Forgive me if I am wrong but I was under the impression that what you described is how machine learning basically operates: reporting and trial and error to achieve better and more efficient results.
At least when analyzing actions. Is that not the case?