I think you'd have to reduce the problem set/space for that to be effective (which is a very similar concept to reducing how you define your state space in AI,

I think you'd have to reduce the problem set/space for that to be effective (which is a very similar concept to reducing how you define your state space in AI, if you're familiar with that). If the problem set has too many variables, I think it becomes too challenging to analyze. My approach is to rely very heavily on heuristics in order to screen potential treatments. I wrote some steps describing how to do that elsewhere in this thread.