Substance and claimed effects

News consumption in this entry refers to the modern, algorithmically-curated, variable-reward delivery of current events and social commentary — the smartphone-era stream of push notifications, infinite-scroll feeds, headline aggregators, cable-news rotations, podcast newscasts, and the comment threads bolted onto each. The substance is not "being informed" — it is the delivery format itself: feeds engineered for engagement on a variable-ratio reinforcement schedule, optimised for emotional arousal, retweetable outrage, and time-on-site. Claimed effects span attention (fragmentation, multitasking residue), mood (anxiety, anger, doom), social comparison (status stress, FOMO), sleep (delayed onset, fragmentation), worldview accuracy (skewed risk perception, polarisation), decision quality (recency bias, threat-driven choices), and time use (hours per day reallocated from sleep, exercise, in-person relationships). The intervention layered on top — deliberate selection and rationing — is the entry's recommended action: curated sources, fixed time windows, removed apps, replaced defaults. The substance and the antidote are covered as one entry because the antidote is defined by the substance it counters.

Evidence by addressing question

Mechanism

Four mechanisms compound, and any one of them would be enough to make the format problematic; together they explain why the habit is hard to moderate by willpower alone.

Variable-ratio reinforcement. Feeds dispense reward (a funny tweet, a notable headline, a flattering photo, an outrage hit) on an unpredictable schedule — the same schedule that produces the most persistent operant behaviour in laboratory animals and the highest hold-time in slot machines. The dopaminergic response is keyed to anticipation of reward, not the reward itself; refreshing a feed produces the same neurochemical pattern as a pull on a lever, and behaviour extinguishes very slowly even after rewards thin out.

Negativity bias in attention and supply. Humans react faster and more strongly to negative stimuli than to positive ones — a 17-country physiological study found the asymmetry in skin conductance and heart-rate responses to news content was present across every population sampled Soroka et al. 2019. Engagement-optimised algorithms learn this and select for it: false news travels roughly six times faster than true news on Twitter, and the spread is driven by novelty and emotional reaction, not by bot activity Vosoughi et al. 2018. Moral-emotional language increases the diffusion of political content by ~20% per emotion word in a tweet Brady et al. 2017.

Social comparison. Feeds present a continuously refreshed sample of curated peer lives. Experimental exposure to Facebook profiles depresses state self-esteem versus a neutral control task Vogel et al. 2014. The mechanism is upward social comparison — the brain treats a stranger's vacation photo as evidence about its own relative standing — and is largely automatic.

Arousal and light at the wrong hour. Evening screen reading from a light-emitting device suppresses melatonin by ~55%, delays circadian phase by ~1.5 hours, lengthens sleep-onset latency, and reduces next-morning alertness in a controlled cross-over trial Chang et al. 2015. The content layered on top — outrage threads, doom headlines — adds cognitive arousal that compounds the light effect.

Evidence

The literature is dense and the effect sizes are mostly modest in observational work, but converge across designs.

Experimental deactivation. A four-week paid Facebook deactivation RCT (n≈2 700) found improvements in subjective wellbeing of roughly 0.09 SD, large reductions in political polarisation, and about an hour per day reallocated to offline activities, with persistent reductions in Facebook use after the experiment ended Allcott et al. 2020. Limiting Instagram, Facebook, and Snapchat to 10 minutes per platform per day for three weeks (n=143 undergraduates) produced significant declines in loneliness and depressive symptoms versus a use-as-usual control Hunt et al. 2018. A two-week Facebook break with active healthy lifestyle reinforcement increased life satisfaction and reduced depressive symptoms at six-month follow-up Brailovskaia & Margraf 2022.

Natural experiment. The staggered rollout of Facebook across U.S. college campuses (2004–2006) caused a measurable increase in severe depression and anxiety among students, identified by Braghieri, Levy, and Makarin via diff-in-diff against pre-rollout cohorts, with the mechanism running through unfavorable social comparison Braghieri et al. 2022. This is one of the cleanest causal designs in the literature because the rollout order was plausibly exogenous.

Observational and longitudinal. Experience-sampling data on Facebook (texted at five random times per day for two weeks) found that more use between samples predicted lower moment-to-moment wellbeing and lower life satisfaction at endline, with the relationship in that direction (Facebook use → wellbeing decline) and not the reverse Kross et al. 2013. The 2010-onset divergence in adolescent depressive symptoms and suicide rates correlates strongly with the smartphone-and-social-media transition Twenge et al. 2018; a competing specification-curve analysis on the same kind of cross-sectional data found the effect of digital screen time on adolescent wellbeing was small (0.4% of variance explained) and comparable to the effect of eating potatoes Orben & Przybylski 2019. Reconciling: pooled all-screen-time effects are small, but social media specifically and heavy use specifically are stronger, and natural-experiment evidence is closer to the larger end.

News-specific exposure. A short laboratory exposure to 14 minutes of negative TV news bulletins increased anxious and sad mood and exacerbated personal worries unrelated to the news content Johnston & Davey 1997. During COVID, hours per day of pandemic news consumption tracked anxiety, depression, and pandemic-specific fear in dose-response fashion in a representative German sample of ~6 500 adults Bendau et al. 2021. Daily diary data during the pandemic showed within-person increases in social-media news exposure predicted same-day increases in PTSD-like symptoms and depressive symptoms Price et al. 2022. A probability-based national U.S. cohort followed across the pandemic's first year found that high baseline media exposure to collective trauma predicted worsening mental and physical health symptoms over the following months Holman et al. 2020.

Attention and task-switching cost. Field studies of knowledge workers show that the average self-interruption recovery time after a notification is on the order of 20 minutes to resume the original task; even brief interruptions raise stress, frustration, and time pressure Mark et al. 2008.

Sleep. A systematic review of 67 studies (most on children and adolescents) found a consistent positive association between screen time near bedtime and shorter sleep duration, later bedtimes, and poorer sleep quality across study designs Hale & Guan 2015. The mechanism is partly photic (see Chang above) and partly arousal-content driven.

Protocol

The intervention with the most experimental support is quantity reduction, not content filtering. Limiting feed apps to ~10 minutes per platform per day produced measurable mood gains within three weeks Hunt et al. 2018; full deactivation for four weeks produced larger effects and persisted post-intervention Allcott et al. 2020. Practitioner-recommended levers in order of behavioural friction needed: remove apps from the phone (use only on desktop); turn off all push notifications; greyscale the phone display; install a blocker (Freedom, Cold Turkey, Screen Time) with a deliberately inconvenient unlock; replace feeds with selected newsletters or a single morning-paper visit; restrict to fixed time windows (e.g., one 30-minute morning window, one 30-minute evening window, nothing in between or in bed); designate device-free rooms (bedroom is the priority) and times (first and last hour of the day). The smaller the friction the rationing requires of in-the-moment willpower, the more it works — the variable-reward mechanism is not something willpower beats reliably.

Contraindications

No medical contraindications. Important caveats: (1) people with diagnosed anxiety, depression, or PTSD may experience stronger acute distress with heavy news exposure Holman et al. 2020 Price et al. 2022, and benefit disproportionately from rationing, but should not abruptly cut all sources of social contact if social media is their primary social tie — substitute first, then ration; (2) acute crises (natural disasters, security threats) may warrant temporarily higher monitoring, but a single trusted source on a scheduled check beats refreshing.

Misconceptions

The two most common: "I need to stay informed." Worldview accuracy does not scale with feed-hours; it scales with source quality and time spent on long-form. Brady et al. and Vosoughi et al. show feed-based exposure selects for emotionally activating and often false content over accurate content Brady et al. 2017 Vosoughi et al. 2018; the marginal hour spent scrolling typically lowers accuracy, not raises it. "I'm just relaxing." The experience-sampling and deactivation literature both find that the felt-experience of scrolling is "neutral" or "fine" in the moment, but predicts lower wellbeing on aggregate Kross et al. 2013 Allcott et al. 2020. The format is bad at signalling its own cost.

Failure modes

Common patterns: (1) Substitute scrolling — quit Instagram, pick up Reddit. The variable-reward delivery is what matters; the platform is interchangeable. Block the format, not the brand. (2) Partial bans — "I'll just check it once" reliably fails because the schedule is variable-ratio; the reward might come this check. Block-mode tools work because they remove the option in the moment. (3) News-as-coping — checking the feed when anxious tends to amplify anxiety in the short term but feels productive ("I'm doing something about it"), creating a self-perpetuating loop seen clearly in COVID-era daily-diary data Price et al. 2022. (4) Falling back during stress — major news events undo months of habit work; pre-committing to a single source and a single check per day during crises is the most-recommended practitioner workaround.

Stakes

What continues to happen if the habit doesn't change: ~2.5 hours per day of feed time at U.S. average rates, of which most users self-report wanting back; a sleep debt accumulated through later bedtimes and fragmented sleep Hale & Guan 2015 Chang et al. 2015; a chronic, low-grade anxiety and threat-monitoring state from prolonged exposure to negativity-selected content Johnston & Davey 1997 Holman et al. 2020; degraded sustained attention from chronic task-switching Mark et al. 2008; persistent social-comparison drift in self-evaluation Vogel et al. 2014 Braghieri et al. 2022; and a worldview increasingly indexed to high-arousal, low-base-rate events.

Payoff

The Allcott deactivation trial gives the cleanest payoff portrait: ~60 minutes per day reallocated, measurable wellbeing improvement (~0.09 SD on a composite happiness/life-satisfaction/anxiety/depression scale), and reduced political polarisation Allcott et al. 2020. Hunt et al.'s 30-minute-per-day cap produced loneliness and depression reductions within three weeks Hunt et al. 2018. Brailovskaia & Margraf show that effects on life satisfaction can persist six months after the intervention ends Brailovskaia & Margraf 2022. Onset latency: mood effects show within days; sleep effects within a week; attention rebuild over weeks to months; relationships shift over months.

Out of scope

Smartphone use generally; gaming; pornography; dopamine-fasting protocols; specific platform bans by jurisdiction; media literacy curricula; the question of regulation. Each of these is a distinct substance with distinct evidence.

The credibility range

Optimist case (the format is fine; the problem is the user). Effect sizes in the largest specification-curve analyses are small — the digital-screen-time / wellbeing association explains less than half a percent of variance in adolescent wellbeing and is comparable to mundane behaviours like wearing glasses or eating potatoes Orben & Przybylski 2019. Many of the strongest correlational claims rely on between-person comparisons confounded by personality, baseline distress, and social context. The format genuinely serves real needs — staying connected to dispersed family and friends, professional networking, news access, civic participation — and a moderation framing may be more accurate than a pathology framing.

Skeptic case (the format is harmful at scale and the trial evidence is now clean). Experimental deactivation trials with thousands of participants (Allcott) and natural-experiment causal designs (Braghieri) both show meaningful negative effects, with the natural-experiment effect on severe depression being non-trivial. Daily-diary and experience-sampling designs eliminate the between-person confound and consistently find within-person negative associations Kross et al. 2013 Price et al. 2022. The mechanism is well-characterised (variable-ratio reinforcement plus negativity-selected content), runs through pathways with independent evidence (sleep, attention, social comparison), and matches felt experience reported broadly. The Orben/Przybylski small-effect finding pools all screen activities (including video calling a grandparent) with feed-scrolling and would be expected to wash out a signal driven by a narrower subset.

Author's call. The skeptic case is the better-supported reading once the experimental and natural-experiment evidence is given priority over the pooled cross-sectional work. The format produces real, replicated, measurable harm on mood, sleep, attention, and time use; the harm is larger for heavier users and for adolescents; the mechanism is overdetermined; and the cheapest intervention (rationing, removing apps, blocking notifications) has the largest evidence base. Controversy remains around effect-size magnitude and around adolescent-specific causal pathways, which is why controversy is non-zero. The entry lands as do (the rationing), with strong evidence for the substance's harm and a moderate-to-strong effect across most dimensions it touches.

Stakeholder and incentive map

• Platform incentive. Meta, X, TikTok, YouTube, Reddit, and the cable-news networks earn revenue proportional to time-on-platform. Their product teams have published, modelled, and patented engagement-maximising recommendation systems; internal documents made public during litigation and whistleblower disclosures show platforms know about wellbeing harms and consistently prioritise engagement.
• Advertising and news-organisation incentive. Ad-supported news has an inverted economic incentive — more clicks, longer sessions, more page-views — that selects against the calmer, slower, lower-engagement format that produces accurate worldviews. Subscription-supported long-form (broadsheets, specialist newsletters, magazines) has a partly aligned reader-success incentive.
• Academic and clinical. Clinical psychology, sleep medicine, and behavioural-addiction researchers broadly recommend rationing; the AAP and similar bodies have issued guidance on adolescent and child screen time. No specialty has financial stake in the recommendation.
• Counter-movement. Digital-minimalism communities (Cal Newport's work, dumbphone communities, the Center for Humane Technology) push the same recommendation from a non-academic angle; their conclusions are largely consistent with the trial evidence.

Population variability

Effects are heaviest at the tails: heavy users (3+ hours/day) and adolescents both show larger associations than light users and adults. Women report greater body-image and social-comparison harm than men; men report greater outrage and political-content-driven anger. Pre-existing anxiety and depression amplify acute negative response to news content Holman et al. 2020. Sleep effects of evening screen use are largest in adolescents and morning chronotypes but appear in every age group studied Hale & Guan 2015. For socially isolated individuals (rural older adults, disabled-at-home, expat communities), the substitution cost of removing social media is higher — the recommendation is to ration the feed but preserve the messaging functions, which most platforms decouple imperfectly.

Knowledge gaps

• Long-term (5+ year) effects of sustained rationing on the same outcome panel — almost all RCTs are 2–4 weeks; the Brailovskaia six-month follow-up is the longer end.
• Effects of TikTok and short-form video specifically — the literature is dominated by Facebook/Instagram-era studies, and the short-form-video format may be qualitatively more intense on the variable-reward axis.
• Mechanism of the apparent dose-response relationship at the heavy end (3+ hours): is harm linear or does it inflect, and does the inflection match a clinical threshold?
• Whether content-filtering (curated sources, no comments, no feed) is sufficient or whether the variable-reward delivery format itself must be removed. Most trials cut quantity, not format.
• Civic and information-quality side effects of broad rationing — would a population that consumes less news be better or worse at decisions that depend on current-events knowledge?

Scope decision. The brief is titled "News Consumption" but the substance treated in the article is the broader variable-reward feed format — news and social media, since the mechanism (variable-ratio reinforcement, negativity-selected content, light-and-arousal at bedtime) is the same and the strongest trial evidence (Allcott, Hunt, Braghieri) is on social-media platforms specifically. Treating them as one substance is what the editorial spec asks for: a substance and all its meaningful consequences in one entry. Title updated from "News Consumption" to "News and Social-Feed Consumption" to be honest about coverage.

All eight brief-named consequences covered. Attention (mechanism + payoff + focus score 4), mood / anxiety (evidence + mood score 4), comparison-driven stress (mechanism + evidence + Braghieri callout), sleep (mechanism + stakes + sleep score 3), world-view accuracy (misconceptions + stakes), decision quality (stakes), time use (stakes + payoff). No silent narrowing.

Action call. Scored as do rather than avoid because the active behaviour is the rationing protocol — a positive ongoing practice — not abstention from news in general. Cadence daily follows: the windows are a daily rhythm.

Evidence at 3, not 4. The trial evidence on rationing is strong (Allcott n≈2,700 deactivation RCT; Hunt RCT; Braghieri natural experiment), but effect-size magnitude is genuinely contested (Orben & Przybylski 2019 specification curve gives a much smaller pooled number). 3 reflects "real, replicated, mechanism well-characterised, magnitude under live debate".

Controversy at 3. The Twenge/Haidt vs. Orben/Przybylski debate is active and the adolescent-causal-pathway question is unresolved. Not 4 because the direction of effect is no longer seriously contested; only the magnitude is.

beauty_cumulative and longevity dropped from 1 to 0. Both initially scored 1 for the indirect pathway through sleep-quality and chronic-stress reduction, but the substance is not meaningfully an aesthetic or longevity intervention in its own right, and the body wasn't going to give either dimension a real home without padding. The spec's guidance to "score 0 freely" is the right call here.

Audience left unscoped. Effects are heaviest in adolescents and heavy users, but the substance is universal and the catalogue's audience scoping is for narrowing the entry, not for noting where effects are stronger. The dossier records the population variability.

Separate-entry candidates.

• Short-form video (TikTok, Reels, Shorts) — the existing literature is mostly Facebook/Instagram-era; the short-form video format is likely qualitatively more intense on the variable-reward axis and may warrant its own write-up once the trial evidence catches up.
• Adolescent feed use specifically — the developmental angle (Twenge, Haidt, Braghieri-on-students) has its own evidence base and policy stakes; a separate entry could go deeper than the universal-audience treatment here allows.
• Smartphone use in general — the device, not the content format. Adjacent but distinct substance.

Future links. Once written: sleep environment / bedroom-phone bans, deep-work habits, the smartphone-as-device entry, and a possible entry on outrage / political media specifically.

Hard call during the write. Considered including the Stieger / Lancet Digital Health social-media-break trial but the literature has enough trial evidence already and the marginal cite didn't change anything. Skipped to keep the dossier focused.