I’m less keen on prosecuting the police here; they are, in this case, just cogs taught to follow small parts of a very complex multi-person process for arresting/booking/jailing someone—a process where no single person has the sort of full top-down view that would allow them to realize that their output is causing a bad overall result.

I’d be more keen here in prosecuting the people who devised the flowchart that the police are following, for not engineering it to be robust against this failure mode. We shouldn’t expect humans to do an extra non-obvious step (checking the criminal’s file against the fingerprints they took) if the flowchart doesn’t require it, especially under time-pressure to “get the job done”; instead, we should make the flowchart require it.

When you’ve devised a system such that the human “components” of your system don’t have the necessary information from their vantage-point to correct the behaviour of the system, the system itself must be designed to be self-correcting. Because, at that point, that’s the only kind of correction it can have.

This is how we do modern medicine: a given doctor or nurse or specialist has no idea who you are or what your story is, outside of the small window they are told to see you in. They’re also just cogs in their machine. But the process itself — run mostly through medical charts — drives patients toward the treatment they need nevertheless; and there are many checks and fallback cases built into both the trained human processes, and the automated processes, to get patients who fall through one crack back on track toward treatment/good health outcomes.

—————

Also, a tangent re: prosecuting those responsible for engineering the process:

Back in Ancient Greece, the Lottery system of electing a leader had one important feature not oft-mentioned in modern discourse, but which was perhaps the keystone to understanding the Lottery system as a whole: if the polity didn’t like a leader’s policies, they’d drag them out and lynch them (or whatever the Ancient Greek equivalent of lynching was.) There was no concept of “impeachment”; it was just a question of surviving your term, by doing things that don’t anger your polity too much. (And as such, probably the whole reason for electing a leader by sortition in the first place, would have been that nobody wants to go into politics if you’re only one misstep away from a lynching at all times. So you have to force people to do the job; and if you’re forcing people to do a job that’s net-negative for them, you should at least be fair in your selection process by making it random.)

In modern times, capital-E Engineers already have to sign off on the designs that pass through their hands; and they can then be held criminally liable if those things fail for reasons they should have been able to foresee. This rarely happens, though, because Engineers, out of equal parts conscientiousness and self-interest, have built up a large body of best-practices, of meta rules and guidelines to follow when defining domain rules and guidelines, that either avoid failures, or at least outright reject “impossible to make sound” designs.

What I’m saying is: legislators and (especially) regulators really need to hold the burden of ultimate responsibility — criminal liability, even — when a defined-by-regulation process fails. They should be “signing off on” processes and workflows they define, the same way a civil Engineer would sign off on a bridge design. We probably shouldn’t be lynching them, but we should at least be taking all those civil suits over wrongful imprisonment, and targeting them directly at these policy-makers. Being a policy-maker should be a duty, not a privilege: something with perhaps more down-side than up-side.