Designing an interface, writing a compelling introductory sentence, and making a diagnosis—many types of Knowledge work depend on cycles of generating possibilities, culling them, and repeating based on observations or feedback.
Anecdotally, my experiences in design and engineering suggest that experts are better at generating possibilities in the right “regions,” at rapidly culling possibilities which are unlikely to work, and at updating their search based on observations/feedback. One possible explanation: Expertise requires building sophisticated chunk recoding schemes