, J.B., Boals, A., . Understanding the role of mind wandering in stress-related working memory impairments 31, 1023–1030. URL:https://doi.org/10.1080/02699931.2016.1...

    work page doi:10.1109/fg.2018.00019 2018

  7. [7]

    IEEE Transactions on Neural Networks 5, 537–550

    Using mutual information for selecting features in supervised neural net learning. IEEE Transactions on Neural Networks 5, 537–550. doi:10.1109/72.298224. Bixler, R., Blanchard, N., Garrison, L., D’Mello, S.,

    work page doi:10.1109/72.298224

  8. [8]

    Automatic detection of mind wandering during reading us- ing gaze and physiology, in: Proceedings of the 2015 ACM on International Conference on Multimodal Interaction - ICMI ’15, ACM Press, Seattle, Washington, USA. pp. 299–306. doi:10.1145/2818346.2820742. 20 Bixler, R., D’Mello, S.,

    work page doi:10.1145/2818346.2820742 2015

  9. [9]

    User Modeling and User-Adapted Interaction 26.0, 33–68

    Automatic gaze-based user-independent detection of mind wandering during comput- erized reading. User Modeling and User-Adapted Interaction 26.0, 33–68. doi:10.1007/s11257-015-9167-1. Bixler, R.E., D’Mello, S.K.,

    work page doi:10.1007/s11257-015-9167-1

  10. [10]

    ACM Symposium on Eye Tracking Research and Applications , 1–12doi:10.1145/3448017.3457386

    Crossed eyes: Domain adaptation for gaze-based mind wandering models. ACM Symposium on Eye Tracking Research and Applications , 1–12doi:10.1145/3448017.3457386. Bodonhelyi, A., Thaqi, E., Özdel, S., Bozkir, E., Kasneci, E.,

    work page doi:10.1145/3448017.3457386

  11. [11]

    From passive watching to active learning: Empowering proactive participation in digital classrooms with ai video assistant, in: Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems, pp. 1–21. doi:https://doi.org/10.1145/3706598. 3713513. Bodonhelyi, A., Wang, M., Bozkir, E., Bühler, B., Kasneci, E.,

    work page doi:10.1145/3706598 2025

  12. [12]

    arXiv preprint arXiv:2602.09904

    Safeguarding privacy: Privacy-preserving detection of mind wandering and disengagement using federated learning in online education. arXiv preprint arXiv:2602.09904 . Bosch, N., D’Mello, S.K., 2021a. Automatic detection of mind wandering from video in the lab and in the classroom. IEEE Transactions on Affective Computing 12, 974–988. doi:10.1109/TAFFC.201...

    work page doi:10.1109/taffc.2019.2908837 2019

  13. [13]

    Bühler, B., Bozkir, E., Deininger, H., Gerjets, P., Trautwein, U., Kasneci, E.,

    doi:10.3390/s20092546. Bühler, B., Bozkir, E., Deininger, H., Gerjets, P., Trautwein, U., Kasneci, E.,

    work page doi:10.3390/s20092546

  14. [14]

    Proceedings of the ACM on Human-Computer Interaction 8, 1–18

    On task and in sync: Examining the relationship between gaze synchrony and self-reported attention during video lecture learning. Proceedings of the ACM on Human-Computer Interaction 8, 1–18. Bühler, B., Bozkir, E., Goldberg, P., Sümer, Ö., D’Mello, S., Gerjets, P., Trautwein, U., Kasneci, E., 2024a. From the lab to the wild: Examining generalizability of...

    work page doi:10.1007/s40593-024-00412-2

  15. [15]

    Bühler, B., Bozkir, E., Deininger, H., Goldberg, P., Gerjets, P., Trautwein, U., Kasneci, E., 2024b

    doi:https://doi.org/10.3390/educsci11120785. Bühler, B., Bozkir, E., Deininger, H., Goldberg, P., Gerjets, P., Trautwein, U., Kasneci, E., 2024b. Detecting aware and unaware mind wandering during lecture viewing: A multimodal machine learning approach using eye tracking, facial videos and physiological data, in: International Conference on Multimodel Inte...

    work page doi:10.3390/educsci11120785

  16. [16]

    2021 , journal =

    Efficient mind-wandering detection system with gsr signals on mm-sart database. 2021 IEEE Workshop on Signal Processing Systems (SiPS) , 199–204doi:10.1109/SiPS52927.2021. 00043. Chen, T., Guestrin, C.,

    work page doi:10.1109/sips52927.2021 2021

  17. [17]

    IEEE Journal of Biomedical and Health Informatics 26.0, 3649–3660

    An effective entropy-assisted mind-wandering detection system using eeg signals of mm-sart database. IEEE Journal of Biomedical and Health Informatics 26.0, 3649–3660. doi:10.1109/JBHI.2022.3187346. Chiossi, F., Ou, C., Gerhardt, C., Putze, F., Mayer, S.,

    work page doi:10.1109/jbhi.2022.3187346 2022

  18. [18]

    URL:https://www

    Designing and evaluating an adaptive virtual reality system using EEG frequencies to balance internal and external attention states 196, 103433. URL:https://www. sciencedirect.com/science/article/pii/S1071581924002167, doi:10.1016/j.ijhcs.2024.103433. Cortes, C., Vapnik, V .,

    work page doi:10.1016/j.ijhcs.2024.103433 2024

  19. [19]

    Learning Technologies and Globalization: Pedagogical Frameworks and Applications , 9–13doi:https://doi.org/10.1007/978-3-319-22963-8_2

    Massive open online courses. Learning Technologies and Globalization: Pedagogical Frameworks and Applications , 9–13doi:https://doi.org/10.1007/978-3-319-22963-8_2. Dhindsa, K., Acai, A., Wagner, N., Bosynak, D., Kelly, S., Bhandari, M., Petrisor, B., Sonnadara, R.R.,

    work page doi:10.1007/978-3-319-22963-8_2

  20. [20]

    PLOS ONE 14.0, e0222276

    In- dividualized pattern recognition for detecting mind wandering from eeg during live lectures. PLOS ONE 14.0, e0222276. doi:10.1371/journal.pone.0222276. D’Mello, S.K., Mills, C.S.,

    work page doi:10.1371/journal.pone.0222276

  21. [21]

    Language and Linguistics Compass 15, e12412

    Mind wandering during reading: An interdisciplinary and integrative review of psychological, computing, and intervention research and theory. Language and Linguistics Compass 15, e12412. doi:10.1111/lnc3.12412. _eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1111/lnc3.12412. Dong, H.W., Mills, C., Knight, R.T., Kam, J.W.Y .,

    work page doi:10.1111/lnc3.12412

  22. [22]

    PLOS ONE 16.0, e0251490

    Detection of mind wandering using eeg: Within and across individuals. PLOS ONE 16.0, e0251490. doi:10.1371/journal.pone.0251490. Ebbert, D.,

    work page doi:10.1371/journal.pone.0251490

  23. [23]

    Advances in Human- Computer Interaction 2021, 1–10

    How did the covid-19 pandemic affect higher education learning experience? an empirical investigation of learners’ academic performance at a university in a developing country. Advances in Human- Computer Interaction 2021, 1–10. doi:https://doi.org/10.1155/2021/6649524. European Commission,

    work page doi:10.1155/2021/6649524 2021

  24. [24]

    URL:https://digital-strategy.ec.europa.eu/en/ policies/regulatory-framework-ai

    Artificial intelligence act. URL:https://digital-strategy.ec.europa.eu/en/ policies/regulatory-framework-ai. accessed: 2025-06-16. Faber, M., Bixler, R., D’Mello, S.K., 2018a. An automated behavioral measure of mind wandering during computer- ized reading. Behavior Research Methods 50.0, 134–150. doi:10.3758/s13428-017-0857-y. Faber, M., Krasich, K., Bixl...

    work page doi:10.3758/s13428-017-0857-y 2025

  25. [25]

    Journal of Experimental Psychology: Human Perception and Performance 46, 1201–1221

    The eye–mind wandering link: Identi- fying gaze indices of mind wandering across tasks. Journal of Experimental Psychology: Human Perception and Performance 46, 1201–1221. doi:10.1037/xhp0000743. Faber, M., Radvansky, G.A., D’Mello, S.K., 2018b. Driven to distraction: A lack of change gives rise to mind wandering. Cognition 173, 133–137. doi:https://doi.o...

    work page doi:10.1037/xhp0000743 2018

  26. [26]

    v10i5.643

    doi:https://doi.org/10.19173/irrodl. v10i5.643. Fox, K.C.R., Spreng, R.N., Ellamil, M., Andrews-Hanna, J.R., Christoff, K., . The wandering brain: meta-analysis of functional neuroimaging studies of mind-wandering and related spontaneous thought processes 111, 611–621. doi:10.1016/j.neuroimage.2015.02.039. Friedman, J.H.,

    work page doi:10.19173/irrodl 2015

  27. [27]

    Annals of statistics , 1189– 1232doi:http://www.jstor.org/stable/2699986

    Greedy function approximation: a gradient boosting machine. Annals of statistics , 1189– 1232doi:http://www.jstor.org/stable/2699986. Fütterer, T., Goldberg, P., Bühler, B., Sikimi ´c, V ., Trautwein, U., Gerjets, P., Stürmer, K., Kasneci, E.,

    work page arXiv

  28. [28]

    Computers and Education: Artificial Intelligence 9, 100483

    Ar- tificial intelligence in classroom management: A systematic review on educational purposes, technical imple- mentations, and ethical considerations. Computers and Education: Artificial Intelligence 9, 100483. URL: https://www.sciencedirect.com/science/article/pii/S2666920X25001237, doi:https://doi.org/ 10.1016/j.caeai.2025.100483. Glorot, X., Bordes, ...

    work page doi:10.1016/j.caeai.2025.100483 2025

  29. [29]

    Guo, X., Hosseini, S.,

    doi:https://doi.org/10.3389/fpsyg.2014.00031. Guo, X., Hosseini, S.,

    work page doi:10.3389/fpsyg.2014.00031 2014

  30. [30]

    Proceedings of the 10th ACM International Conference on Bioinformatics, Computational Biology and Health Informatics , 314–319doi:10.1145/3307339.3342176

    Deep convolutional neural network for automated detection of mind wandering using eeg signals. Proceedings of the 10th ACM International Conference on Bioinformatics, Computational Biology and Health Informatics , 314–319doi:10.1145/3307339.3342176. Gwizdka, Jacek,

    work page doi:10.1145/3307339.3342176

  31. [31]

    Neural computation 9, 1735–1780

    Long short-term memory. Neural computation 9, 1735–1780. doi:10.1162/ neco.1997.9.8.1735. Hollis, R.B., Was, C.A.,

    work page 1997

  32. [32]

    Learning and Instruction 42, 104–112

    Mind wandering, control failures, and social media distractions in online learning. Learning and Instruction 42, 104–112. doi:https://doi.org/10.1016/j.learninstruc.2016.01.007. Hosmer Jr, D.W., Lemeshow, S., Sturdivant, R.X.,

    work page doi:10.1016/j.learninstruc.2016.01.007 2016

  33. [33]

    out of the fr-eye-ing pan

    "out of the fr-eye-ing pan": Towards gaze-based models of attention during learning with technology in the classroom. Proceedings of the 25th Confer- ence on User Modeling, Adaptation and Personalization , 94–103doi:10.1145/3079628.3079669. Hutt, S., Mills, C., White, S., Donnelly, P.J., D’Mello, S.K.,

    work page doi:10.1145/3079628.3079669

  34. [34]

    One size does not fit all: Considerations when using webcam-based eye tracking to models of neurodivergent learners’ attention and comprehension, in: Proceedings of the 15th International Learning Analytics and Knowledge Conference, Association for Computing Machinery. pp. 24–35. URL:https://dl.acm.org/doi/10.1145/3706468.3706472, doi:10.1145/3706468.3706...

    work page doi:10.1145/3706468.3706472

  35. [35]

    Cognitive, Affective, & Behavioral Neuroscience 19.0, 1059–1073

    Predicting task-general mind-wandering with eeg. Cognitive, Affective, & Behavioral Neuroscience 19.0, 1059–1073. doi:10.3758/s13415-019-00707-1. Jin, C.Y ., Borst, J.P., Van Vugt, M.K.,

    work page doi:10.3758/s13415-019-00707-1

  36. [36]

    Journal of Neural Engineering 20, 026024

    Decoding study-independent mind-wandering from eeg using convolu- tional neural networks. Journal of Neural Engineering 20, 026024. doi:10.1088/1741-2552/acc613. Kane, M.J., Smeekens, B.A., von Bastian, C.C., Lurquin, J.H., Carruth, N.P., Miyake, A.,

    work page doi:10.1088/1741-2552/acc613

  37. [37]

    doi:10.1037/xge0000362

    A combined ex- perimental and individual-differences investigation into mind wandering during a video lecture 146, 1649–1674. doi:10.1037/xge0000362. place: US Publisher: American Psychological Association. Kasneci, E., Gao, H., Ozdel, S., Maquiling, V ., Thaqi, E., Lau, C., Rong, Y ., Kasneci, G., Bozkir, E.,

    work page doi:10.1037/xge0000362

  38. [38]

    arXiv preprint arXiv:2404.15435

    In- troduction to eye tracking: A hands-on tutorial for students and practitioners. arXiv preprint arXiv:2404.15435 . Kawashima, I., Kumano, H.,

    work page arXiv

  39. [39]

    Khan, A.R., Bokhari, S.M., Khosravi, S., Hussain, S., Ghannam, R., Imran, M.A., Zoha, A., 2022a

    doi:10.3389/fnhum.2017.00365. Khan, A.R., Bokhari, S.M., Khosravi, S., Hussain, S., Ghannam, R., Imran, M.A., Zoha, A., 2022a. Feature selection mechanism for attention classification using gaze tracking data. 2022 29th IEEE International Conference on Electronics, Circuits and Systems (ICECS) , 1–4doi:10.1109/ICECS202256217.2022.9970936. Khan, A.R., Khos...

    work page doi:10.3389/fnhum.2017.00365 2017

  40. [40]

    Psychonomic bulletin & review 23, 842–848

    Mind wandering during film comprehension: The role of prior knowl- edge and situational interest. Psychonomic bulletin & review 23, 842–848. doi:https://doi.org/10.3758/ s13423-015-0936-y. Kuvar, V ., Blanchard, N., Colby, A., Allen, L., Mills, C., 2023a. Automatically detecting task-unrelated thoughts during conversations using keystroke analysis. User M...

    work page doi:10.1007/s11257-022-09340-z

  41. [41]

    URL:https://www.mdpi.com/1424-8220/ 21/22/7569, doi:10.3390/s21227569

    When eyes wander around: Mind-wandering as revealed by eye movement analysis with hidden markov models. URL:https://www.mdpi.com/1424-8220/ 21/22/7569, doi:10.3390/s21227569. 24 Lee, T., Kim, D., Park, S., Kim, D., Lee, S.J., 2022a. Predicting mind-wandering with facial videos in on- line lectures, in: Proceedings of the IEEE/CVF Conference on Computer Vi...

    work page doi:10.3390/s21227569

  42. [42]

    Li, Z., Li, F., Lee, C.H., Su, H.,

    doi:https://doi.org/10.1186/s40359-025-02667-3. Li, Z., Li, F., Lee, C.H., Su, H.,

    work page doi:10.1186/s40359-025-02667-3

  43. [43]

    doi:10.3233/ATDE240924

    Catching the wanderer: Temporal and visual analysis of mind wandering in digital learning. doi:10.3233/ATDE240924. Liu, R., Walker, E., Friedman, L., Arrington, C.M., Solovey, E.T.,

    work page doi:10.3233/atde240924

  44. [44]

    Journal on Multimodal User Interfaces 15.0, 257–272

    fnirs-based classification of mind-wandering with personalized window selection for multimodal learning interfaces. Journal on Multimodal User Interfaces 15.0, 257–272. doi:10.1007/s12193-020-00325-z. included in systematic review. Liyanagunawardena, T.R.,

    work page doi:10.1007/s12193-020-00325-z

  45. [45]

    Multimodal detection of external and internal attention in virtual reality using EEG and eye tracking features, in: Proceedings of Mensch und Computer 2024, Association for Computing Ma- chinery. pp. 29–43. URL:https://doi.org/10.1145/3670653.3670657, doi:10.1145/3670653.3670657. McMahan, B., Moore, E., Ramage, D., et al.,

    work page doi:10.1145/3670653.3670657 2024

  46. [46]

    Mollahosseini, A., Hasani, B., Mahoor, M.H.,

    doi:https://doi.org/10.1136/bmj.b2535. Mollahosseini, A., Hasani, B., Mahoor, M.H.,

    work page doi:10.1136/bmj.b2535

  47. [47]

    IEEE Transactions on Affective Computing16(4), 2560–2578 (2025) https://doi.org/10.1109/TAFFC

    Affectnet: A database for facial expression, valence, and arousal computing in the wild. IEEE Transactions on Affective Computing 10, 18–31. doi:10.1109/TAFFC. 2017.2740923. Moy, V ., Yakoub, M.S., Selouani, S.A., Alharbi, Y ., Alotaibi, Y .A.,

    work page doi:10.1109/taffc 2017

  48. [48]

    Classification of the neural correlates of mind wandering states in EEG signals, in: 2021 44th International Conference on Telecommunications and Signal Processing (TSP), IEEE. pp. 126–129. URL:https://ieeexplore.ieee.org/document/9522608/, doi:10.1109/TSP52935.2021.9522608. Mézière, D., Hautala, N., Heikkilä, T., Kaakinen, J.K., . Eye-movement markers of...

    work page doi:10.1109/tsp52935.2021.9522608 2021

  49. [49]

    org/10.1037/stl0000060

    doi:https://doi. org/10.1037/stl0000060. 25 Pain, S., Chatterjee, S., Sarma, M., Samanta, D.,

    work page doi:10.1037/stl0000060

  50. [50]

    URL:https: //www.sciencedirect.com/science/article/pii/S0045790624009303, doi:10.1016/j.compeleceng

    MSSTNet: A multi-stream time-distributed spatio-temporal deep learning model to detect mind wandering from electroencephalogram signals 122, 110005. URL:https: //www.sciencedirect.com/science/article/pii/S0045790624009303, doi:10.1016/j.compeleceng. 2024.110005. Papoutsaki, A., Sangkloy, P., Laskey, J., Daskalova, N., Huang, J., Hays, J.,

    work page doi:10.1016/j.compeleceng 2024

  51. [51]

    URL: https://www.imrpress.com/journal/JIN/24/3/10.31083/JIN26508, doi:10.31083/JIN26508

    Elevated blink rates predict mind wandering: Dopaminergic insights into attention and task focus 24, 26508. URL: https://www.imrpress.com/journal/JIN/24/3/10.31083/JIN26508, doi:10.31083/JIN26508. number: 3 Publisher: IMR Press. Robbins, H., Monro, S.,

    work page doi:10.31083/jin26508

  52. [52]

    librosa/librosa: 0.6.3,

    Asreview: Active learning for systematic reviews. doi:10.5281/zenodo. 3345592. Schuster, M., Paliwal, K.K.,

    work page doi:10.5281/zenodo

  53. [53]

    Schuster and K.K

    Bidirectional recurrent neural networks. IEEE transactions on Signal Processing 45, 2673–2681. doi:10.1109/78.650093. Dias da Silva, M.R., Postma, M.,

    work page doi:10.1109/78.650093

  54. [54]

    URL:https://www.sciencedirect.com/science/article/pii/ S0747563220302053, doi:10.1016/j.chb.2020.106453

    Wandering minds, wandering mice: Computer mouse tracking as a method to detect mind wandering 112, 106453. URL:https://www.sciencedirect.com/science/article/pii/ S0747563220302053, doi:10.1016/j.chb.2020.106453. Simamora, R.M.,

    work page doi:10.1016/j.chb.2020.106453 2020

  55. [55]

    Bellver, C

    Mind-wandering while reading: Attentional decoupling, mindless reading and the cascade model of inattention. Language and Linguistics Compass 5, 63–77. URL:https://onlinelibrary.wiley. com/doi/abs/10.1111/j.1749-818X.2010.00263.x, doi:10.1111/j.1749-818X.2010.00263.x. Smallwood, J., Schooler, J.W.,

    work page doi:10.1111/j.1749-818x.2010.00263.x 2010

  56. [56]

    Smallwood, J., Schooler, J.W.,

    doi:https://doi.org/ 10.1037/0033-2909.132.6.946. Smallwood, J., Schooler, J.W.,

    work page doi:10.1037/0033-2909.132.6.946

  57. [57]

    Proceedings of the 2017 ACM Conference on User Mod- eling, Adaptation and Personalization , 23–27doi:10.1145/3079628.3079633

    Face forward: Detecting mind wandering through facial expressions and building a feedback intervention. Proceedings of the 2017 ACM Conference on User Mod- eling, Adaptation and Personalization , 23–27doi:10.1145/3079628.3079633. Sümer, Ö., Goldberg, P., D’Mello, S., Gerjets, P., Trautwein, U., Kasneci, E.,

    work page doi:10.1145/3079628.3079633 2017

  58. [58]

    Frontiers in Human Neuroscience 17.0, 1182319

    Mind wandering state detection during video-based learning via eeg. Frontiers in Human Neuroscience 17.0, 1182319. doi:10.3389/fnhum.2023.1182319. Tasika, N.J., Haque, M.H., Rimo, M.B., Al Haque, M., Alam, S., Tamanna, T., Rahman, M.A., Parvez, M.Z.,

    work page doi:10.3389/fnhum.2023.1182319 2023

  59. [59]

    2020 IEEE Region 10 Symposium (TENSYMP) , 1474–1477doi:10.1109/TENSYMP50017.2020.9230790

    A framework for mind wandering detection using eeg signals. 2020 IEEE Region 10 Symposium (TENSYMP) , 1474–1477doi:10.1109/TENSYMP50017.2020.9230790. Toisoul, A., Kossaifi, J., Bulat, A., Tzimiropoulos, G., Pantic, M.,

    work page doi:10.1109/tensymp50017.2020.9230790 2020

  60. [60]

    Consciousness and Cognition 20, 1882–1886

    The eyes know what you are thinking: Eye movements as an ob- jective measure of mind wandering. Consciousness and Cognition 20, 1882–1886. URL:https:// www.sciencedirect.com/science/article/pii/S1053810011002182, doi:https://doi.org/10.1016/ j.concog.2011.09.010. from Dreams to Psychosis: A European Science Foundation Exploratory Workshop. Van De Schoot, ...

    work page 2011

  61. [61]

    frontiersin.org/journals/human-neuroscience/articles/10.3389/fnhum.2019.00348/full, doi:10.3389/fnhum.2019.00348

    URL:https://www.frontiersin.orghttps://www. frontiersin.org/journals/human-neuroscience/articles/10.3389/fnhum.2019.00348/full, doi:10.3389/fnhum.2019.00348. publisher: Frontiers. Wammes, J.D., Ralph, B.C., Mills, C., Bosch, N., Duncan, T.L., Smilek, D.,

    work page doi:10.3389/fnhum.2019.00348/full 2019

  62. [62]

    Computers & Education 132, 76–89

    Disengagement during lectures: Media multitasking and mind wandering in university classrooms. Computers & Education 132, 76–89. doi:https: //doi.org/10.1016/j.compedu.2018.12.007. Wang, M., Bodonhelyi, A., Bozkir, E., et al.,

    work page doi:10.1016/j.compedu.2018.12.007 2018

  63. [63]

    15546–15554

    Turbosvm-fl: Boosting federated learning through svm aggregation for lazy clients, in: Proceedings of the AAAI Conference on Artificial Intelligence, pp. 15546–15554. URL: https://ojs.aaai.org/index.php/AAAI/article/view/29481, doi:10.1609/aaai.v38i14.29481. Withammer-Ekerhovd, E.,

    work page doi:10.1609/aaai.v38i14.29481

  64. [64]

    URL:https: //ntnuopen.ntnu.no/ntnu-xmlui/handle/11250/3167337

    Classifying states of attention and mind wandering from EEG signals. URL:https: //ntnuopen.ntnu.no/ntnu-xmlui/handle/11250/3167337. accepted: 2024-11-28T18:24:01Z. 27 Wong, A.Y ., Smith, S.L., McGrath, C.A., Flynn, L.E., Mills, C., 2022a. Task-unrelated thought during educational activities: A meta-analysis of its occurrence and relationship with learning...

    work page doi:10.1016/j.cedpsych.2022.102098 2024

  65. [65]

    Human factors 56, 260–269

    Driving with the wandering mind: The effect that mind-wandering has on driving performance. Human factors 56, 260–269. doi:https://doi.org/10.1177/0018720813495280. Zhang, H.,

    work page doi:10.1177/0018720813495280

  66. [66]

    Sleep Medicine 112, 197–202

    The influence of neuroticism on insomnia: The chain mediating effect of mind wandering and symptom rumination. Sleep Medicine 112, 197–202. doi:https://doi.org/10.1016/j.sleep.2023.10.012. Zhang, Z., Ye, Y ., Zhou, J., Wang, Y .,

    work page doi:10.1016/j.sleep.2023.10.012 2023

  67. [67]

    Child Abuse & Neglect 170, 107788

    The detrimental effect of combined risks on mind-wandering: A response surface analysis of childhood family abuse and peer bullying. Child Abuse & Neglect 170, 107788. doi:10.1016/j.chiabu.2025.107788. Zhu, J., Zou, H., Rosset, S., Hastie, T., et al.,

    work page doi:10.1016/j.chiabu.2025.107788 2025

  68. [68]

    Proceedings of the 13th ACM International Conference on Bioinformatics, Computational Biology and Health Informatics , 1–10doi:10.1145/3535508.3545533

    Topographynet: a deep learning model for eeg-based mind wandering detection. Proceedings of the 13th ACM International Conference on Bioinformatics, Computational Biology and Health Informatics , 1–10doi:10.1145/3535508.3545533. 28 Appendix A. Summary of mind wandering detection modalities The following table provides an overview of the sensing modalities...

    work page doi:10.1145/3535508.3545533 2025

  69. [69]

    mind−wandering

    facial self-caught Bühler et al. (2024a); Bosch and D’Mello (2021b); Stewart et al. (2017) video probe-caught Bühler et al. (2024b); Bühler et al. (2024a); Lee et al. (2022a) keyboard/ self-caught Kuvar et al. (2023a) mouse probe-caught Dias da Silva and Postma (2020) fNIRS correct Liu et al. (2021) Appendix B. Search String The following search string wa...

    work page 2017

  70. [70]

    Naive Bayes(Zhang, 2004), includingGaussian Naive Bayes, comprises probabilistic classifiers based on Bayes’ theorem with the assumption of conditional feature independence

    prevents overfitting in high-dimensional settings and class weighting addresses imbalance, the model’s linear decision boundary may limit performance on complex, non-linear patterns without explicit feature engineering or basis expansions. Naive Bayes(Zhang, 2004), includingGaussian Naive Bayes, comprises probabilistic classifiers based on Bayes’ theorem ...

    work page 2004

  71. [71]

    construct an ensemble of decision trees using bootstrap sampling and random feature selec- tion, with predictions aggregated via majority voting. This approach reduces variance and overfitting and performs well with minimal tuning, but can be less interpretable than single trees and may struggle with highly sparse or high-dimensional feature spaces. XGBoo...

    work page 2009

  72. [72]

    Mind Wandering was measured by using self-caught reports, making it possible to indicate instances that do not correspond to specific probes (Kopp et al., 2016)

    which encompasses task unrelated thoughts and task related inferences, thus defining Mind Wandering as not attentively following the video content. Mind Wandering was measured by using self-caught reports, making it possible to indicate instances that do not correspond to specific probes (Kopp et al., 2016). This method was chosen instead of the probe-cau...

    work page 2016