Interpret evidence and uncertainty

Select one claim that matters

Start with an accessible saved assessment or a published case. This walkthrough uses Assess; sign-in and record access are required there. Public catalogue sources can be inspected in Explore without private assessment access

Pick a claim that could change your decision: a stated benefit, performance level, safeguard or readiness assumption. Bring the underlying source when available, and be prepared to distinguish missing information from contradictory information

Open an assessment to inspect

Trace the claim in six steps

1. Define the claim precisely

   Write what is claimed, for whom, in what setting and under which conditions. “Enables communication” leaves important questions open about reliability, independence, support and daily use. The illustrative communication-at-home question does not establish any of those outcomes

2. Identify the kind of input

   Distinguish a documented fact, a self-reported claim, a model input, an assumption, a requirement and a missing value. A recorded or simulated data source is a feature of the study description; entering it does not independently verify the study

3. Read the source in context

   Check whether the source concerns the same system, population, setting and outcome. Look at dates and the difference between planned work and reported results. Evidence supporting a narrow test should not silently become a claim about general use

4. Inspect the reasoning around the result

   Where the model applies, read Decision Surface drivers, uncertainty and evidence gaps together. Numerical summaries depend on the entered information and the model. An unavailable numerical output for a dataset study is a scope boundary, not a negative finding

   

5. Ask which assumption could change the decision

   Use supported stress-test controls where relevant, and record the assumption varied and the result you observed. Treat a scenario as conditional reasoning; it does not establish the probability of that scenario. The exploratory Impact map likewise leaves likelihood unestimated

6. Form an evidence request or condition

   State exactly what would resolve the uncertainty: a source, test, population, outcome, safeguard or review. Identify what remains unresolved even if the requested evidence arrives. Use the result to improve the assessment or your own review record

Separate support from certainty

A useful interpretation explains which claim is supported, at what scope, by which source and with what remaining uncertainty. Avoid turning a range into a guarantee or the absence of a documented harm into evidence that harm cannot occur

Use the method for the wider reasoning framework, or continue to Prepare a Passport to carry the evidence and qualifications into a discussion

About this guide edition

Reviewed 2 October 2026 against the current interface and maintained product guide. Screenshots show either the public GESDA catalogue or a local demonstration using fictional Aurora data, as identified in each caption

Local captures illustrate the interface, not sign-in, server persistence or successful export. Use Print / save PDF for a reading copy and follow the live application for current availability