Principlev1PR-1092
Hold measurement protocol constant across before-and-after…
Hold measurement protocol constant across before-and-after conditions while varying only the system being optimized—changing both simultaneously eliminates your ability to attribute effects to causes.
Why This Is a Principle
This is a principle for experimental design derived from the axiom that learning occurs through prediction error (Learning occurs when outcomes differ from predictions…). To generate valid prediction error, you must isolate what changed. The axiom states how learning works; the principle prescribes how to structure measurement to enable that learning.