hidden variables

Often there are variables or constants deeply embedded in code where there are 'best guess' estimates, but where machine learning could easily be used to improve or personalise them. The challenge is to connect these hidden variables deep in the code with the interface behaviours they influence and human feedback that would enable training.