It occurs to me that when your program results in a situation where recidivism is "very low", it might become much easier to distinguish those who fell upon bad circumstances from the ones with more deep-rooted mental issues-- like treatment resistant compulsive behavior, or perhaps even more importantly, psychopathic traits.

In my view, these are the real dynamos of the 'criminal world', where predatory traits are baked into their behavior patterns at a lower level. There are compelling opportunities for more research here (https://modlab.yale.edu/news/can-psychopaths-be-cured) and isolating variables is a good way to help move things forward.