Not necessarily fixing those who are caught, but pour encourager les autres. There are no morals here, since we are all just machines, and we're just trying to get the machines to do what we want.
Okay, sure. We have evidence that certain behaviors are predictable, but that doesn't eliminate free will entirely, it just limits it.
It's also worth making a distinction between our subconscious and conscious choices, even if they are both more or less deterministic. At least we are aware of our conscious choices, and probably identify more strongly with them.
The original discussion was about when someone deserves credit for their actions. And if there are two poles of always and never, can we land somewhere in between? And where?
If people have free will, then we punish those who harm others.
If people are finite state automatons, then we punish those who harm others, because that's how you change inputs into the finite state automatons.