It seems to me that you are being unfairly downvoted. At some point there has to be a balance between punishing provably false reports, punishing (or not) good-faith false reports, and not discouraging true reports. This kind of balance is something that folks in AI need to take into consideration when deciding model performance (ROC curves), but it gets a lot more complex in real life because now you need resources to distinguish between false-reports, good-faith-false-reports, true-reports etc which easily veer into territories of determining intent and downstream harm.

If those resources do get allocated, there's a good chance these same folks downvoting you would complain about wasted government resources because the nuance in allocating those resources doesn't shine through in obvious ways.

What's the solution to this? I don't know, unfortunately. AI scientists have to take every problem that involves this nuance on a per-case basis and work with their business partners (or whoever the "client" is) to decide what's a good point on the curve for the operational model. Similarly social scientists (and politicians) influencing government operations presumably have to determine where that balance lies. And likely the downvoters don't have those answers either. Maybe the real solution is to have a not-too-complex-or-lengthy redress mechanism for those caught unfairly in the "system", with the ease-of-redress driven by how far to one side or other the balance tips.