Solutions · Think tanks + nonprofits

Research that shapes
public debate needs
this standard.

Think tank publications and advocacy positions enter the public record and stay there. A methodology weakness that a hostile commentator finds, a statistic that doesn’t survive scrutiny from a rival institution, or a policy recommendation whose evidence base has been superseded — these become the story. Augle runs the adversarial review before publication, not after.

Research directors reviewing pre-publication policy papers
Advocacy teams stress-testing campaign evidence bases
Communications leads preparing for hostile media questions
Programme officers evaluating grant-funded research claims
Session configuration
Guardian modeAcademic + Editorial
DepthStandard · Deep
DocumentsPolicy papers · research reports · briefings
SVS checksRetraction DB · source independence · preprint status · statistical claims
OutputFinding · confidence grade · unresolved objections · audit trail
The problem

Published research is permanent. Corrections aren’t.

A think tank’s credibility is its only asset. A policy paper that makes a statistical claim a hostile statistician can dismantle in a tweet, or an advocacy position whose primary study has been superseded, damages that credibility in ways a correction doesn’t repair. The standard for publication should be higher than the standard for friendly peer review. Augle applies the adversarial standard before you publish.

Questions think tanks run on Augle

What is the strongest methodological objection to our headline claim, and can we defend against it?
Are any of the studies our recommendation depends on retracted, superseded, or subject to replication failure?
What will a hostile commentator at a rival institution say about this paper within 48 hours of publication?
Does our statistical framing accurately represent what the cited study actually found?
Which of our policy recommendations have the weakest evidence base and need the most caveating?
Statistical claims that don’t survive hostile scrutiny

The Methodologist evaluates every statistical claim against the study design that generated it — whether the effect size is appropriately qualified, whether relative risk is being presented without the absolute baseline, whether a correlation is being presented as causal. These are the objections that become quote-tweets within 48 hours of publication.

Primary studies that have been retracted or superseded

The Guardian’s SVS checks every citation against retraction databases and assesses replication status. A policy recommendation that depends on a study with known replication failures, or a statistic drawn from a retracted paper, is flagged before it enters the published record — not discovered by an opponent after publication.

Advocacy positions with asymmetric evidence

The Contrarian evaluates the full evidence landscape, not just the evidence the paper cites. A position paper that presents a contested claim as settled, or that cites studies selectively while ignoring materially relevant contrary evidence, will not survive a well-resourced critic. Augle surfaces both sides before publication.

Recommendations whose generalisability doesn’t hold

The Methodologist evaluates external validity — whether the evidence base for a recommendation applies to the population, context, and scale the recommendation targets. A policy drawn from a randomised trial in a different country, or a programme evaluation with a non-representative sample, requires explicit qualification.

How it works in practice

Three sessions.
Three publication moments.

Each scenario illustrates realistic deliberation behaviour across policy publication, advocacy preparation, and research quality review.

Use case 01 of 03 · Think tanks + nonprofits
The Pre-Publication Review
Research Director · Health policy think tank · Pre-launch paper review
Standard depth · document ingestion

Does our paper’s headline claim — that universal free school meals reduce childhood obesity rates by 18% — survive methodological scrutiny, and what will critics say?

Guardian SVS28 citations verified. One primary study — the source of the 18% figure — is a preprint that has not completed peer review. Flagged SVS_UNVERIFIED. Evidence node capped at Probable. The paper presents the figure without this qualification.
MethodologistExternal validity problem: the primary study was conducted in a Danish municipality with universal baseline nutrition education. The UK context the paper targets has no equivalent baseline. The effect size is not directly transferable without adjustment. Confidence ceiling: Contested for UK-specific claims.
ContrarianStrong objection (Unresolved): "The 18% figure is from a preprint, presented as established fact, derived from a context with materially different baseline conditions. A hostile commentator with basic methodological knowledge will identify all three problems in the abstract. This will be the story, not the recommendation."
Pragmatist(1) Add preprint qualification to the 18% citation. (2) Qualify the UK generalisability claim explicitly. (3) Present a range based on studies with stronger external validity. Three edits prevent the paper from being reduced to a methodology critique rather than a policy contribution.
Session output
SVS flagPrimary source is unreviewed preprint — presented as established fact
FindingContested for UK-specific obesity reduction claims
Key objectionContext mismatch — Danish baseline nutrition conditions not present in UK target context
Edits neededPreprint qualification + UK generalisability caveat + effect size range
The value

The paper was two days from publication. The preprint status of the primary source hadn’t been flagged in internal review. Three targeted edits moved the paper from methodologically vulnerable to defensible — and kept the policy recommendation, which was sound, from being buried under a methodology story.

Use case 02 of 03 · Think tanks + nonprofits
The Advocacy Position Review
Campaign Director · Criminal justice reform nonprofit · Pre-launch advocacy materials
Standard depth

Is the evidence base for our campaign’s central claim — that community supervision is more effective than incarceration at reducing reoffending — strong enough to withstand a hostile media cycle?

CartographerSettled: community supervision outperforms incarceration on reoffending in low-to-medium risk offender populations in controlled studies. Contested: whether this effect holds for high-risk offenders and violent offence categories. Unknown: long-run community safety outcomes beyond 5-year follow-up windows.
MethodologistScope problem: the campaign materials present the finding as universal. The primary evidence base is limited to low-to-medium risk, non-violent offender populations. Extending the claim to all offender categories — which the campaign materials do implicitly — is not supported. Confidence ceiling for universal claim: Contested.
ContrarianModerate objection: "A journalist covering a high-profile violent reoffending case will immediately ask whether community supervision applies to violent offenders. The campaign has no answer because the materials don’t scope the claim. This is a predictable and entirely avoidable vulnerability."
Session output
FindingEstablished for low-to-medium risk non-violent populations · Gap for high-risk and violent categories
Key vulnerabilityUniversal claim not scoped — exposes campaign to predictable hostile question
Next stepScope the claim explicitly in all materials. Prepare a one-page FAQ for the violent offender question.
The value

The campaign’s evidence base was strong — for the population the research actually covered. The materials had silently extended it to all offenders. One scoping edit and a prepared FAQ for the predictable hostile question turned a vulnerability into a demonstration of research rigour.

Use case 03 of 03 · Think tanks + nonprofits
The Grant Research Evaluation
Programme Officer · Foundation · Evaluating grantee research claims pre-funding
Deep depth · document ingestion

Does the evidence base cited in this grant application support the causal claims the applicant makes about their programme’s impact on educational attainment?

CartographerSettled: structured mentoring programmes improve attendance and engagement in the cited studies. Contested: whether attendance improvements translate to attainment gains at the scale proposed. Unknown: whether the effect holds in the specific demographic and geographic context the applicant targets.
Guardian SVSAll 14 citations verified. One study cited in the impact section was published in a journal that has since been flagged for peer review concerns by the Committee on Publication Ethics. Flagged SVS_UNVERIFIED. Evidence node downgraded.
ContrarianStrong objection (Unresolved): "The application presents attendance improvement as a causal pathway to attainment. The cited studies establish correlation, not causation. The causal mechanism is asserted, not evidenced. This is the gap a rigorous external evaluator will find."
Session output
FindingProbable for attendance effect · Contested for causal attainment claim
SVS flagOne citation from journal flagged by COPE for peer review concerns
Key gapCausal mechanism asserted not evidenced — correlation presented as causation
DecisionFund conditional on applicant addressing causal mechanism gap with additional evidence or scoped claim
The value

The application was compelling and the programme was promising. The causal claim overreach and the journal integrity flag were not visible in a standard desk review. A conditional grant with a scoped claim requirement protected the foundation’s evaluation record — and gave the applicant a clear path to funding.

How Augle works for think tanks

The adversarial review
before publication.

1
Submit your research materials

Upload policy papers, research reports, briefings, and supporting studies. Academic and editorial integrity modes activate — the Guardian checks retraction status, preprint and peer review classification, source independence, and statistical claim validity before any agent receives the evidence.

2
The ensemble maps what’s settled and what’s contested

The Cartographer classifies every claim as Settled, Contested, or Unknown. The Methodologist assesses whether the evidence supports the conclusions drawn — effect size qualification, external validity, causal inference. Confidence bounds are set before deliberation begins, not after.

3
The hostile critic’s objections are run

The Contrarian takes the role of a well-resourced critic at a rival institution or a hostile journalist — surfacing the strongest version of every objection. Every challenge specifies a resolution condition. Unresolved Strong objections appear verbatim with a resolution condition. These become the pre-publication edit list.

4
You receive a defensible publication

The Synthesizer produces a finding anchored to the evidence base — not the position the paper argues. Confidence grades per claim, unresolved objections verbatim, and reopen conditions that specify what new evidence would change the assessment. The paper that emerges is the one that survives hostile scrutiny.

Think tank session · configuration
Guardian mode
Academic + Editorial — retraction database, preprint status, peer review classification, source independence, statistical misrepresentation detection
Document types
Policy papers · Research reports · Briefings · Supporting studies · Grant applications · Advocacy materials
Contrarian focus
Hostile commentator framing · statistical overreach · causal inference gaps · generalisability limits · selective citation
Output package
Confidence grade per claim · unresolved objections verbatim · SVS verification record · reopen conditions · exportable audit trail
Session depth
Standard for policy papers and advocacy materials · Deep for flagship annual reports and major commissioned research
Why Augle for think tanks

Credibility is the only
asset that matters.

Runs the hostile critic before you publish

The Contrarian is calibrated to surface the objection a well-resourced critic at a rival institution, a hostile journalist, or a hostile politician’s advisor will deploy. Unresolved objections appear verbatim with resolution conditions. The paper that goes out is the one that has already survived the worst-case scrutiny scenario.

Catches retracted and superseded sources automatically

The Guardian’s SVS checks every citation against retraction databases, assesses preprint and peer review status, and flags source independence concerns. A primary study with known replication failures, or a statistic drawn from a retracted paper, doesn’t survive SVS. It shouldn’t survive your review process either.

Grades every claim by the evidence that supports it

Every claim receives a confidence grade — Established, Probable, Contested, or Gap — based on the evidence base, not the advocacy position. The difference between Probable and Contested is the difference between a paper with appropriate caveats and one that makes claims it can’t defend. Knowing which is which before publication is the entire value.

Publish with confidence.
Not regret.

Join the waitlist and run a session on your next policy paper before it goes out.