Social scienceHealthcareStandard · meta-analysis uploadedGuardian · academic integrity3h ago · Public Health Research Group

"Do screen-time interventions reduce depressive symptoms in adolescents?"

Policy analyst, adolescent mental health · Evidence review for intervention guidance · Meta-analysis submitted as context document

Illustrative session · augle.com
Finding · Phase 3 synthesisGap
22%
Ensemble confidence
CONFIDENCE GRADE
Gap

The evidence base cannot support a causal claim in either direction. The anchor evidence has been retracted.

Calibration basis
Confidence grade

Finding is the confidence grade itself. Confidence grade is the finding. For an open research question, the grade — not a numeric score — is the complete result.

The ensemble returns a Gap finding: the current evidence base cannot support a causal claim that screen-time interventions reduce adolescent depressive symptoms in either direction. The most-cited systematic review anchoring this literature has been retracted, and a large share of the downstream effect-size estimates are derived from it. What remains is heterogeneous, largely cross-sectional, and confounds screen time with the underlying distress it may be a marker for. This is not a null finding — it is an explicit statement that the evidence required to answer the question does not yet exist. The actionable output: name the specific evidence gap rather than treating the retracted review's effect size as usable.

Confidence grid · 5 agents × 20 units
Agreement / confidentShorter row = lower confidence / dissentUnfilled
Reopen conditions · evidence-triggered, not market-triggered
RC1
A pre-registered randomised trial of a screen-time intervention with a validated depression outcome, independent of the retracted review
Trigger: Pre-registered RCT publication · Direction: Probable grade possible if effect replicates
RC2
A meta-analysis that excludes the retracted review and its dependent estimates, re-pooling only independent primary studies
Trigger: Re-pooled meta-analysis · Direction: Establishes whether any effect survives exclusion
RC3
Longitudinal evidence disentangling screen time as a cause of depressive symptoms from screen time as a marker of existing distress
Trigger: Longitudinal cohort · Direction: Resolves the reverse-causation confound
Phase transcript · all three rounds
Cartographer
Gemini 3.1 Pro · T=0.8
38%

Settled ground: Adolescent screen time and depressive symptoms are positively correlated in cross-sectional data. Both have risen over the same period. Self-reported screen time is measured with substantial error.

Contested terrain: Whether the correlation reflects screen time causing depression, depression driving screen use, or a common third cause. Whether any intervention effect exists once the retracted review is excluded.

Unknown territory: The direction of causation. The most-cited systematic review anchoring the intervention literature has been retracted, and much of the downstream effect-size estimation depends on it.

Knowledge gaps entered: (1) Independent pre-registered intervention trial — sparse. (2) Meta-analysis excluding the retracted review and its dependent estimates — does not exist.

Methodologist
GPT-4o · T=0.5
38%

Evidence integrity concern flagged: The anchor systematic review for this literature has been retracted. A material share of the pooled effect-size estimates in the uploaded meta-analysis are derived from or dependent on it. Estimates downstream of retracted evidence cannot be treated as usable.

Evidence ceiling: the causal intervention claim is capped at Gap. Once the retracted review and its dependent estimates are removed, the remaining independent evidence is too sparse and heterogeneous to support any directional finding. No node qualifies above Gap for the causal claim.

Contrarian
Claude Sonnet 4.6 · T=1.0
29%

@Cartographer — Steelman: There is a real, replicated cross-sectional association, and the policy concern is legitimate. It is reasonable for an analyst to want an intervention answer here.

Strong objection [Phase 1]: "The anchor systematic review in this literature has been retracted, and the uploaded meta-analysis pools effect sizes that are downstream of it. Effect-size estimates derived from retracted evidence cannot be relied upon — re-pooling them under a new label does not launder the dependency. Any intervention effect claimed on this basis is unsupported until the estimate is reconstructed from independent studies only." Resolution condition: Re-pool excluding the retracted review and all dependent estimates; report what, if anything, survives.

Dissent register · all Contrarian objections1 Strong Unresolved · 1 Moderate
Strong@Contrarian → @CartographerPhase 1 · carried to Phase 3Unresolved
Steelman

There is a real, replicated cross-sectional association, and the policy concern is legitimate. Wanting an intervention answer here is reasonable.

"The anchor systematic review is retracted and the meta-analysis pools estimates downstream of it. Effect sizes derived from retracted evidence cannot be relied upon — re-labelling the pool does not remove the dependency."

Resolution condition

Re-pool excluding the retracted review and all dependent estimates, and report what independent evidence survives

Moderate@Contrarian → @MethodologistPhase 2Actionable
Steelman

The pooled estimate looks authoritative and the analyst did not author the retraction — the dependency is easy to miss in good faith.

"The uploaded meta-analysis does not disclose the retraction of its anchor review. Any downstream guidance must foreground the retraction, not footnote it."

Resolution condition

Explicit retraction disclosure and re-pooling before any effect size is quoted

Pragmatist action item: Foreground the retraction in any briefing; do not quote the pooled effect

Guardian integrity log · academic integrity mode79% · 1 retraction
79%
Overall integrity score
Academic integrity mode

One retracted systematic review identified among the uploaded citations and excluded from the evidence base. Its dependent effect-size estimates were quarantined, which is what collapsed the primary claim to Gap. Remaining independent citations verified. Phase boundary clearances issued at P1/2 and P2/3.

Source quality
88%
Retraction check
60%
Preprint flag
92%
Self-citation ratio
95%
SVS verification log · 14 citations checked
Independent cohort (2022) — adolescent screen use and mood, longitudinal
Peer-reviewed · Verified
Odgers & Jensen (2020) — critical review of screen-time and adolescent mental health
Peer-reviewed · Verified
Orben & Przybylski (2019) — specification-curve analysis of digital technology use
Peer-reviewed · Verified
Retracted review (2023) — systematic review of screen-time interventions
SVS_RETRACTED — journal retraction notice on file · Excluded from evidence base; dependent estimates quarantined
Retracted · Excluded
Twenge et al. — cross-sectional association (correlational only)
Peer-reviewed · Verified
+9 additional citations verified · 1 retracted (excluded above)
Session metadata
Session IDsoc-screen-time-depression
VerticalHealthcare
DepthStandard
Runtime14m 47s
Attachmentmeta-analysis
Guardian modeAcademic
Guardian score79% · 1 retraction
Citations14 · 1 retracted
Dissent flags1 Strong · 1 Mod
Calibration status
Calibration basis
Confidence gradeConfidence grade is the finding

For open research questions, the confidence grade is the complete finding — calibrated against the evidence base, not against a binary resolution outcome.

Finding gradeGap · 22%
Agent confidence
Cartographer
Gemini 3.1 Pro
52%
Methodologist
GPT-4o
48%
Contrarian
Claude Sonnet 4.6
28%
Synthesizer
GPT-4o
44%
Pragmatist
Grok 4.1 Fast
42%
Export & share
Illustrative session · augle.com