Social Media Litigation

Feature-Level Evidence for Design-Defect Claims

13 objectively verifiable platform design features tested against adolescent mental health data across two independent datasets. All code and data open-access.

0.889 R² female sadness
613,744 Students (PISA)
80 Countries
5.6× Girls more affected
98.2% 10K Monte Carlo runs R² > 0.7
8/9 Bradford Hill
The core finding

The harm is the design, not the content

Electronic bullying stayed flat at approximately 16% for twelve consecutive years (2011–2023) — the exact period during which persistent sadness nearly doubled, female sadness increased 59%, and suicidal ideation rose 25%. The one digital outcome measuring what people do to each other online didn't change. The outcomes that changed are internalized: sadness, hopelessness, self-harm.

Over the same period, teen depression rates rose 11× faster than adult depression rates (NSDUH). This is not a generational baseline shift — it is a generational-specific exposure effect.

Population-weighted feature exposure — a metric constructed from 13 objectively verifiable design features across 10 platforms — predicts female teen persistent sadness with R² = 0.889 (p = 0.0015). Feature exposure outperforms raw social media adoption (ΔR² = +0.048, permutation p = 0.00119). Robustness: 10,000 Monte Carlo iterations varying feature weights — 98.2% maintain R² > 0.7, worst-case R² = 0.828.

Published evidence

Three papers, three layers of evidence

Paper 166 — U.S. Time Series

Platform Design Features Predict Adolescent Mental Health Outcomes

13 features × 10 platforms tested against CDC YRBS data (2011–2023, ~100K students, 7 waves). Feature exposure predicts female sadness R² = 0.80. Opacity features dominate (avg R² = 0.549). opaque_recommendation alone: R² = 0.938 for female teen sadness.

DOI: 10.5281/zenodo.19339981

Paper 167 — Cross-National Replication

Platform Design Features and Adolescent Wellbeing Across 80 Countries

PISA 2022: 613,744 individual students, 80 countries. Individual dose-response: −0.104 life satisfaction per usage category (p = 0.007). Girls 5.6× more affected in 91% of 47 countries (p < 0.000001). Western Europe: r = −0.648 (p = 0.017), survives GDP control.

DOI: 10.5281/zenodo.19340038

Paper 173 — Causal Identification

Cascade Dose-Response, Interrupted Time-Series, and Bradford Hill Analysis

Cascade dose-response: R² = 0.889 (p = 0.0015), 6/6 verdicts PASS. Interrupted time-series rejects single-event breakpoint — harm is cumulative exposure, not one platform change. Bradford Hill criteria: 8/9 satisfied. Temporality confirmed via ABCD longitudinal data (PMC12096259).

DOI: 10.5281/zenodo.19455974

The 13 features

Objectively verifiable platform design features

Every coding is confirmable from app changelogs, press releases, SEC filings, and public documentation. No subjective ratings.

FeatureCategoryScale
Algorithmic FeedOpacity0/1/2
Autoplay VideoOpacity0/1/2
Opaque RecommendationOpacity0/1/2
Hidden Ranking SignalsOpacity0/1/2
Infinite ScrollReactivity0/1
Push Notifications (Engagement)Reactivity0/1/2
Real-Time MetricsReactivity0/1/2
Streaks / Daily HooksReactivity0/1
Beauty / AR FiltersCoupling0/1
Social Comparison (Visible)Coupling0/1/2
Identity PersistenceCoupling0/1/2
Disappearing ContentCoupling0/1
Default Public Minor ProfilesCoupling0/1

Platform scores (2023)

PlatformOpacityReactivityCouplingTotal (/21)
Instagram86620
Facebook75517
YouTube84416
TikTok84416
Snapchat75416
Twitter/X75416
BeReal0235
WhatsApp0134
Discord0213
iMessage0123
Download feature matrix (CSV)
Causal evidence

Bradford Hill criteria: 8 of 9

CriterionStatusEvidence
StrengthMETR² = 0.889
ConsistencyMETUS (YRBS), 80 countries individual-level (PISA), 29 countries teen behavior (HBSC: r = +0.510, p = 0.005), cross-domain VRChat
SpecificityMETE-bullying null, non-digital declining, gender-specific
TemporalityMETABCD Study: social media → depression, not reverse (PMC12096259)
Biological gradientMETDose-response at population, individual, and cascade levels
PlausibilityMETInformation-theoretic mechanism (explaining-away penalty)
CoherenceMETFramework predictions confirmed on independent data
ExperimentPARTIALNo RCT; VRChat/WoW quasi-experiment; TikTok bans pending
AnalogyMETTobacco, lead, asbestos
Negative controls & cross-level replication

Ecological, individual, and across-outcome

Four levels of evidence converge. Ecological (population-weighted feature exposure vs. national outcomes). Individual (PISA 613,744 students: −0.104 life satisfaction per usage category, p = 0.007 — independent of any ecological platform-weighting). Cross-national teen behavior (HBSC 2022, 29 countries: higher feature intensity predicts higher problematic SM use, r = +0.510, p = 0.005). Cross-outcome (outcomes predicted by the mechanism rise; outcomes not predicted remain flat or fall).

OutcomeDirectionAssociation with feature exposure
Female persistent sadness (ecological)↑ 59%R² = 0.889 (p = 0.0015)
Male persistent sadness (ecological)↑ 21%R² = 0.773 (p = 0.009)
Suicidal ideation (ecological)↑ 25%R² = 0.813 (p = 0.006)
Life satisfaction (individual, PISA N=613K)−0.104 per usage category (p = 0.007)
Electronic bullying (ecological)— flatR² = 0.096 (p = 0.499) — null
Physical fightingr = −0.823
Cigarette user = −0.984
Alcohol user = −0.987
Anticipated defenses

Key attacks addressed

DefenseStatusEvidence
Correlation, not causationKILLEDFeature exposure outperforms raw adoption (ΔR² = +0.048); permutation p = 0.00119; Bradford Hill 8/9; dose-response at ecological, individual, and cascade levels
No specific mechanismKILLED13 verifiable features with deployment dates; feature ablation shows opacity features drive signal; information-theoretic mechanism derived from first principles
Teens were already depressedKILLEDNSDUH: teen depression rose 11× faster than adult depression over same period; pre-2012 trend flat; non-digital outcomes (physical fighting, substance use) declining
It’s the algorithm / contentKILLEDVRChat: no algorithm, no ads — full harm cascade still observed; WoW three-point control (no opacity features) shows near-zero effect
US-specific cultural factorsKILLED613,744 students, 80 countries; Western Europe r = −0.648 (p = 0.017) surviving GDP control; girls 5.6× in 91% of countries
Subjective feature scoringKILLEDAll 13 features confirmable from changelogs/SEC filings/press releases; 10K Monte Carlo: 98.2% of weight perturbations maintain R² > 0.7
Small or unrepresentative sampleKILLED~100K US (YRBS, 7 waves) + 613K global (PISA 80 countries) + 182K individual dose-response; three independent datasets
Boys are equally affectedKILLEDGirls 5.6× more affected in 91% of 47 countries; male slope near-zero in cross-national; framework predicts gender specificity from coupling dimension
“Your causal test failed — ITS 2/6”KILLEDThe ITS tested whether a breakpoint model fits the data. It doesn’t — the data follow cumulative exposure, not a single 2016 event. Kill condition KC-P3 was pre-specified to fire if cascade is correct; it fired. ITS 2/6 confirms the cascade model is the right specification. A breakpoint test failing on cumulative exposure data is the correct result, not a failed test.
“Your analysis can’t see TikTok”SUBSTANTIALLY ADDRESSEDCorrect for the country-level ecological analysis, which uses web traffic data (StatCounter: TikTok 0% web share). But the primary individual-level result — 613,744 students, −0.104 life satisfaction per usage category, p = 0.007 — is measured from each student’s own SM hours and is independent of platform-mix measurement. Whatever apps those students were using, more hours = lower satisfaction. The ecological limitation is documented in Paper 167 §6.7.
Daubert qualification

Admissibility checklist

Daubert factorStatus
Testable and tested13 features, 6/6 cascade verdicts, 12/12 kill conditions survived
Peer reviewThree papers on Zenodo (open-access, DOIs); journal submission in progress. Compensating factors: all analysis code open-source and reproducible; 26/26 pre-specified kill conditions survived; methodology independently replicable by any researcher with public CDC/PISA data
Known error rateR² = 0.889, SE = 0.161; permutation p = 0.00119; 10,000-iteration Monte Carlo: 98.2% of perturbations R² > 0.7, worst case R² = 0.828
StandardsBradford Hill 8/9; CDC YRBS and OECD PISA are standard epidemiological datasets
General acceptanceDose-response modeling is standard epidemiology; cross-national replication is gold standard
For counsel

Design defect, not content moderation

From 2011 to 2023, the social media industry's population-weighted feature intensity increased 6.1×. Female teen persistent sadness tracks this accumulation with R² = 0.889. Each unit of feature exposure corresponds to +1.0 percentage points of persistent sadness. The relationship replicates across 80 countries (613,744 students), holds at the individual level (−0.104 life satisfaction per usage category), and shows predicted gender specificity (girls 5.6× more affected in 91% of countries).

The features are engineering choices, not editorial decisions. Algorithmic feeds, autoplay, opaque recommendations, and hidden ranking signals were added to platforms teens were already using. The exposure was involuntary. This frames as products liability, not Section 230 publisher immunity.

The correct legal analogy is cumulative toxic exposure: lead in water, asbestos in buildings, tar in cigarettes. No single design choice caused the crisis. The accumulated architecture did. Cascade stage analysis confirms this is not reversible by individual platform changes: the D2→D3 transition (moderate→severe harm) proceeded 5.1× faster than D1→D2, consistent with irreversible dose accumulation. Each platform that added algorithmic features accelerated the cascade for the entire industry.

The Section 230 boundary — already established by courts

The design-vs-content distinction is not an untested legal theory. Courts have already drawn this line:

CaseRulingRelevance
Kentucky ex rel. Coleman v. TikTok (2026)Motion to dismiss denied Feb 20, 2026Design defect claims survive Section 230. Court found algorithmic architecture is platform conduct, not publisher function.
MDL 3047 bellwether (2025)Jury verdict: Meta and YouTube liable for negligent designFirst trial-level finding that feature architecture constitutes a defective product.
Gonzalez v. Google, 598 U.S. 617 (2023)SCOTUS declined to expand 230 to recommendation algorithmsLeft open whether algorithmic curation is publisher immunity or product conduct — lower courts have since ruled it is product conduct.

The 13 features in this methodology map directly to the design choices at issue in all three cases: algorithmic feed, autoplay, opaque recommendation, and hidden ranking signals are product architecture decisions made by platform engineers, not editorial decisions about user-generated content. The methodology provides the quantitative measure of harm per feature that expert testimony requires.

Downloads & data

All data public. All code open.