Conditioned Description Is Not Independent Validation
Label a judgment that was given the measured answer as conditioned description rather than independent corroboration.
These examples and illustrative results are independently authored teaching materials, not measured model results.
Use case
A media measurement supplies teaching contrast 34, then a model receives 34 and describes it. Repeating 34 is conditioned explanation, not a second independent measurement.
Mechanism
Record the materials, metrics, target values and prompts seen. Label value-conditioned outputs for description/interpretation. Corroboration requires an independent method operating on the same stable source with defined units/algorithm and shared dependencies recorded. Report disagreement too, without counting quoted values as observations.
Bad example
Tell the model 34, then report two independent validations because it also says 34.
Good example
Label prose conditioned on supplied contrast 34 and retain prompt dependencies. Independently compute from stable media under a declared method, or record no corroboration if unrun. Explain source/shared dependencies rather than treating repetition as correctness proof.
Why the change matters
Inherited answers naturally agree without independently supporting them. Dependency records expose why, while preserving descriptions’ legitimate interpretive role.
Observable expectation
The teaching table separates measurement A=34 from B quoting A, counting one measurement. If independently calculated C=32, investigate units/algorithms instead of forcing 34. These are hypothetical dependency values.
Limits
Methods can share errors, so agreement does not guarantee truth. Define the actual contrast metric; teaching 34 is neither an accessibility ratio nor measured efficacy.