Performance-Learning Distinction

The performance-learning distinction is the most important conceptual separation in Bjork's framework and the one most systematically violated by contemporary evaluation systems. Performance is observable behavior during training—accuracy, speed, output quality, visible competence under the conditions in which practice occurs. Learning is the change in long-term capability that supports retention after delays and transfer to novel contexts. The two can align, but under many conditions they diverge or invert: practices that maximize performance (massed study, blocked problems, immediate feedback, easy retrieval) often minimize learning, while practices that maximize learning (spaced study, interleaved problems, delayed feedback, effortful retrieval) often suppress performance during training. This inversion produces a systematic evaluation error: if you measure only performance, you are measuring the wrong thing when your goal is learning. AI tools are performance-maximizing engines that operate without regard for learning consequences, because learning cannot be measured in the moment and performance can.