"It's Like Drinking from a Fire Hose": Understanding and Characterizing Video Learning Experiences for Individuals with ADHD

• Video lectures offer flexible access and playback controls, but static visuals, long duration, and dense information can challenge ADHD viewers’ attention.Source: §1 · p. 1
• General lecture-design findings may not fit ADHD viewers: picture-in-picture can benefit general students but can distract viewers with ADHD.Source: §1 · p. 1
• The study asks how lecture elements shape ADHD viewers’ challenges, how challenges appear in behavioral signals, and what coping strategies and design preferences they have.Source: §1 · pp. 1–2
• Sixteen participants with ADHD watched two lecture segments each and reflected afterward, while researchers recorded behavior and used gaze playback to support recall.Source: §1 · p. 2
• The study identified inaccessible lecture designs, behavioral signals of challenges, coping strategies and their limits, and learning practices beyond video watching.Examples include overfocusing on underexplained visuals, prolonged fixations during understimulating segments, and using AI to check notes for knowledge gaps.Source: §1 · p. 2
• The authors describe this as the first systematic examination of ADHD viewers’ video lecture experiences with behavioral evidence, characterizing multimodal barriers and deriving design implications.Source: §1 · p. 2

2.1 Challenges of Video Learning for Individuals with ADHD

• Video lectures combine visual and auditory information that can increase cognitive and attentional demands, and prior ADHD video studies focused on social media rather than lecture formats.Source: §2.1 · p. 2

2.2 Technologies to Support Learning for Individuals with ADHD

• ADHD support technologies include tools for attention, task management, and communication, while video-accessibility work has simplified general videos or added auxiliary lecture supports rather than adapting lecture content itself.Source: §2.2 · pp. 2–3

2.3 Accessible Learning via Adaptation and Personalization

• Adaptive learning personalizes materials to learner needs, but ADHD poses distinct challenges because strategies such as increasing stimulation may overwhelm or distract, and engagement may appear differently.Source: §2.3 · p. 3
• Prior ADHD personalization work addressed text materials, while the paper identifies video lecture adaptation as an unaddressed domain.Source: §2.3 · p. 3

3 Methodology

• Sixteen adults with ADHD watched lecture videos and reflected post hoc, using observed behavior such as gaze trajectories and interaction logs as memory-recall probes.Source: §3 · p. 3

3.1 Participants

• The 16 participants were aged 20–51; fourteen reported clinical diagnoses and two were undergoing clinical diagnosis after university health-services screening.Source: §3.1 · p. 3

3.2 Apparatus

• Eight lectures were curated to vary content, instructor and slide presence, visual formats, and speech style, then trimmed to 13–17 minutes to limit fatigue while allowing attentional challenges to emerge.Source: §3.2.1 · p. 3, §3.2 · p. 3

3.2.1 Video Lecture Selection.

• Participants excluded topics they preferred to avoid, then received randomly assigned videos from the remaining pool, with each video viewed by four participants in counterbalanced order.Source: §3.2.1 · p. 4

3.2.2 Study Interface.

• During viewing, researchers monitored gaze, screen activity, and participants’ behavior; afterward, timestamped video and gaze playback supported retrospective reflection.Source: §3.2.2 · pp. 4–5, Figure 1 · p. 5

3.2.3 Implementation.

• The interface used scene segmentation with manual verification and reset gaze visualizations at scene boundaries to prevent gaze data carrying across scenes.Source: §3.2.3 · p. 5

Initial Interview

• Participants first discussed their backgrounds and video-learning experiences, then completed gaze calibration.The initial interview covered demographics, ADHD background, and general challenges and coping strategies.Source: §3.3 · p. 5

Video Learning

• Participants watched two lecture segments with control over playback and activities, then answered factual-recall quiz questions.Researchers monitored behavioral data and noted events such as pauses and rewinds during viewing.Source: §3.3 · pp. 5–6

Retrospective Think-aloud

• After each task, participants reflected on the video while viewing it slide by slide and referencing their gaze behaviors when applicable.Researchers used targeted questions to address notable behaviors participants had not mentioned spontaneously.Source: §3.3 · p. 6

Exit Interview

• An exit interview explored whether experiences and preferences generalized to other lectures and how participants would improve video design.Participants were encouraged to consider improvements without technological constraints.Source: §3.3 · p. 6

3.4 Analysis

• The quantitative analysis summarized ratings, quiz accuracy, pauses, and rewinds, and tested correlations between content absorption and design effectiveness, interest, and familiarity.Spearman correlations used Holm-Bonferroni correction.Source: §3.4.1 · p. 6

3.4.2 Qualitative Analysis.

• For qualitative analysis, researchers jointly coded behavioral signals and retrospective feedback, developing and applying a shared codebook through independent coding and discussion.Two researchers independently open-coded four shared samples (25% of data) before dividing the remaining data and periodically resolving discrepancies.Source: §3.4.2 · p. 6

4 Findings

• ρ = 0.75 (p < 0.001) for perceived video-design effectiveness and content absorption, compared with ρ = 0.51 (p < 0.01) for interest and ρ = 0.42 (p = 0.016) for familiarity.Across videos, perceived absorption averaged 4.32 (SD = 1.42), ranging from 2 to 7; quiz accuracy averaged 92.2% (SD = 18.4%).Source: §4 · p. 6
• Despite comparable familiarity and interest, learners rated absorption in the public policy lecture higher than in the neuroscience lecture (6.25 vs. 3.75), alongside higher design-effectiveness ratings (6.75 vs. 3.25).Familiarity was 3.00 vs. 2.75 and interest was 5.75 vs. 5.50, respectively.Source: §4 · p. 6
• P11 said an interesting topic could still be unappealing when its presentation was not adapted to how he learns.The study frames accessible, adaptive video design as important to learners with ADHD.Source: §4 · p. 6

4.1 Cognitive Overload from Overwhelming Content

• Cognitive overload from overwhelming multimodal presentation and information was a significant barrier to effective video learning for participants with ADHD.Participants identified overly dense, unexplained, and irrelevant visuals, as well as low-quality audio, as overwhelming presentation designs.Source: §4.1 · p. 6, §4.1.1 · p. 6

4.1.1 Overwhelming Presentation.

• Dense slides split attention between text and speech, producing lagging gaze, content skipping, or apparent reading without comprehension.Some participants preferred dense slides as a fallback when speakers were ineffective, and tolerance could depend on content familiarity.Source: §4.1.1 · p. 7
• Large unexplained visuals could draw attention from speech, while irrelevant images diverted gaze and provoked task-irrelevant thoughts.Participants differed: some found illustrative images distracting, while others valued them as memory aids.Source: §4.1.1 · p. 7
• Background noise and disfluent speech distracted some participants, while fast, dense information could overwhelm them and prompt disengagement.One participant looked away more after speeding up the video; another took a break after being unable to absorb more content, attributing the burden to holding unrelated contexts in mind.Source: §4.1.1 · p. 7, §4.1.2 · pp. 7–8

4.1.3 Coping Strategies and Limitations.

• Participants used visual and auditory signifiers to align attention with speech or filter what seemed worth attending to.New images attracted gaze in 90.0% of instances, compared with 41.3% for new text and 40.1% for pointing gestures; cues could be absent, obscured, or distracting.Source: §4.1.3 · p. 8
• Participants paused or rewound to separate competing information sources, but this could extend viewing time and make refocusing difficult.They paused 113 times and rewound 69 times; 72.6% of pauses occurred on three videos identified as containing all three types of overwhelming presentation elements.Source: §4.1.3 · pp. 8–9
• Some participants focused on one element, such as the speaker, to avoid being overwhelmed by complex slides, but a usable attention anchor depended on video design.In one example, the screen-recorded lecture left a participant looking at blank space while trying to listen.Source: §4.1.3 · p. 9

4.2 Boredom from Understimulating Presentation

• 11 participants described boredom across six videos, with understimulation arising from both speech and visual channels.Seven participants showed prolonged fixed-point staring associated with “checking out” (Mean = 3.26s, SD = 1.78s), mainly in the art history and mathematics videos.Source: §4.2 · p. 9, §4.2.1 · p. 9, Figure 3 · p. 10

4.2.1 Understimulating Lecture Designs.

• Slow, monotone slide-reading reduced prosodic cues for important content, while tight speech-slide synchronization could lower the effort of reconciling sources.Participants also described static slides and missing visual supplements during tangents as dull or hard to follow.Source: §4.2.1 · p. 9

4.2.2 Coping Strategies and Limitations.

• 10 participants shifted gaze briefly to secondary visuals, such as captions, to reboot attention before returning to the slide.This strategy depended on having something suitable to look at: overly simple visuals offered little to switch to, while complex ones risked distraction; familiar, repetitive components could reduce added cognitive load.Source: §4.2.2 · p. 9
• Seven participants intentionally let their minds wander during boring or unimportant segments while retaining enough peripheral attention to refocus on important content.This strategy could fail when monotone speech made it difficult to detect that something had changed and re-engage.Source: §4.2.2 · pp. 9–10
• All participants described multitasking to increase stimulation, but seven acknowledged that it could distract them from learning.Note-taking could shift effort toward recording rather than understanding, while hands-busy activities could make note-taking difficult.Source: §4.2.2 · p. 10
• Participants sped up videos to increase stimulation; the mathematics video was the only one all participants sped up and one of only two sped up to 2×.Faster playback could make speech feel unnatural or encourage rushing through rather than absorbing content.Source: §4.2.2 · p. 10

Forced Scene Switches

• Five participants found forced switches between speaker and slide views disrupted recall and train of thought, while some others valued changes for dynamism or as cues to attend to different information sources.After a switch back to slides, one participant had to rescan content already viewed.Source: §4.3 · p. 10

Unexpected Topic Changes

• Abrupt or poorly signaled topic changes disrupted participants’ ability to reconcile speech and visuals, with gaze scattering after one mid-sentence shift.A separate concept absent from the original flowchart also threw a participant off.Source: §4.3 · pp. 10–11

Coping Strategies and Limitations

• After abrupt content changes, participants reported no effective strategy beyond pausing and re-examining, but frequent pauses and restarts could compound disruption.Source: §4.3 · p. 11

4.4 Confusion from Missing Key Content

• Nine participants experienced confusion from missing context, background knowledge, or learning goals, which can be especially challenging when fragmented information must be organized into coherent representations.Source: §4.4 · p. 11
• In four videos, participants struggled with concepts introduced without definitions or a broader framework; three also described uncertainty when lectures lacked clear learning goals.Source: §4.4 · p. 11
• Participants sought external resources to fill knowledge gaps, but leaving the video could invite distraction or lead to information overload.Source: §4.4 · p. 11

4.5 Addressing Higher-Level Video Learning Needs with Generative AI

• Higher-level needs included restructuring inaccessible content, checking understanding after learning, and repurposing learning for personal goals; five participants used generative AI, while more expressed concerns.Source: §4.5 · p. 11

4.5.1 Addressing Learning Needs via Generative AI.

• All five participants who used AI for video learning relied on it to restructure content into summaries, highlights, notes, or readings that felt more digestible.Source: §4.5.1 · p. 11
• Two participants used AI after learning to compare their notes with a transcript and check for missed content, while one used learning objectives to generate an exam-focused study guide.Source: §4.5.1 · p. 11
• Nine participants hesitated to use AI over concerns that outputs could miss instructors’ intended goals or be inaccurate, and validating outputs against other sources was tedious and difficult to sustain.Participants noted the risk of trusting outputs after repeated correct responses without checking the next one.Source: §4.5.2 · pp. 11–12

4.5.3 A Missed Opportunity: Externalizing Attention through AI-

• Three participants saw eye-tracking feedback as a way to monitor their attention, improve self-awareness, and identify material to review.Source: §4.5.3 · p. 12

5 Discussion

• The study links inaccessible lecture designs to behavioral patterns and coping strategies, including strategies used beyond viewing, while noting their limitations.The discussion frames improved ADHD-friendly video learning as a shared effort between educators and personalized adaptation systems.Source: §5 · p. 12

5.1 Navigating Tensions in Video Adaptation Towards ADHD-friendly Video Learning

• Video-level adaptation can preserve content structure, learner agency, and trust, while beyond-video abstraction can improve efficiency but risks omitting content or misrepresenting educator intent.Even participants using AI-generated summaries still wanted to watch the video to ensure understanding.Source: §5.1 · p. 12
• Both overly simple and visually rich designs can undermine engagement, and stimulation needs shift with content difficulty, fatigue, and learning goals.The discussion therefore recommends dynamically modulating stimulation using behavioral signals rather than optimizing statically for simplicity or richness.Source: §5.1 · p. 12

5.2 Informing ADHD Attention and Engagement with Gaze Behaviors

• Gaze signals are ambiguous: attentive-looking patterns can accompany mindless reading, while dispersed or stationary gaze can reflect intentional attention management as well as disengagement.The authors recommend interpreting behavior with video context and moving beyond binary attention-versus-distraction classifications.Source: §5.2 · pp. 12–13

5.3 Design Implications for ADHD-friendly Video Learning

• Educators can lower broad access barriers, while personalized systems address individual needs that universal design cannot fully resolve.The discussion notes that participants had diverse and sometimes conflicting video-design preferences.Source: §5.3 · p. 13, §5.3.2 · p. 13

5.3.1 For Educators: Toward ADHD-friendly Video Lecture Design.

• Educators should provide stable visual anchors, multimodal topic cues, aligned speech and visuals, clear context, and balanced visual content.These recommendations aim to support attention, clarify dense material, and avoid both clutter and understimulation.Source: §5.3.1 · p. 13

5.3.2 For Adaptation System Designers: Toward Personalized Video

• Adaptation systems should use behavior and video context, preserve viewer agency, augment attention cues, offer in-situ clarification, and keep AI adaptations aligned with educator intent.Suggested approaches include preference settings, user-selected secondary visuals, and links between adapted content and the original video.Source: §5.3.2 · pp. 13–14

5.4 Limitations and Future Work

• The lab setting and assigned videos may not reflect everyday self-selected learning, while the exploratory study cannot quantify individual design effects or broader and long-term learning outcomes.The authors call for longitudinal field deployments and larger controlled studies with scalable attention measures and learning and retention assessments.Source: §5.4 · p. 14