The science behind PullBack.

PullBack is built on a specific bet: that the cheapest place to break a doomscroll is during it, not before. That bet rests on a stack of behavioral science we've spent some time reading. Below is what the actual papers say — what holds up, what's smaller than the headlines suggest, and where PullBack's mechanism fits in.

1. Attention Residue & Interruption-Cost Asymmetry

What the research says. When you switch tasks, a chunk of your attention stays glued to the previous one. Sophie Leroy coined this "attention residue" in her 2009 Organizational Behavior and Human Decision Processes paper, showing across two lab studies that performance on Task B suffers when participants haven't psychologically finished Task A. Gloria Mark's field work at UC Irvine — shadowing knowledge workers minute-by-minute — produced the widely-cited but often-misquoted finding that workers spent an average of just three minutes on a task before switching, and that returning to a disrupted task can take roughly 23 minutes when measured across spheres of work, not 23 minutes for every minor interruption. Both bodies of work concern involuntary interruptions to focused work; neither studied the cost of being yanked out of a passive scroll, which is structurally different — there's no Task A to leave residue from when Task A was watching strangers cook eggs.

Why it matters for PullBack. The "interruption is bad" trope is a real finding from a different context. Pulling someone out of deep work is expensive. Pulling someone out of an unintentional 47-minute Reels session is the exact opposite: you're terminating residue, not creating it. The asymmetry is the whole point — the cost of a deliberate interruption (a few seconds of mild annoyance) is dwarfed by the cost of the binge it ends.

Leroy, S. (2009). Organizational Behavior and Human Decision Processes, 109(2), 168–181.
Mark, G., Gudith, D., & Klocke, U. (2008). The Cost of Interrupted Work: More Speed and Stress. CHI 2008.
Mark, G., Iqbal, S. T., Czerwinski, M., Johns, P., & Sano, A. (2016). Focused, Aroused, but so Distractible: A Temporal Perspective on Multitasking and Communications.

2. Variable-Ratio Reinforcement & Infinite-Scroll Design

What the research says. B.F. Skinner identified the variable-ratio schedule — rewards delivered after an unpredictable number of responses — as the most resistant-to-extinction reinforcement schedule he tested. Slot machines exploit it deliberately. Anthropologist Natasha Dow Schüll spent 15 years embedded in the Las Vegas gambling industry and documented in Addiction by Design (Princeton, 2012) how machine designers explicitly engineer "the zone" — a dissociative trance where players lose track of time, money, and bodily needs. Tristan Harris and the Center for Humane Technology have argued, drawing directly on Schüll, that infinite-scroll feeds replicate the same architecture: pull-to-refresh as the lever, algorithmically-spaced novelty as the variable reward. The neurochemistry is the part most pop articles get wrong: dopamine is primarily a "wanting" / seeking signal, not a pleasure signal — Kent Berridge's incentive-salience work (Robinson & Berridge) clarified this distinction. Noradrenaline handles arousal and vigilance, which is why doomscrolling can feel both wired and joyless at the same time.

Why it matters for PullBack. You're not weak. The feed is engineered, by people with PhDs, to keep you in the zone. The same mechanism that makes a Bally slot machine sticky makes TikTok sticky. PullBack doesn't try to make the feed less appealing — that's a losing battle against billion-dollar optimization budgets. It just makes the session shorter than the variable-reward schedule needs to do its work.

Schüll, N. D. (2012). Addiction by Design: Machine Gambling in Las Vegas. Princeton University Press.
Hsu, H. (2013). Machine gambling in the 'zone': Natasha Dow Schüll's Addiction by Design. Social Studies of Science.
Burhan, R., & Moradzadeh, J. (2020). Neurotransmitter Dopamine and its Role in the Development of Social Media Addiction.
Center for Humane Technology — The CHT Perspective.

3. Time-Blindness in Flow & During Media Consumption

What the research says. Mihaly Csikszentmihalyi's Flow (1990) identified "transformation of time" as one of nine core characteristics of the flow state — subjective time can compress or stretch when attention is fully absorbed. More recent lab work (Rivkin & Bouwer, Creativity Research Journal, 2018) demonstrated experimentally that moderate attentional demand produces downward time distortion (the "where did the hour go" effect). The doomscroll case is messier — it's not classical flow, it's closer to what Schüll calls the zone — but the time-distortion outcome is the same. Empirical work on smartphone use specifically: Andrews et al. (PLOS ONE, 2015) found participants underestimated their actual screen time, and a 2023 study found participants underestimated lockdown-era smartphone use by roughly 71 minutes per day on average, with inactive users underestimating by up to 85 minutes. Self-reported screen time is, charitably, a rough guess — a meta-analysis found only 3 of 49 comparisons placed self-reports close to logged means.

Why it matters for PullBack. Users genuinely don't know how long they spent in the app. Asking them to "be mindful" is asking them to perceive something perception is bad at. A timer is a sensory prosthetic — it does the noticing the user can't reliably do themselves.

Rivkin, A., & Bouwer, L. (2018). Distorted Time Perception during Flow as Revealed by an Attention-Demanding Cognitive Task. Creativity Research Journal, 30(3).
Andrews, S., Ellis, D. A., Shaw, H., & Piwek, L. (2015). Beyond Self-Report: Tools to Compare Estimated and Real-World Smartphone Use. PLOS ONE.
Parry, D. A., et al. (2023). Discrepancies Between Self-reported and Objectively Measured Smartphone Screen Time: Before and During Lockdown.
Sewall, C. J. R., et al. (2022). Duration, frequency, and time distortion: Which is the best predictor of problematic smartphone use in adolescents? A trace data study. PMC.

4. The Efficacy of Friction-Based Interventions

What the research says. The most-cited recent evidence is Löchner et al., "Directing smartphone use through the self-nudge app one sec" (PNAS, 2023), out of Heidelberg University and the Max Planck Institute. In a 6-week field experiment with 280 participants, the app — which interposes a brief friction screen with a deliberation prompt and a dismiss option before the target app opens — produced a 57% reduction in actual app openings by week 6, and participants attempted to open target apps 37% less often than in week 1. A preregistered follow-up experiment (N=500) decomposed the mechanism: the option to dismiss drove the strongest effect, the time delay produced a smaller but real reduction, and the deliberation message alone did basically nothing. So "friction works" is roughly true, but not the whole story — friction works because it gives people a moment to bail out, not because reading a mindfulness sentence changes minds. Other RCTs have shown smaller and less consistent results. A 2025 npj Mental Health Research trial found that limiting digital screen use improved well-being and mood; nudge-based RCTs have shown effects ranging from null to ~29 minutes/day reduction on the most problematic app. Effect sizes vary; the design and the user's pre-existing motivation matter.

Why it matters for PullBack. The PNAS paper validates the category — small, well-timed friction can reliably reduce app use — and it identifies the specific feature that does the work: making "give up and close" easy. PullBack's mid-session interrupt is essentially that mechanism, but moved from the entry to the exit: rather than offering an off-ramp at the door, it forces one before the binge gets long. The "dismiss option" finding is the exact theoretical basis for not making the close screen punitive.

Löchner, J., et al. (2023). Directing smartphone use through the self-nudge app one sec. PNAS, 120(8), e2213114120.
Open Access mirror on PMC.
Pieh, C., et al. (2025). Smartphone screen time reduction improves mental health: a randomized controlled trial.

5. Why Mid-Session Beats Pre-Open & Hard-Block

What the research says. Three converging lines of evidence support the mid-session position. First, habituation: behavioral science has known since Thompson & Spencer (1966) that repeated weak stimuli lose effect. A pre-open friction prompt that fires every single time you tap Instagram is a textbook habituation setup — you mash through it on autopilot within a week. Second, psychological reactance (Brehm, 1966; Steindl et al., 2015): when freedom is restricted, people are motivated to restore it. Hard schedule blockers (Freedom, AppBlock, Cold Turkey) are excellent at producing reactance — users disable them, find workarounds, or simply uninstall when they "really need" the app. Third, self-permission: the mid-session interrupt fires after the user has already opened the app. There's no restriction on "I want to check Instagram" — the user did. The interrupt only restricts "I want to be on Instagram for 90 minutes," which the user almost never consciously chose.

Why it matters for PullBack. This is the whole product thesis in one paragraph. Pre-open friction trains users to ignore it. Hard blocks trigger reactance, get disabled, never come back. A mid-session interrupt fires after self-permission (no reactance) but before the variable-reward schedule has done most of its damage. It's the only window where the behavioral economics work in your favor.

Brehm, J. W. (1966). A Theory of Psychological Reactance. Academic Press.
Steindl, C., Jonas, E., Sittenthaler, S., Traut-Mattausch, E., & Greenberg, J. (2015). Understanding Psychological Reactance: New Developments and Findings. Zeitschrift für Psychologie.
Rosenberg, B. D., & Siegel, J. T. (2018). A 50-year review of psychological reactance theory.

6. Doomscrolling — What Is It Actually?

What the research says. The term first surfaced on Twitter in October 2018 and was popularized by Quartz reporter Karen Ho during the COVID-19 lockdown in early 2020; Merriam-Webster added it to "Words We're Watching" the same year. Real research caught up later. Sharma, Lee, and Johnson (2022, in Technology, Mind, and Behavior) developed and validated the 15-item Doomscrolling Scale, defining doomscrolling as "habitual, immersive scanning for timely negative information on social media newsfeeds." Subsequent peer-reviewed work has linked higher Doomscrolling Scale scores to: lower life satisfaction and harmony (Satici et al., 2023); existential anxiety and pessimism in a cross-cultural Iran/US sample (Shabahang et al., 2024); and elevated depressive symptoms and future anxiety (McCutcheon et al., 2024). Personality predictors: high neuroticism, low conscientiousness, low self-control. Methodological caveat: most studies are cross-sectional and self-report, so causal direction is murky and effect sizes are modest.

Why it matters for PullBack. Doomscrolling is a real, measurable construct with replicated mental-health correlations — not a moral panic. The literature also tells us who's most at risk (high neuroticism, low self-control), which is exactly the population least likely to stop themselves and most likely to benefit from an external mechanism that does the stopping for them.

Sharma, B., Lee, S. S., & Johnson, B. K. (2022). The Dark at the End of the Tunnel: Doomscrolling on Social Media Newsfeeds. Technology, Mind, and Behavior, 3(1).
Satici, S. A., Gocet Tekin, E., Deniz, M. E., & Satici, B. (2023). Doomscrolling Scale: its Association with Personality Traits, Psychological Distress, Social Media Use, and Wellbeing. Applied Research in Quality of Life.
Shabahang, R., et al. (2024). Doomscrolling evokes existential anxiety and fosters pessimism about human nature? Evidence from Iran and the United States. Computers in Human Behavior Reports.
Merriam-Webster — Doomsurfing and Doomscrolling: Words We're Watching.

7. Phone-Use Stats With Real Provenance

What the research says. The "142 unlocks per day" figure floating around marketing decks is a Deloitte estimate from earlier U.S. consumer surveys; Deloitte's Connected Consumer Survey 2023 cited an average of roughly 52 phone checks per day for U.S. adults. The most rigorous teen data comes from Common Sense Media's 2023 report Constant Companion: A Week in the Life of a Young Person's Smartphone Use, which used an installed monitoring app on the phones of 203 U.S. participants ages 11–17 over nine days — actual passive measurement, not self-report. Median teen pickups: 51/day. 44% of 16- to 17-year-olds picked up their phone more than 100 times a day. Median notifications: 237 per day; some teens received nearly 5,000 in 24 hours.

Why it matters for PullBack. When you cite numbers, cite the ones with methodology you can defend. The Common Sense passive-tracking dataset is the gold standard in the U.S. teen population. Skip the unattributed "average person checks their phone 96 times a day" stats — those usually trace back to a single Asurion press release.

Common Sense Media (2023). Constant Companion: A Week in the Life of a Young Person's Smartphone Use.
Common Sense Media press release: Teens Are Bombarded with Hundreds of Notifications a Day.
Deloitte (2023). Connected Consumer Survey.
Sewall, C. J. R., et al. (2022). Duration, frequency, and time distortion.

Plain English, three paragraphs

You're not weak. The feed is built by teams of engineers using the same variable-reward mechanics slot machines use to keep people in a trance — anthropologist Natasha Schüll documented all of it. PullBack doesn't try to out-design Instagram. It just sets a timer, lets you in, and pulls you out before the trance gets long. Most users underestimate their actual scroll time by over an hour a day, which is exactly the gap the timer fills.

Apps that nag you before you open Instagram get ignored within a week — your brain habituates. Apps that hard-block trigger reactance: you turn them off when you "really need" the app, and forget to turn them back on. PullBack interrupts you after you've already opened it. There's no fight about whether you're allowed in; you let yourself in. The timer just decides how long is long enough. That's the only window where the psychology works in your favor.

A 2023 PNAS study tested a friction-based intervention — a brief pause before opening a target app, with an easy option to bail out. Six weeks in, app openings were down 57%. The mechanism that mattered wasn't a mindfulness message; it was the bail-out option itself. PullBack applies the same exit-option psychology, but mid-session — we let you in, then give you the exit. Different timing, same mechanism, no replication study yet on our exact variant.

Citations are real, primary, and clickable. If you find a study that contradicts something here, email support@pullback.works — we'd rather be corrected than wrong.