Beyond Screen Time: Algorithmic Reward Unpredictability as the Locus of Attentional Risk in Short-Form Video Use Among Adolescents with ADHD

Authors: Aryaveer Agarwal

Published: June 30, 2026

Introduction

Adolescence is the period in which sustained, goal-directed attention is most vigorously consolidated. Working memory, inhibitory control, and the capacity for top-down guidance of attention mature gradually across the second decade of life, as the fronto-parietal and fronto-striatal circuits that support them continue to develop through this protracted phase. It is precisely in this malleable stage of brain development that adolescents have become the world's most voracious consumers of short-form video (SFV), a medium explicitly engineered to sustain continuous engagement through algorithmic selection and rapid, swipe-based delivery of clips. The collision between a developmentally incomplete control system and an environment optimized for engagement is therefore not a peripheral cultural worry but a question of developmental consequence.

Debate in both public and scientific forums has polarized around two framings of the problem. On one side is the view of SFV as a corrosive, effort-displacing force that trains the brain toward a distraction-prone, weakened attentional system. On the other is the position that the concern is overstated, that the measured effects are vanishingly small, and that the faster processing SFV entails is better understood as adaptation. The debate is sharpest for adolescents with ADHD, whose defining features include atypical reward processing and weaker attentional control—and because the ADHD population is large, even small effects can carry broad societal consequences.

This paper operates within developmental cognitive neuroscience, in dialogue with media psychology. Its core thesis is that both prevailing views err at a fundamental level: the truly central variable is neither the duration of the clips nor the amount of time spent consuming them, but the unpredictability of the rewards those clips are designed to deliver—the variable-ratio reward schedule that algorithmic selection has built into the very act of swiping. On this view, the risk to attention arises from SFV's reward architecture, and adolescents with ADHD, by virtue of their particular reward profile, face outsized risk from it.

The argument proceeds in six stages. Section 2 sets out epidemiological data establishing the scope of both phenomena—the prevalence of ADHD and the intensity of SFV use. Section 3 reconstructs the historical background and delineates the two contemporary framings before isolating the gap that remains in the literature. Section 4 poses a bounded, formal research question. Section 5 develops the paper's own position, together with its logical basis and empirical support. Section 6 draws out the clinical, design, and research implications, and Section 7 turns the paper's analytic framework back on the argument itself.

Background and Significance

The problem arises from the convergence of two large and growing quantities: the prevalence of ADHD and the intensity of adolescents' engagement with short-form video. ADHD is among the most common neurodevelopmental conditions of childhood and adolescence.

A meta-analysis applying DSM diagnostic criteria estimates a pooled prevalence of 5.6% among twelve- to eighteen-year-olds and 7.6% among younger children (Salari et al., 2023), while an umbrella review synthesizing thirteen meta-analyses and data from more than three million participants places the global figure near 8%, with the predominantly inattentive presentation the most common of the three subtypes (Ayano et al., 2023). The World Federation of ADHD International Consensus Statement reports a comparable prevalence of roughly 5.9% (Faraone et al., 2021). That the inattentive presentation predominates matters for the present argument, because the SFV literature bears most directly on attention rather than hyperactivity. Against this clinical backdrop, short-form video has rapidly become a near-default feature of teenage life.

A survey of a nationally representative sample indicates that roughly three-quarters of teenagers use YouTube daily, about six in ten use TikTok daily, one in three uses at least one platform almost constantly, and nearly half report being online almost constantly; about four in ten describe the algorithmically generated feed as very or extremely interesting (Faverio & Sidoti, 2024). Industry analytics place average daily time on a single short-form platform in the range of one to one-and-a-half hours for teenage users. Exposure on this scale, sustained across the developmental window in which attentional control matures, is what lends the question its urgency. The underlying logic is one of aggregation: even a small fraction of the roughly six percent of adolescents with ADHD represents, in absolute terms, a very large number of young people, and for that number to sustain attentional costs from a medium nearly all of their peers use daily constitutes a substantial and concentrated burden on an already disadvantaged group. Establishing that the burden exists, however, is not the same as identifying its cause, and it is on the question of mechanism that the current literature divides.

Literature Review

Concern that screen media erodes young people's attention long predates the smartphone. Cultural critics of the television age argued that a medium organized around sustained, passively absorbed stimulation was making the public more susceptible to distraction and less capable of sustained thought (Postman, 1985), and early empirical work tied heavy television exposure in early childhood to later attentional problems (Christakis et al., 2004). The thread connecting that era to the present is a steady compression of the unit of content and a steady increase in the viewer's control over what comes next: from the fixed schedules of broadcast, through the on-demand archives of streaming, to the user-curated infinite scroll of SFV. Music videos that once ran three to four minutes gave way to sixty-second clips and then to fifteen-second loops.

What distinguishes SFV from its broadcast predecessors is therefore not the speed of the content but the structure of the interaction it invites. Television delivered scheduled programming the viewer could not influence in real time; the SFV feed delivers content the viewer co-produces through swiping, and the reward attached to each swipe is selected by a recommendation system to maximize the likelihood of continued engagement. This is the architecture that so-called digital-dopamine researchers are attempting to name: the feed as an engine of unpredictable, personalized reinforcement (Satani et al., 2025). Framing the contemporary literature against this history makes clear that the salient discontinuity is the move from predictable scheduling to algorithmic unpredictability, a point to which the argument returns.

The displacement–deficit position, holds that algorithmically driven short-form video harms adolescents' attentional development. On this account, rapidly changing content that captures attention from the bottom up disrupts the top-down control that deliberate study and reading require, while habituating the viewer to a pace of stimulation that ordinary learning cannot match. A systematic review and meta-analysis links social-media use to ADHD-type symptoms in adolescents (Thorell et al., 2022); experimental studies report reduced performance on executive-control and conflict-processing tasks following brief SFV exposure (Yan et al., 2024; Xie et al., 2023); a meta-analysis finds elevated problematic digital-media use in individuals with ADHD relative to typically developing peers (Werling et al., 2022); and cross-sectional evidence ties heavier SFV use to more frequent inattentive symptoms (Chiencharoenthanakij et al., 2025). Collectively, this evidence supports a cognitive-load interpretation: the bottom-up processing demand of the feed exceeds the capacity of the executive systems tasked with regulating it, in adolescents with ADHD and typically developing adolescents alike (Conte et al., 2024; Dekkers & van Hoorn, 2022).

The cognitive-adaptation and neural-plasticity position, challenges these claims on both empirical and conceptual grounds. Its principal empirical evidence is that the measured association between digital-technology use and adolescent well-being is very small. A large cross-dataset analysis found that technology use accounts for only about 0.4% of the variance in adolescent well-being, a proportion the authors argue is comparable in magnitude to trivial correlates and far too modest to justify widespread alarm, while emphasizing that exposure models keyed to time spent are the wrong instrument for the question (Orben & Przybylski, 2019). This position's second, conceptual strand draws on the digital-native thesis, which holds that those raised amid numerous, rapidly accessed, multi-sourced information streams develop skills adapted to that ecology: efficient filtering and rapid switching (Prensky, 2001). On this reading, what deficit-oriented researchers interpret as impaired attention is instead a different and efficient mode of processing suited to an environment that rewards breadth over depth, and the appropriate response is to calibrate to this mode of use rather than to restrict it.

The two positions are not symmetrically right or wrong. Each grasps something real about the medium: One that the environment is taxing attention, though not what feature does so or where; Another that models built on raw exposure are uninformative. They talk past each other because each has fastened onto a different surface attribute of the system rather than onto its deeper, structural cause.

The failure to pinpoint the underlying cause of SFV's effects produces two interacting gaps. The first is evidentiary. The most direct experimental evidence for an attentional deficit is drawn largely not from clinically defined adolescent samples—the population most likely to suffer adverse outcomes, but from generally developing groups, or from adults and non-clinical controls. The population for whom the stakes are highest is thus the one least directly studied, and the experiment the field most needs is one that exposes clinically characterized ADHD adolescents, alongside matched neurotypical controls, to calibrated algorithmic feeds. The second gap is conceptual, and it is the precondition for that experiment: the current literature confounds all of SFV's salient features. In any deployed platform, clip length, content density, and reward unpredictability are perfectly confounded—every short clip is also high-density and delivered on a variable reward schedule,  so no existing study can isolate which feature does the causal work. Until an experiment holds brevity and density constant while manipulating reward predictability, the debate between the two contemporary positions cannot be resolved.

Research Question

This paper poses a narrow but precisely bounded question: to what extent is the variable-ratio reinforcement schedule that structures algorithmic short-form video, as distinct from clip duration or overall daily use, responsible for reliable, measurable reductions in executive attentional control among adolescents, and to what extent is any such effect larger in adolescents with a clinical ADHD diagnosis than in their neurotypical peers? Two sub-questions follow. First, can the effect of reward unpredictability be dissociated from the effects of clip length and content density? Second, does the dopaminergic reward dysregulation characteristic of ADHD moderate the relationship, such that identical algorithmic exposure produces a larger decrement in executive attentional control in this group? The question is deliberately constrained so that each clause is, in principle, directly testable.

Argument

The claim is this: the attentional hazard of short-form video arises not from clip length, and not from sheer volume, but from the unpredictable reinforcement schedule inherent to algorithmic curation. This position is distinct from both established perspectives. On this account, "short-form" is a red herring. An equal quantity of long video delivered on the same schedule of unpredictable reinforcement would be equally harmful, whereas an equal quantity of short video delivered predictably—say, a manually curated playlist—would be comparatively harmless. This move immediately resolves the standoff between the two contemporary positions. The displacement–deficit position is right about the medium's negative effect on attentional control but wrong about its cause; The cognitive-adaptation and neural-plasticity position is right about the failure of exposure-time accounts but wrong to conclude that the medium is benign. Relocating the cause to the reinforcement schedule makes their opposing conclusions congruent.

The claim rests on the behavioral and neurobiological status of variable-ratio reinforcement. Among schedules of reinforcement, the variable-ratio schedule produces the highest, steadiest rate of response and the greatest resistance to extinction, because every response carries a non-zero, unpredictable probability of reinforcement. Algorithmic feeds instantiate this schedule with unusual precision: each swipe is a low-cost action, and each outcome - a clip of unpredictable salience, arriving at an unpredictable point in the sequence - is tuned by the algorithm to be as engaging as possible. The neurobiological basis of the mechanism is phasic dopamine release indexed to reward prediction error, the difference between expected and received reward, which is maximized precisely when reward is uncertain. An unpredictable feed is therefore not merely pleasurable; it is a continuous source of prediction-error signals—the signals that drive habit formation and that bias attention toward the reward from which they arise.This interpretation reframes The displacement–deficit position's observations about cognitive load and digital-dopamine effects as downstream consequences of the schedule rather than independent causes.

The existing corpus, read as evidence for this premise, is consistent with it at all three analytic levels. At the behavioral level, longitudinal associations between heavy digital-media use and worsening attention are bidirectional (Thorell et al., 2022; Dekkers & van Hoorn, 2022; Wang et al., 2024); and, most tellingly for the present argument, it is the inattentive symptom cluster, not the hyperactive one, that is selectively associated with short-form media use (Chiencharoenthanakij et al., 2025). A prediction-error account predicts exactly this asymmetry: a stimulus engineered to capture attention should manifest as reduced focus on competing tasks rather than as generalized over-arousal. At the cognitive level, studies finding that executive-control performance declines after exposure to algorithmic feeds (Yan et al., 2024; Xie et al., 2023) point in the same direction—though not conclusively, since these studies confound reward unpredictability with clip length and content density. At the neurocognitive level, EEG analyses following short-form video exposure reveal reduced frontal theta activity and elevated gamma activity, indexing a shift from top-down attentional control toward a stimulus-driven pattern (Yan et al., 2024)—precisely the signature expected when attention has been captured by a feed engineered to be maximally engaging. The correspondence between this signature and the fronto-striatal and fronto-parietal impairments documented in ADHD (MacDonald et al., 2024; Werling et al., 2022) is the crucial link to the between-group prediction: because the dopaminergic dysregulation characteristic of ADHD steepens the response to reward prediction error, identical unpredictable reinforcement should exert a stronger pull on adolescents with the condition (MacDonald et al., 2024).

Earlier accounts describe a bidirectional relationship in which more inattentive adolescents gravitate toward attention-grabbing content, and that use in turn deepens their inattention. My argument goes further by specifying the engine of the loop. The cycle is driven by schedule-induced attentional reactivity, and it sustains itself through schedule-based learning that amplifies attentional sensitivity in adolescents with ADHD - precisely because they are more prone both to selecting into such feeds and to being conditioned by them. This yields a prediction that no rival account generates. Interventions that reduce the schedule's unpredictability - substituting a chronological feed for an algorithmic one, batching videos into predictable blocks, disabling autoplay, and removing dynamic recommendation - should reduce attentional cost more than interventions that merely cap total viewing time or lengthen individual clips. The cognitive-adaptation and neural-plasticity position would prescribe abstinence or time limits; The cognitive-adaptation and neural-plasticity position, finding effects negligible, would prescribe nothing; my argument prescribes redesigning the delivery schedule, a prescription that emerging experimental paradigms can, and likely will, put to rigorous test.

In clinical practice, the reframing argues for what might be called schedule literacy rather than for blanket bans. Because the risk resides in the unpredictability of reward rather than in screens per se, clinicians working with ADHD adolescents would do better to help families reorganize how content is delivered - disabling autoplay and algorithmic recommendation, substituting chosen playlists, and consolidating viewing into discrete blocks - than to impose global time targets that fail to address the causal variable and that, at best, prove only transiently sustainable.

For platform design, the analysis favors a predictability-by-default posture for all content consumed by minors - chronological or manually curated feeds rather than reward-optimized recommendation, and the removal of features whose only function is to make the next reward less predictable.

For research, the priority is the experiment implied by the gap analysis: one that recruits clinically characterized ADHD adolescents alongside well-matched neurotypical controls and tests reward predictability directly, holding clip length and content density constant and combining objective attentional measures with neural ones. This is the essential next step.

Conclusion

This paper has argued that the polarized debate over short-form video and adolescent attention has fastened onto the wrong variable. Reframing the hazard as the variable-ratio reward schedule of algorithmic feeds - rather than the brevity of clips or the quantity of exposure - reconciles the otherwise contradictory positions of the deficit and adaptation positions, and it predicts that adolescents with ADHD bear a disproportionate cost because their dopaminergic profile renders them unusually sensitive to unpredictable reinforcement. The contribution is not a new datum but a relocation of the causal question, and with it a falsifiable prediction and a precise specification of the experiment the field requires.

Intellectual honesty requires turning the same critical lens on this argument. First, an algorithm-centered account courts technological reductionism. If the schedule is the trigger, the dopaminergic dysregulation of ADHD is the loaded chamber, and the two are not separable: identical feeds would be far less consequential in the absence of the underlying neurobiology, which means a purely design-level remedy is necessarily incomplete and risks underweighting biological predisposition, comorbid conditions, and the family and developmental context in which use occurs. Second, the limitations of the rival prescriptions cut in both directions and constrain the present one as well. The displacement–deficit position's abstinence and time-limit remedies may fail not only because they target the wrong variable but because restriction displaces use toward other unpredictable feeds; The cognitive-adaptation and neural-plasticity position's laissez-faire stance under-protects a vulnerable subgroup, since a population-average effect of 0.4% can conceal a substantial effect concentrated in a high-sensitivity tail - an aggregation problem that the average conveniently hides. The argument is not immune to analogous error: its emphasis on the schedule could, if pressed too far, neglect that content itself, sleep displacement, and social comparison contribute to outcomes independently of reinforcement structure.

Third, and most consequentially, the argument's signature prediction rests on a dissociation - between unpredictability on the one hand and brevity and density on the other - that no deployed platform permits and that has not yet been engineered into an experiment; until it is, the ADHD-specific claim remains an extrapolation from non-clinical evidence, effect sizes remain small, and confounding remains a live threat. The appropriate posture, then, is precautionary but evidence-calibrated. The value of relocating the debate to the reward schedule is not that it settles the question but that it makes the question answerable: it names the variable to manipulate, the population to study, and the result that would prove the argument wrong.

References

  1. Ayano, G., et al. (2023). The global prevalence of attention deficit hyperactivity disorder in children and adolescents: An umbrella review of meta-analyses. Journal of Affective Disorders, 339, 860–866. 

  2. Chiencharoenthanakij, R., et al. (2025). Short-form video media use is associated with greater inattentive behaviors in children. Brain and Behavior, 15(7), e70656.

  3. Christakis, D. A., Zimmerman, F. J., DiGiuseppe, D. L., & McCarty, C. A. (2004). Early television exposure and subsequent attentional problems in children. Pediatrics, 113(4), 708–713.

  4. Conte, G., et al. (2024). TikTok and adolescent mental health: A scoping review. International Journal of Environmental Research and Public Health, 21, 11234.

  5. Conte, G., et al. (2024). TikTok and adolescent mental health: A scoping review. International Journal of Environmental Research and Public Health, 21, 11234.

  6. Faraone, S. V., et al. (2021). The World Federation of ADHD International Consensus Statement: 208 evidence-based conclusions about the disorder. Neuroscience & Biobehavioral Reviews, 128, 789–818.

  7. Faverio, M., & Sidoti, O. (2024). Teens, social media and technology 2024. Pew Research Center.

  8. MacDonald, H. J., et al. (2024). The ADHD dopamine hypothesis: Recent evidence and critical review. Frontiers in Psychiatry, 15, 1492126.

  9. Orben, A., & Przybylski, A. K. (2019). The association between adolescent well-being and digital technology use. Nature Human Behaviour, 3(2), 173–182.

  10. Postman, N. (1985). Amusing ourselves to death: Public discourse in the age of show business. Viking Penguin.

  11. Prensky, M. (2001). Digital natives, digital immigrants. On the Horizon, 9(5), 1–6.

  12. Salari, N., et al. (2023). The global prevalence of ADHD in children and adolescents: A systematic review and meta-analysis. Italian Journal of Pediatrics, 49(1), 48.

  13. Satani, A., et al. (2025). Modern day high: The neurocognitive impact of social media usage.

    Cureus, 17(7), e87496.

  14. Werling, A. M., et al. (2022). Problematic use of digital media in children and adolescents with ADHD compared to controls: A meta-analysis. Journal of Behavioral Addictions, 11(2), 305–325. 

  15. Xie, J., et al. (2023). Short-form video addiction and attentional control. Frontiers in Psychology, 14, 1178902.

  16. Yan, S., et al. (2024). Mobile phone short video use negatively impacts attention functions: An EEG study. Frontiers in Human Neuroscience, 18, 1383913.

  17.  Wang, J. L., et al. (2024). Longitudinal associations between internet addiction and ADHD symptoms in adolescents. Journal of Behavioral Addictions, 13(1), 55–67.
  18. Thorell, L. B., et al. (2022). Social media use and ADHD symptoms in adolescents: A systematic review and meta-analysis. Neuroscience & Biobehavioral Reviews, 132, 983–994.