Start with the proposition
Before evaluating an artifact, state what needs to be established. “A request was received”, “the requested action completed” and “the result was correct” are different propositions. Evidence sufficient for one does not automatically support the others.
The Conceptual Core calls this the Target Proposition. Its subject, instance, time and context matter. A finding about one transaction is not a finding about every transaction in a deployment.
Connect materials and arguments
Witness Materials are inputs to an evaluation. An evidence-and-argument graph connects those materials to the proposition, making the relevant observations, bindings, assumptions, warrants and dependencies available for examination.
A well-formed graph is not itself a finding that the proposition is true. A reader also needs to know what each material represents and whether the argument connecting it to the proposition holds under the stated conditions.
Evaluate under an explicit profile
An Evaluation Profile specifies the context in which support is assessed. Different purposes can require different evidence. A diagnostic investigation and a contested financial decision may rely on the same logs while asking different questions of them.
A Support Assessment records the result of that evaluation. A reliance decision then considers what to do with the result, alongside authority, risk and policy. Technical verification, support assessment and permission to act remain distinguishable.
Two published entry points
Witnessability Conceptual Core 1.0 defines the stable conceptual vocabulary and consistency conditions at the model level. This website provides a readable rendering alongside the original release bytes and verification materials.
Witnessability Model 1.1 is a separate published research document. Use its version-specific DOI when citing it. The version number of the Model does not change the version of the Conceptual Core.
Read the scope before the label
This page is an introductory map, not a replacement for either publication. It intentionally avoids presenting one scalar label as an answer to every evidentiary question. Consult the exact source and its assumptions when using a model result or defining a profile.