Verified against ChatGPT · 2026-08-14
Synthesize research findings that disagree with each other without quietly picking a favorite
Produces a research synthesis across multiple findings or reports that keeps genuine disagreement visible and traces it to a specific cause, rather than resolving conflicting evidence into one clean narrative.
The prompt
Ready to copy — highlighted parts are example details you can swap.
Synthesize the findings below into a single coherent brief. Some of these findings disagree with each other, and the synthesis needs to represent that honestly rather than resolving the disagreement into one clean story. FINDINGS TO SYNTHESIZE Three internal analyses on whether a pricing change increased churn — one team's cohort analysis says yes, another team's survey data says users didn't notice a difference, a third says it's too early to tell. WHAT THIS SYNTHESIS NEEDS TO SUPPORT Deciding whether to roll back the pricing change company-wide next quarter. KNOWN DIFFERENCES BETWEEN THE SOURCES The cohort analysis only covers customers who joined after the price change; the survey included long-tenured customers who may be less price-sensitive. Group the findings by the specific claim they address, not by which report they came from. Where two findings addressing the same claim agree, state that plainly and move on without over-elaborating. Where they disagree, do three things: state the disagreement explicitly, check whether The cohort analysis only covers customers who joined after the price change; the survey included long-tenured customers who may be less price-sensitive. explains it (different methodology, different population, different time period), and if it doesn't cleanly explain it, say so rather than inventing a resolution. Never write a summary sentence that blends two disagreeing findings into a single averaged claim neither source actually made — "studies suggest a moderate effect" is not an acceptable synthesis of one study finding a large effect and another finding none. State which findings are most load-bearing for Deciding whether to roll back the pricing change company-wide next quarter. and which are tangential, since not every finding in the source list carries equal weight for the actual purpose of this synthesis. WHAT NOT TO DO Do not present the synthesis as more unified or conclusive than the underlying findings actually are — a genuinely split evidence base should read as split, not smoothed into false consensus for the sake of a cleaner-sounding brief. OUTPUT FORMAT 1. Findings grouped by claim, each marked as convergent or conflicting. 2. For each conflict, the explanation checked and whether it resolved the disagreement. 3. A short "bottom line for Deciding whether to roll back the pricing change company-wide next quarter." section that states plainly what's well-supported, what's contested, and what remains genuinely unknown.
Customize
Optional — swap in your own details for the highlighted parts above.
Why this works
Asked to produce one coherent brief from multiple findings, a model's cleanest path is to write a unifying narrative sentence that sits comfortably between disagreeing results — "studies suggest a moderate effect" when the actual inputs were one strong effect and one null result — because that sentence is fluent and sounds authoritative even though no individual source actually supports it; it's synthetic in the worst sense, a claim manufactured by the act of averaging rather than found in any evidence. Grouping by claim rather than by source is what makes disagreement visible in the first place, since findings about the identical question end up adjacent instead of scattered across separate per-source summaries where a reader has to notice the contradiction themselves. Checking each disagreement against named source differences gives the model a legitimate, evidence-grounded way to explain a conflict without picking a winner arbitrarily — and explicitly allowing "this doesn't cleanly explain it" as a valid answer matters because not every disagreement has a tidy methodological explanation, and forcing one where none genuinely exists just relocates the false-resolution problem one level down. The final "bottom line" structure that separates well-supported, contested, and genuinely unknown directly serves the stated decision purpose: a decision-maker reading this needs to know not just what the evidence says but how confident to be in each piece of it, and collapsing that distinction into one smooth paragraph would strip out exactly the information needed to weigh a split evidence base responsibly.
What you get back
Claim: the pricing change increased churn. Conflicting — cohort analysis shows a measurable churn increase among new customers; survey data shows long-tenured customers reporting no noticed change. Explanation checked: source difference (new vs. long-tenured customers) plausibly explains part of the gap, since new customers have less switching-cost inertia, but doesn't fully resolve it since the survey may simply be less sensitive than behavioral cohort data. Bottom line: churn impact on new customers looks reasonably well-supported; impact on the existing base remains genuinely unclear and needs more data before a company-wide rollback decision.
Verified against
ChatGPT GPT-5.1 · 2026-08-14
Changelog
- 2026-08-14 — Initial publish, verified against ChatGPT GPT-5.1.
Need this built into your business?
If a prompt isn't enough — what Scult builds, built and maintained for you — that's Scult's day job.
EXPLORE WHAT SCULT BUILDS
