For non-experts judging unfamiliar scientific topics, AI debate increased accuracy over prior beliefs, unlike single-expert consultancy.
Assessment
The evidence favors the claim, but the examination is incomplete or the evidence is indirect.
The claim comes from a single preregistered-style experiment (Carro et al., arXiv 2610.28539, October 2026, accepted at NeurIPS 2026) in which Prolific participants judged AI debates or AI consultancy on subtopics of ten scientific and technological discoveries. Participants' prior beliefs were elicited within the same procedure, and the design aimed for about half to start with the correct general belief and half with the incorrect one.
In the in-subtopic debate condition, accuracy on the specific question rose from 41.3% before the debate to 57.6% after it (McNemar's test, Holm-adjusted p = .028). In the consultancy condition, where a single AI argued for one answer, accuracy moved from 51.25% to 48.75% (p = 1.0). The reported figures fit roughly 92 debate judges and 80 consultancy judges. The study did collect what the claim describes, and the within-condition result survives the authors' multiple-comparison correction.
The evidence is weaker than the headline suggests, for three reasons. First, debate judges started below chance on a two-option question, and 57.6% after the debate is not clearly above chance at this sample size: a rough binomial check gives p of about 0.15. Some of the gain may be movement from a below-chance start rather than strong truth-tracking. Second, the contrast with consultancy rests on one condition being significant and the other not. The paper reports no significant direct test that debate beat consultancy, and the two conditions started from different baselines (41% and 51%). Third, a companion cross-subtopic debate condition improved only numerically (50.6% to 58.4%, not significant).
Earlier human-judge work points the same way under conditions of information asymmetry. Michael et al. (2023) and Khan et al. (2024) found debate beat consultancy for judges who could not see a hidden source text. Parrish et al. (2022) found that short debates did not help. So the finding is plausible and consistent with that literature, but it rests on one modest-sized study, and no direct between-protocol test has been reported. A replication with a larger sample and a direct debate-versus-consultancy comparison would settle it.
Full reasoning: the evidence and decisions behind this verdict
Read Sections 1, 4.1 and 4.2 of arXiv 2610.28539 (arxiv.org/html/2610.28539). Design: 5 x 2 between-subjects (protocol x grounded/erroneous prior). An a priori power analysis targeted N=200 (20 per cell). Total human N=507 across 5 experiments. Analysis: paired McNemar and Wilcoxon tests, with Holm-Bonferroni correction within and across protocols.
Results: in-subtopic debate 41.3% to 57.61%, p=.028. Cross-subtopic 50.56% to 58.43%, binomial vs chance p=1.0, bootstrap CI [49%, 67%]. Consultancy 51.25% to 48.75%, p=1.0. Group sizes inferred from the percentages: 38/92 to 53/92 for debate, and 41/80 to 39/80 for consultancy. These were not stated in the text read. Binomial check of 53/92 against 0.5: z ≈ 1.46, two-sided p ≈ 0.14.
Cross-protocol Fisher tests are reported only for debate versus debate (Δ = −0.8pp). The text read gives no significant debate-versus-consultancy contrast. The post-hoc gap of about 9pp at n≈92 vs 80 would probably not be significant.
Below-chance starting accuracy, despite the intended balance of grounded and erroneous priors, leaves room for regression toward chance. Weighted-voting gains (45.8% to 65.3%) are reanalyses of the same debate data, not independent replications.
Prior literature, from memory and the paper's related-work section: Parrish 2022 found null results. Michael 2023 and Khan 2024 found debate beat consultancy under hidden-information asymmetry.
Status is supported rather than verified because the result is a single study and the contrast between protocols is not directly tested. Credence 0.62 that the claim holds as a general finding about non-expert judges.
Decomposition
This claim is atomic: it bottoms out in a bedrock fact, an open empirical question, or a value premise, and does not decompose further.
Provenance
Where this claim has been said, linked to its canonical form.
Overall, we find that debate increases accuracy relative to prior beliefs for human judges (from 41.3% to 57.6%, p=.028), whereas the consultancy baseline, in which a single expert argues for one answer, does not.
Abstract; non-expert Prolific participants judging AI debates on unfamiliar scientific topics.
The assertion outruns the source's own evidence. The within-condition debate gain survives Holm correction. However, judges began below chance, and the paper reports no significant direct comparison of debate with consultancy. The phrase "whereas consultancy does not" rests on one condition being significant and the other not. Worth reading closely: Appendix C.3 and the public data would show the exact per-condition sample sizes and whether a direct test of debate against consultancy was run. The quoted passage was not found in the stored copy of this source.
Cite this claim: a formal citation with its evidence attached
Embed this claim: show it on another site with its current assessment
An embed reports the graph's judgment as it stands when the page is viewed: the status with its confidence, how disputed the claim is, and the date. It links back here.
<blockquote class="minerval-claim" data-claim-id="8f1936b6-a556-492b-8711-f64bbed3f5b4"><a href="https://minerval.ai/claims/8f1936b6-a556-492b-8711-f64bbed3f5b4">For non-experts judging unfamiliar scientific topics, AI debate increased accuracy over prior beliefs, unlike single-expert consultancy.</a></blockquote> <script async src="https://minerval.ai/embed.js"></script>
<iframe src="https://minerval.ai/embed/claims/8f1936b6-a556-492b-8711-f64bbed3f5b4" title="Minerval: For non-experts judging unfamiliar scientific topics, AI debate increased accuracy over p…" width="480" height="225" style="border:0;max-width:100%" loading="lazy"></iframe>
[](https://minerval.ai/claims/8f1936b6-a556-492b-8711-f64bbed3f5b4)
The badge now: 
Contribute
Every judgment on this page is open to challenge. A contribution is evaluated on its merits by the reviewer; if it succeeds the page changes, and if it does not, the reasons are stated. Either way the exchange becomes part of the claim’s public record.
The attention this claim received was paid for by a funded mandate. Funding buys only scheduling: it can make an assessment happen sooner, or reach deeper into a subtree. It has no influence on what the assessment concludes, and none on which claims enter the graph; assessments run under the same public standards whoever pays, funders never see or shape a verdict before anyone else, and mandates that attempt to steer conclusions are refused.
Created by extractor · Nov 10, 2026. Every judgment on this page is accompanied by a reasoning trace.