PaperlayerView exact revision on arXiv

Source-linked AI summary

Quantifying the Engagement Trap: Impact of Short-form Video Recommender Systems on Users with ADHD

Vedad Misirlic, Gregor Mayr, Elisabeth Lex

arXiv:2609.07795v1cs.HCcs.AI

TL;DR

The effects of short-form video recommendations on users with ADHD have limited quantitative evidence. This paper studies 302 participants and frames the interaction as an Engagement Trap. Participants with ADHD report greater difficulty disengaging and higher post-use regret and emotional costs, despite recommendations matching their interests.

  • Problem

    Quantitative evidence about how short-form video recommendations affect users with ADHD remains limited.

    Source: §2.2 · p. 3

  • Method

    The paper conducts a stratified study of short-form video consumption and operationalizes the Engagement Trap across participants with and without ADHD.

    Source: §1 · p. 1, §7 · p. 10

  • Results

    Recommendations match participants’ viewing interests, but participants with ADHD report greater difficulty disengaging, higher post-usage regret, and higher emotional costs.

    Source: §5 · p. 7

  • Takeaways & Limitations

    The findings provide quantitative evidence of systemic differences in short-form video recommendation experiences between people with and without ADHD.

    Source: §5 · p. 7

    The study relies on retrospective self-reported data.

    Source: §6 · p. 10

Abstract

from arXiv · show

Short-form video platforms use recommender systems to maximize engagement through highly efficient personalized recommendations. However, the impact of these recommendations on users with ADHD compared to users without ADHD remains underexplored. Through this study, we introduce and operationalize the Engagement Trap, illustrating how recommender systems, while successfully optimizing for engagement, disproportionately disadvantage users with ADHD. This stratified study of 302 participants, recruited via the online platform Prolific, compares experiences between participants with and without ADHD. Our results show that while recommendations are perceived as relevant across groups, participants with ADHD report significantly higher levels of time blindness, post-usage regret, and emotional distress when consuming recommendations. Moreover, we collect feedback for several proof-of- concept, theoretical design interventions for neuro-inclusive design principles. These findings provide quantitative evidence of systemic differences in engagement-optimized recommender systems and highlight the unbalanced negative effects and interactions these systems create for participants with ADHD. We argue for neurodiversity-aware, human-centered design approaches that mitigate such algorithmic harms and support more equitable experiences.

1 Introduction

Short-form video recommenders deliver personalized, relevant streams optimized for engagement, but their negative effects on users with ADHD have been underexplored. A study of 302 participants compares groups and introduces the Engagement Trap to describe their differing experiences.

1 Introduction

  • Short-form video recommenders maximize engagement by adapting recommendations to interaction data, producing relevant, continuously novel streams that can be difficult to leave.The cited examples of interaction data include watch time, re-watch rates, and interaction latency.Source: §1 · p. 1
  • The study addresses a research gap in quantifying recommender-system effects on users with ADHD through a stratified quantitative study of 302 participants.It compares medically diagnosed ADHD, self-reported ADHD, and no-ADHD groups, examining disengagement, time blindness, post-usage regret, and emotional impact.Source: §1 · pp. 1–2
  • Participants with ADHD report greater difficulty disengaging, time blindness, post-usage regret, and negative emotional impact, although recommendations align with interests and intent across groups.The paper names this interaction the Engagement Trap and reports that it affects users with ADHD more negatively than users without ADHD.Source: §1 · p. 2
  • The Engagement Trap describes the interplay between highly engaging recommender-generated streams and users’ need to invest substantial effort to disengage.The paper presents a stratified study operationalizing the model and empirical evidence of differences in disengagement, time perception, and post-usage experiences between participants with and without ADHD.Source: §1 · p. 2
  • The paper also explores participants’ perceptions of several proof-of-concept, theoretical intervention features for short-form video platforms.These interventions are considered as part of the paper’s proposed neuro-inclusive design contribution.Source: §1 · p. 2

2 Related Work

The related work frames the Engagement Trap at the intersection of recommender mechanics, ADHD-related digital vulnerabilities, and dark patterns. It distinguishes personalized continuous video streams from deceptive interface patterns and identifies limited quantitative evidence on ADHD experiences.

2.1 Mechanics of Short-Form Video Recommendations

  • Short-form recommenders use real-time preference prediction and personalized feedback loops to maximize engagement, mixing appealing and less appealing content to sustain attention.The review describes explorative content and intermittent reinforcement as ways these streams remain unpredictable yet rewarding.Source: §2.1 · p. 2

2.2 ADHD and Digital Interfaces

  • The review links ADHD-related challenges in digital environments to executive-function demands, stopping rules, time blindness, and impulse control.Prior work also suggests recommender systems may exploit neurobiological vulnerabilities, while characterizing them as sociotechnical systems that can sustain usage.Source: §2.2 · p. 3
  • Quantitative evidence validating reported ADHD experiences with short-form recommendations remains limited, with only a few studies connecting them to inattentiveness symptoms.The paper positions its study against this evidence gap.Source: §2.2 · p. 3

2.3 Dark Patterns

  • The review contrasts deceptive interface dark patterns with short-form platforms’ continuous personalized streams, which may tax active decision-making and executive function.It notes prior findings that users with ADHD can be more adept at recognizing and avoiding traditional social-media dark patterns, while arguing they may be more susceptible to short-form streams.Source: §2.3 · p. 3

3 Study Design

The study uses an online questionnaire to compare short-form video experiences across medically diagnosed ADHD, self-reported ADHD, and no-ADHD groups. It combines three-group and two-group tests while relying on participants’ self-reported ADHD status.

3 Study Design

  • The study recruited 302 participants through Prolific and separated ADHD participants into medically diagnosed and self-reported groups alongside participants without ADHD.Participants disclosed ADHD status themselves; the study did not collect medical records or medication status.Source: §3 · pp. 3–4
  • A self-developed questionnaire, checked by two psychologists for face validity, used seven-point Likert questions and covered demographics, platform use, experiences, and inclusive-feature preferences.The thematic areas included time blindness and disengagement, regret, emotional responses, and design-feature preferences.Source: §3 · p. 4
  • For three-group comparisons, the researchers used Kruskal-Wallis tests and Holm-Bonferroni-corrected Dunn post-hoc tests, reporting epsilon-squared and Cliff’s Delta effect sizes.They also combined the ADHD groups for Mann-Whitney U tests and Rank-Biserial Correlation effect sizes.Source: §3 · p. 4
  • The two-group analyses were used to validate the three-group results and assess robustness given self-reporting and the lack of verification of medical diagnoses.The study relied on participants’ accuracy and honesty when reporting medically diagnosed or self-diagnosed ADHD.Source: §3 · pp. 3–4

4 Results

Across groups, participants said recommendations aligned with their viewing intentions, but those with ADHD reported greater difficulty disengaging and more adverse post-use experiences. Participants also responded positively to the described inclusive design features, especially fixed-quantity mode and enforced breaks.

4 Results

  • YouTube Shorts was the most-used platform (75.9%), followed by Instagram Reels (67.9%) and TikTok (59.2%).Participants could select multiple platforms, so the cumulative count exceeded the number of participants.Source: §4 · p. 4, Figure 1 · p. 5
  • Recommendations aligned with viewing intentions across all groups, with no significant group difference (pQ6 = .999), while time blindness, guilt, frustration, and post-usage regret differed significantly.The three-group tests found overall differences on all but one question; post-hoc tests found no significant difference between self-reported and medically diagnosed ADHD groups.Source: §4 · pp. 4–5
  • Participants with ADHD reported significantly more time blindness, difficulty disengaging, frustration, and post-usage regret than participants without ADHD.Both ADHD groups reported greater likelihood of watching longer than intended and difficulty stopping; post-usage regret and mood after long sessions also differed.Source: §4 · pp. 5, 7
  • Median disengagement difficulty was 5 for both ADHD groups and 3 for participants without ADHD.Users with ADHD also reported higher frustration related to difficulty disengaging and higher time blindness.Source: §4 · p. 6
  • Participants expressed positive attitudes toward all described inclusive features, with fixed-quantity mode and enforced breaks receiving the highest median ratings.The features were presented theoretically in the questionnaire rather than implemented.Source: §4 · p. 7

5 Discussion

Recommendations generally match users’ viewing intentions, yet participants with ADHD report greater difficulty disengaging and higher post-usage regret, making the emotional cost of use higher. The discussion frames this disparity as an interaction between relevant content and self-regulation demands, and considers design and policy responses.

5 Discussion

  • Participants with ADHD, whether medically diagnosed or self-reported, report more negative effects from short-form video recommendations than participants without ADHD.Across groups, recommended content generally matches users’ intentions, but those with ADHD report more disengagement difficulty and post-usage regret.Source: §5 · p. 7

5.1 Relevance Does Not Imply Control

  • Recommendations generally match users’ intentions, suggesting the reported negative outcomes are linked to how relevant content is delivered rather than irrelevant recommendations.The discussion identifies a trade-off between engaging, relevant content and supporting user control.Source: §5.1 · p. 8

5.2 Disengagement, Time Perception, and Post-Usage Outcomes

  • Participants with ADHD report greater difficulty disengaging, time blindness, and post-usage regret; the discussion links this pattern to hyperfocus and reliance on self-regulation.When no external stopping signals are available, disengagement requires users to rely on self-regulation.Source: §5.2 · p. 8

5.3 Post-Usage Regret and Lowered Mood

  • Participants with ADHD report higher post-usage regret and lower mood after consuming recommendations, with prolonged consumption associated with more negative outcomes.The discussion presents the latter relationship as an association, not a demonstrated cause.Source: §5.3 · p. 8

5.4 The Engagement Trap as an Interaction Model

  • The Engagement Trap describes relevant recommendation streams paired with disengagement that depends mainly on users’ self-regulation or external triggers.The discussion connects this interaction to ADHD participants’ reported time-perception and disengagement difficulties.Source: §5.4 · p. 9

5.5 Inclusive Features and Tools

  • Participants responded positively to conceptual features including focus mode, reminders, enforced breaks, continuation prompts, time indicators, and recommendation interventions.The proposed features could provide external stopping signals, reduce reliance on self-regulation, limit time spent, and foster time awareness.Source: §5.5 · p. 9

5.6 Public Policy Implications

  • The authors argue that accessibility frameworks should consider how interaction dynamics and optimization objectives influence outcomes across populations, not only access to digital systems.They frame addressing these disparities as important to protecting neurodivergent users’ rights and supporting inclusive, accessible, and transparent online environments.Source: §5.6 · p. 9

6 Limitations & Future Work

The study’s findings are bounded by participant-reported ADHD status and retrospective self-reports of time perception and emotional state. The authors also note that medication was not measured and that inclusive features were assessed hypothetically, calling for longitudinal and behavioral follow-up.

Limitations

  • ADHD status relied on participant disclosure, without medical-record verification, although diagnosed and self-reported groups showed statistical alignment in symptom severity.The authors report divergence between those groups on some questions.Source: §6 · p. 10
  • Retrospective self-reports make perceived time and emotional state subjective; future work could add behavioral measures such as comparing perceived with actual usage time.The authors also propose longitudinal studies to capture longer-term effects.Source: §6 · p. 10
  • The study did not collect ADHD medication intake, so it cannot isolate whether medication moderated participants’ platform interactions or susceptibility to the Engagement Trap.The authors recommend collecting medication data in future longitudinal studies.Source: §6 · p. 10
  • Participants rated the inclusive features from descriptions rather than interacting with working tools, so their actual effects on usage and disengagement remain unassessed.The authors call for translating the concepts into working applications.Source: §6 · p. 10

Future Work

  • The authors call for longitudinal research on algorithmic harms and engagement traps, and for further development of neuro-inclusive features and fair, transparent platforms.They describe neuro-inclusive design features as important to accessible design.Source: §6 · p. 10

7 Conclusion

Survey findings show that recommendations align with viewing intentions, yet participants with ADHD report greater disengagement difficulty, time blindness, and negative emotional outcomes. The paper calls this interaction the Engagement Trap and argues for neurodiversity-aware design that supports time awareness, intentional disengagement, and user control.

7 Conclusion

  • Recommendations align with viewing intentions, yet participants with ADHD report greater disengagement difficulty, time blindness, and more negative emotional outcomes.The comparison is based on survey data from 302 participants.Source: §7 · p. 10
  • The Engagement Trap describes engagement-optimized recommenders sustaining attention in ways that undermine intentional disengagement and time perception.Source: §7 · p. 10
  • The authors call for neurodiversity-aware, human-centered design and evaluation that supports time awareness, intentional disengagement, and user control for people with ADHD.Source: §7 · p. 10

8 Declaration of Generative AI and AI-assisted Technologies in Writing

The authors used language and generative AI tools for writing-related assistance, then reviewed and edited the content and retained responsibility for it.

8 Declaration of Generative AI and AI-assisted Technologies in Writing

  • LanguageTool, Grammarly, and Gemini were used for grammar and spelling checks.Source: §8 · p. 11
  • The tools also helped improve style, paraphrase and reword text, and format tables.Source: §8 · p. 11
  • The authors reviewed and edited the content as needed and take full responsibility for the publication’s content.Source: §8 · p. 11

A Questionnaire

The questionnaire covers participant demographics and short-video use, reported viewing experiences, and demand for inclusive features.

A Questionnaire

  • Participants were asked about age, gender, ADHD diagnosis, platforms used, and daily short-video viewing time.Source: A Questionnaire · p. 13
  • Items assess time blindness and disengagement difficulty, including losing track of time, difficulty stopping, and whether recommendations match viewing intentions.Source: A Questionnaire · p. 13
  • Further items ask about regret, perceived value, displaced tasks, recall, mood, frustration, guilt, and feeling overwhelmed by recommendations.Source: A Questionnaire · p. 13
  • Inclusive-feature options include feed removal, reminders, enforced breaks, fixed-quantity viewing, manual advance, time indicators, and diversity injection.Source: A Questionnaire · p. 13

B Kruskal-Wallis and Epsilon-Squared Effect Size per Question

The appendix identifies two tables of per-question statistical results: one for Kruskal–Wallis tests and ε², and one for Dunn post hoc tests and Cliff’s delta.

B Kruskal-Wallis and Epsilon-Squared Effect Size per Question

C Mann-Whitney U-test and Rank-Biserial Correlation per Question

Table 3 reports Mann–Whitney U-test statistical significance results and Rank-Biserial Correlation effect sizes for each question.

C Mann-Whitney U-test and Rank-Biserial Correlation per Question

Loading 2609.07795v1…