Principlev1
Track agent outcome accuracy separately from execution
Track agent outcome accuracy separately from execution speed, classifying errors as systematic bias or random noise to determine appropriate correction mechanisms.
Why This Is a Principle
Derives from Illusion of Explanatory Depth (self-assessment unreliable), Systematic Overconfidence Taxonomy (overconfidence), and Domain-Specific Calibration Development (calibration domain-specific). Prescribes decomposition of error types to enable targeted accuracy improvement.