Evidence

Bibliography

If we cite a figure anywhere on this site, the source is listed here. Check any of it.

5 min read
In short

This page lists every source cited across Emotuit's guides and articles, grouped by topic: engagement and disengagement research, emotion recognition and affective computing, learning measurement and L&D data, and EU AI Act and privacy regulation.

Engagement and disengagement research

ScienceDirect (2024) — Cyberloafing as a predictor of disengagement in online courses. Finds tab-switching to be the single strongest predictor of disengagement, outperforming self-regulation and satisfaction measures. sciencedirect.com

Boote, D., Agarwal, A. & Mostow, J. (2021) — Predicting disengagement before it occurs. Reports 73.3% prediction accuracy with 40% of the activity still remaining. pmc.ncbi.nlm.nih.gov

OUCI / DNTB — Multimodal engagement classification in online learning. Reports classification accuracy improving from 91.5% to 94.6% when behavioural signals are combined with facial expression analysis. ouci.dntb.gov.ua

PMC (2021) — MOOC engagement framework. Real-time system perceiving learner engagement by recognising facial expressions and mapping them to content positions. pmc.ncbi.nlm.nih.gov

ScienceDirect (2020) — Facial expression analysis of 67 students to identify which lecture segments caused disengagement. sciencedirect.com

Emotion recognition and affective computing

Barrett, L. F., Adolphs, R., Marsella, S., Martinez, A. M. & Pollak, S. D. (2019)Emotional Expressions Reconsidered: Challenges to Inferring Emotion From Human Facial Movements. Psychological Science in the Public Interest. The most substantial published challenge to the basic-emotion model underlying FACS-based classification. journals.sagepub.com

Frontiers in Psychology (January 2026) — On calibration in affective computing: "Calibration typically involves capturing a neutral facial expression and adjusting subsequent data accordingly." Notes particular relevance for passive viewing scenarios. frontiersin.org

Springer, Smart Learning Environments (2025) — Confirms FACS as the standard framework for AI-driven educational emotion recognition. link.springer.com

PMC / Multimedia Tools and Applications (2022) — Deep learning models (Inception-V3, VGG19, ResNet-50) calculating engagement indices from facial emotion recognition data. pmc.ncbi.nlm.nih.gov

Ekman, P. & Friesen, W. V. — Facial Action Coding System. The taxonomy of facial Action Units underlying essentially all commercial facial expression analysis.

Russell, J. A. — A circumplex model of affect. The valence–arousal dimensional alternative to discrete emotion categories.

Learning measurement and L&D data

LinkedIn — Workplace Learning Report. Source for the finding that 67% of L&D leaders struggle to demonstrate training impact and 29% feel confident proving ROI. learning.linkedin.com

TalentLMS — 2026 L&D Report: The State of Workplace Learning. talentlms.com

AIHR — Learning and development statistics. Source for several completion and measurement benchmarks. aihr.com

eLearning Industry — On why completion rates of online courses are low, and how they vary by format. elearningindustry.com

Kirkpatrick Partners — The Kirkpatrick Model and the New World Kirkpatrick Model. kirkpatrickpartners.com

Sweller, J. — Cognitive Load Theory. The intrinsic, extraneous and germane load distinction underlying content design guidance.

EU AI Act, privacy and regulation

Regulation (EU) 2024/1689 — the AI Act. Article 5 sets out prohibited practices; Article 5(1)(f) covers emotion inference in workplace and education contexts. Recital 18 defines the exclusions. Article 3(34) defines biometric data; Article 3(39) defines emotion recognition systems. artificialintelligenceact.eu/article/5

Future of Privacy ForumRed Lines under the EU AI Act: Unpacking the Prohibition of Emotion Recognition in the Workplace and Education Institutions. The most thorough public analysis of the prohibition's scope, including the unresolved secondary-functionality question. fpf.org

Wolters KluwerThe Prohibition of AI Emotion Recognition Technologies in the Workplace under the AI Act. legalblogs.wolterskluwer.com

Technology's Legal EdgeEU AI Act: Spotlight on Emotional Recognition Systems in the Workplace. technologyslegaledge.com

Information Commissioner's Office (UK) — Guidance and public statements on biometric technologies and emotion detection, including warnings on scientific reliability and discrimination risk. ico.org.uk

GDPR — Regulation (EU) 2016/679. Article 4(14) biometric data definition; Article 9 special category conditions; Article 35 DPIA requirement; Article 36 prior consultation.

Market and category

Fortune Business Insights — Emotion AI market size and forecast. fortunebusinessinsights.com

GM Insights — Emotion AI market analysis. gminsights.com

EFFStudents Are Pushing Back Against Proctoring Surveillance Apps. Context on the cultural and reputational dimension of monitoring in education. eff.org


Frequently asked questions

Why publish a bibliography?
Because vendor claims in this category are usually unsourced, and a number without a source is a marketing assertion. Listing everything in one place means you can check any figure we quote rather than taking the summary on trust.
Do you cite research that contradicts your product?
Yes — the Barrett review challenging the basic-emotion model is on this list, and it is the most serious published objection to the framework our facial analysis layer uses. Leaving it off would have been the more comfortable choice and the less honest one.
How current is this?
The regulatory sources are checked when the compliance pages are reviewed, currently quarterly. Research citations are stable — papers do not change — but the field moves, and anything materially superseded gets noted rather than quietly dropped.

See where your content loses people

Book a walkthrough and we will show you the engagement data on your own content.

Request a demo
Get Started

See what completion rates can't tell you

Find out exactly where your content works, where it fails, and what disengagement looks like before people leave.

Request a Demo