It's also worth considering why a participant might "behave poorly". Even if you select for "good people", those in circumstances bad enough to need an ISA are

It's also worth considering why a participant might "behave poorly". Even if you select for "good people", those in circumstances bad enough to need an ISA are probably more likely to hit harsh times in the future and simply be unable to repay the loan - they suffer from generational poverty, for example, which means if a close loved one becomes injured or badly ill they may have no choice but to abandon their education and work multiple low wage jobs to support their family. At that point the ISA investment is wasted because they're unlikely to finish that education anytime soon or be able to convert it into a high-wage job. People in poverty situations also often have worse health, which means you are dealing with higher odds of the participant being unable to pay it back because they get sick, or die. I would not be shocked if it turns out that in practice ISAs are a money-losing proposition due to all the risks involved, even if they can be an effective way to make education accessible to people. Education being so ridiculously expensive in the US is still a relatively new thing.