Quantifying the Engagement Trap: Impact of Short-form Video Recommender Systems on Users with ADHD
Vedad Misirlic, Gregor Mayr, Elisabeth Lex
cs.HC · cs.AI
Quick verdict
N=302 survey: ADHD users report more time blindness, regret, distress from short-video recommendations; credible but self-report only, effects small.
Summary
This survey study asks whether short-form video recommender systems affect users with and without ADHD differently. The authors recruited 302 Prolific participants — 152 without ADHD, 65 with medically diagnosed ADHD, and 85 with self-reported ADHD — and administered a self-developed, psychologist-reviewed 7-point Likert questionnaire covering time blindness, disengagement difficulty, post-usage regret, guilt and frustration, and attitudes toward seven hypothetical inclusive-design features. Analysis used Kruskal-Wallis tests with Dunn's post-hoc comparisons (Holm-Bonferroni corrected), plus a merged-group Mann-Whitney robustness check. All groups rated perceived recommendation relevance equally (no significant difference), but participants with ADHD reported significantly higher time blindness, disengagement difficulty, post-usage regret, and negative mood, with mostly small effect sizes (epsilon-squared 0.026–0.064) and a medium effect only for frustration about disengagement (Q12, 0.108). All groups endorsed the theoretical interventions (median 5), with fixed-quantity mode and enforced breaks rated highest. The authors coin the 'Engagement Trap' interaction model and argue for neurodiversity-aware design and policy. Main limitations, largely acknowledged by the authors: unverified ADHD self-report (no severity instrument or medication data), retrospective self-report rather than behavioral measures, hypothetical rather than implemented interventions, and small cross-sectional effects that cannot establish causality.
Review complete
Assessment at a glance
5
risk flags identified
● high 0
● medium 3
● low 2
4 strengths · 5 weaknesses · 5 author questions
AI text signal = Binoculars zero-shot detector score over 20 sampled passages (Fast-DetectGPT co-reported) — absence does not prove human authorship, and a flag does not imply misconduct.
Levels reflect verification complexity — not a quality score. Evolve never rates your paper.
Key claims under review
- Perceived recommendation relevance does not differ between ADHD and non-ADHD participants (Q6 shows no significant group difference), so negative outcomes stem from engagement dynamics rather than recommendation quality.
- Participants with ADHD report significantly higher time blindness, disengagement difficulty, post-usage regret, guilt/frustration, and lower post-session mood than participants without ADHD.
- Medically diagnosed and self-reported ADHD subgroups do not differ significantly on any questionnaire item, supporting their statistical alignment.
- Participants across all groups endorse seven theoretical neuro-inclusive features (median 5), with fixed-quantity mode and enforced breaks rated highest.
- These patterns define the 'Engagement Trap': engagement-optimized feeds shift the burden of disengagement onto users' self-regulation, which burdens users with ADHD disproportionately.
Strengths
Finer ADHD stratification than prior work
Splitting medically diagnosed (n=65) from self-reported ADHD (n=85) and validating with parallel 3-group (Kruskal-Wallis/Dunn, Tables 1-2) and merged 2-group models (Mann-Whitney, Table 3) goes beyond the usual binary self-report design; the subgroups align statistically on all items.
Transparent, complete statistical reporting
Per-item nonparametric tests with corrected post-hocs and three effect-size families (epsilon-squared, Cliff's delta, rank-biserial); even the negligible/negative effect is shown (Q6, epsilon-squared = -0.007), not hidden.
The Q6 null is genuinely informative
Candid limitations and data-quality hygiene
Weaknesses
Post-hoc power analysis as design justification
Causal rhetoric exceeds small effects
Questionnaire validity limited to face-validity
Exposure confound left unanalyzed
Hypothetical intervention ratings carry little evidential weight
Full Review Report
$1.99 · one-timeUnlock the complete preliminary analysis Evolve produced for this paper.
Report preview
Preliminary Review: Quantifying the Engagement Trap (arXiv 2609.07795)
The remaining sections cover Methodology Assessment, Statistical Evidence, Results & Reproducibility, Limitations & Risks, and Verdict & Recommendations.
- All 4 strengths & 5 weaknesses, analyzed in detail
- 5 risk flags with severity, section & reviewer questions
- Methods & statistical evidence assessment
- 5 questions a peer reviewer would ask
- Full six-dimension evaluation report & recommendations
One-time purchase · this paper only
One-time purchase, this paper only. AI-generated preliminary analysis — not peer review.
Frequently asked questions
What is the 'Engagement Trap' in this paper?
The authors' term for an interaction pattern in which engagement-optimized recommender systems deliver highly relevant short-form video streams while providing no external stopping cues, so disengagement depends entirely on the user's self-regulation. Because ADHD involves executive-function and time-perception differences, users with ADHD report significantly more difficulty escaping this loop — more time blindness, post-usage regret, and distress — even though they rate the recommendations as equally relevant as non-ADHD users do.
How many people participated and how were they grouped?
302 participants recruited via the Prolific platform: 152 without ADHD, 65 with medically diagnosed ADHD, and 85 with self-reported (undiagnosed) ADHD. ADHD status was self-disclosed; no medical records were collected or verified.
Did participants with ADHD find the recommendations bad or irrelevant?
No. All three groups rated the perceived relevance of recommendations essentially identically, with no statistically significant difference (item Q6). The paper's point is that harm arises from how engaging, relevant content is delivered — infinite, frictionless streams with no stopping cues — not from irrelevant recommendations.
What design fixes did participants want?
Participants rated seven hypothetical features positively (median 5 on a 7-point scale): focus mode (feed removal), usage reminders, enforced breaks, fixed-quantity mode, manual advance (removing infinite scroll), a visual time indicator, and diversity injection. Fixed-quantity mode and enforced breaks received the highest ratings. None of the features were implemented or tested in real use.
About this review
This is an AI-generated preliminary review by Evolve.Pub, produced from the openly available text of the paper. It is not peer review and does not represent the authors' views. The canonical record of the paper is at the publisher link above.
- Figures 2 and 3 are images: the per-item response distributions and neutral-percentages shown inside them are not present in the extracted text (only captions and a few medians), so distribution shapes could not be checked.
- No data or code availability statement appears anywhere in the provided text; 'codeAvailable: false' reflects the absence of any statement rather than an explicit refusal to share.
Go deeper on this paper
Wrote a paper? Stress-test it before reviewers do.
Upload your manuscript for the same structured scrutiny — reasoning, statistics, evidence, figures, and reproducibility.