The first try at a question is either

• A certain answer:
  • if answered correctly then that's that.
  • if answered incorrectly then a great deal can be learned quickly
• A blind guess:
  • if answered correctly, then this is bad because we don't really know why.
  • if answered incorrectly then this is not so bad. TODO: Why.
• An educated guess:
  • if answered correctly, then this may be confirming some false assumptions. A lucky guess contains lots of uncertainty, but a green check serves to imply those uncertainties don't exist.
  • if answered incorrectly, then this is good because we have a starting point and can branch out to explore why. Some system of partial credit would be useful. partial credit can be had by breaking the question down into smaller questions or more incisive analysis by the backend. If the backend can pick apart expressions and relate the inacuracies back to a structured model then some
• The result of many hours of study
  • if answered correctly, then this is excellent.
  • if answered incorrectly, then this is disheartening.

is there room for explicit learning control flow, can teachers use guided adaptive programs to determine the order and depth required for best learning outcomes? There is an adaptive lisp course that does this, but only because the material is machine parsable with unambiguous semantics. How can courses with less structured material, like poetry be adaptive? What is the spectrum of adaptivity and how can machines understand users well enough to guide them?