On the Carnegie Foundation site, I read Lee Shulman's distinctions between thinking and acting:
During my decade of work on medical diagnosis, I studied gifted internists to understand how they made diagnostic judgments. A good friend, the Australian surgeon Ken Cox, came to me one day and said, "Lee, you're doing pioneering work on internists, trying to learn how their diagnoses lead to courses of action, but there's a big difference between internists and surgeons."
"Internists," he said, "make a diagnosis in order to act. Surgeons act in order to make a diagnosis."
Interesting, because the surgeons Shulman talks about are always working within a horizon of probabilities, and a repertoire of skills & acquired knowledge. All of this leads them to act, so in a sense understanding (and especially the understanding that you don't quite understand the situation you're in) always precedes or at least accompanies action.
The problem for folk designing learning environments is knowing when this might happen, and how to support students so that their 'horizon' knowledge can be enriched or extended. This is a problem for PBL designers, but it's even more of a problem for simulation designers. Get it wrong, and students just thrash around. Helping them to move from thrashing to threshing is the heart of it. And there are times when, as a designer/implementer, you can't wait for the results of complex tests: you need to act like a surgeon, not an internist, in order to create successul simulations.