Research with a destination | A question to investigate
A stronger answer begins with a better question.
A medicines evidence summary can look complete while leaving the decisive uncertainty untouched. We want to investigate the difference between a polished document and a review that helps a professional reach a defensible next step.
The missing work inside a summary.
Imagine a medicines committee receiving a clear account of an intervention. The relevant studies are present, but their populations differ from the local service, a competing option has been omitted and the delivery assumptions are hidden. These are different problems from slow document preparation. Our proposed research separates coverage, interpretation and usefulness so that an improvement in presentation cannot quietly stand in for an improvement in reasoning.
Give quality something concrete to mean.
A first collaboration would select one recurring evidence task with an academic methodologist and a professional who uses its outputs. Before preparing any new summaries, we would agree an assessment rubric: relevance of the sources, fidelity to their findings, handling of conflicting results, visibility of assumptions and completeness of the comparison. Reviewers should be able to explain why a judgement changed, including when a source is accurate but unsuitable for the decision.
Compare against a credible existing approach.
The comparator should be the team's actual review process, with its usual tools and experienced staff. Artificially weakening that process would produce an impressive but unhelpful result. We propose matched evidence questions, a common source cut-off and assessors who do not know which preparation method produced each draft wherever feasible. Complexity, previous familiarity and access to information would be recorded rather than assumed equivalent across tasks.
Measure the whole review, including correction.
Time to produce a first draft is only the opening measurement. We would also record checking time, corrections, discussion and the effort needed to turn the review into an agreed decision paper. Error severity matters: a missing citation and a distorted conclusion should not receive identical weight. The proposed output is a quality-and-effort comparison, with examples of disagreements and a clear account of what the sample can support.
A public decision brief | An illustrative example
Open the question.
Inspect the reasoning.
Choose a business question, then look beneath its claim. See what a public source supports, what depends on the local setting and what a useful next step could be.
An academic or development team examining the next experiment after a published model comparison.
What does this result allow us to ask next?
A promising figure can sharpen a research question. Its value depends on keeping the model, observation and next hypothesis distinct.
A proposition to investigate
A clearer question can be a useful output.
An evidence brief could help a team distinguish a published observation from the experiment needed to investigate translation. The proposed value is a more precise comparison and an explicit evidence gap. It does not turn an external model result into a clinical claim or a NEUVIOR discovery.
This is a constructed research-planning question, not an agreed study or product-development result.
What the next brief would contain
A decision someone
can commission.
- The decision
- Whether a specific follow-on question has a credible experimental comparison.
- The comparator
- A scientifically justified model or condition relevant to the proposed hypothesis.
- The evidence
- Full methods, model relevance, measurements, replication and a feasible authorised protocol.
- Stop or redesign
- Stop or redesign if the proposed model cannot answer the question or the required work is not feasible.
Constructed public example | Not PHARMORIS-generated software output or a measured customer result.
Keep this example briefMake difficult cases part of the design.
An evaluation built only around straightforward questions could miss the situations where support is most useful or most risky. We would deliberately include contradictory evidence, thin evidence and questions where the right response is to defer. Agreement between reviewers would be examined alongside average scores. If experienced assessors disagree about the rubric itself, the first research result may be a better definition of the task rather than a claim for the technology.
A partnership with room for an inconvenient answer.
Academic collaborators could contribute study design and independent interpretation; professional teams could establish realistic tasks and consequences. NEUVIOR would contribute the proposed preparation approach and document its boundaries. UKRI's research-integrity guidance informs the conduct we would expect, while the Magenta Book provides evaluation context (UKRI, n.d.; HM Treasury, n.d.). Publication rights, source licences, conflicts of interest and access to assessable materials would need agreement before a study begins.
The decision this study should unlock.
The bounded next step is a jointly scoped protocol using a small, authorised set of completed or simulated organisational questions. It should end with a recommendation to refine, expand or stop the proposed approach. Progress would require acceptable review quality and a meaningful reduction in total burden, using thresholds agreed before results are seen. We would stop or redesign if serious distortions persist, correction cancels the benefit or the comparison cannot be made fairly.
A practical decision paper could display the competing options, the evidence that favours each and the assumptions most likely to reverse the recommendation. Asking reviewers to use that paper in a structured discussion would test whether the preparation supports an actual choice, rather than merely improves a document's appearance or makes its prose easier to read.
Sources and context
UKRI (n.d.) Research integrity. Available at: https://www.ukri.org/manage-your-award/good-research-resource-hub/research-integrity/ (Accessed: 14 October 2026).
HM Treasury (n.d.) The Magenta Book. Available at: https://www.gov.uk/government/publications/the-magenta-book (Accessed: 14 October 2026).
Another useful angle.
- Reproducible medicines analysis
Explore reproducible medicines analysis that connects source data, calculation and interpretation, giving research partners a result they can check and update.
- AI and professional judgement
Examine how professionals check, challenge and use AI outputs in medicines work, with proposed studies of appropriate reliance, review quality and total effort.
- Academic and commercial translation
Connect an academic finding with a credible user, comparator and next experiment, while keeping research rights and commercial translation questions clear.
