Black-Box Conversation

A black-box system may expose enough evidence to evaluate some claims while providing little evidence about internal execution.

For example, an evaluator may observe:

• the input provided to the system
• the output returned by the system
• limited request metadata

For a claim that can be resolved entirely from the externally visible output, additional internal evidence may provide no additional claim-material capability.

This illustrates an important property of Structural Assurability:

More instrumentation does not imply greater assurability for every claim.

Example Result

In the formal output-property example, the capability model treats externally returned output as the only claim-material evidence.

The opaque service, instrumented agent, and assurability-oriented architecture therefore enable the same modeled output-classification capability for that assurance context.

The example establishes equality of capability profiles. It does not prove that the observed output resolves a nontrivial output-property claim.

text S_O ~q S_A ~q S_R

This is claim-relative equivalence. It does not imply that the architectures are equivalent for other claims.

Source of Truth

The Lean source is authoritative:

• Case2_OutputProperty.lean
• Architectures.lean
• Equivalence.lean