If you applied this same level of scrutiny to the existing prison system, I'm sure you could come up with far worse outcomes using far less speculation. For example: What's to stop state prison guards from torturing inmates? Answer: Nothing, because they have qualified immunity.[1]
1. https://www.newsweek.com/prison-inmate-qualified-immunity-pr...
If you read more into the cash for kids scandal, you'll see it was the tip of the iceberg. The level of nepotism in Luzerne county was absurd and the state did nothing to fix it. There were judges abusing their positions, using staff for personal errands, showing blatant bias against certain attorneys, fixing cases, etc. By the time the FBI was done, around a half dozen judges in the county were convicted of various offenses.
My takeaway from the whole thing was that judges have too much power and too little oversight. They're the one relic of the royalty/nobility system that we kept after the American revolution.
Some of that will be corrupt.
It's not that I don't think incentives can play a role; I just wanted to emphasize how incentives, profit, and creativity can go awry.
That said, I think this is still pretty fraught.
Incentives are just rules enforced with a carrot instead of a stick; we're just as likely to come up with bad incentives as we are to develop ineffective regulations and write loophole-ridden or misapplied laws. At sufficient scale, money tends to inspire a troublesome kind of creativity.
If the private prisons have sufficient profit motive and the problems they face prove tractable for data and bespoke or ML solutions, they'll analyze likely convicts as thoroughly as possible and do what they can to skim the most-profitable convicts off the top of the pool and leave public prisons with whatever is left. Maybe they'll just directly pay the prisoner or their family. Maybe they'll have a miserable as-bad-as-regulations-allow experience for uninvited plebians and a special experience to offer prisoners they want.
Will a certain incentive lead them to bribe judges, prosecutors, defense lawyers, or witnesses to get (or avoid) prisoners? Would they do the same to keep a former inmate from being convicted again within 3 years of release?
And then some donation-funded corner of the think-tank ecosystem will endlessly torture the data to "prove" private prisons get better outcomes than public ones and shop boilerplate legislation to restrict public and promote private prisons.
And once the prisoners are within their walls, they'll be sucking up all of the data they can to hone their prediction of whether they'll reoffend when they get out. Will they use it nobly to target support, or to ensure the whoever the algorithms pick as bad apples end up in constant manufactured trouble?