The thing that Flock does that's alarming is that it provides operators with a search engine for arbitrary vehicle descriptions which include but are not limited to license plates, with history stretching back; misuse scenarios are obvious, the search histories allow you to track the movements of specific people with fine granularity.

The thing that Flock does that is actually immediately problematic is that it operationalizes BOLO/hotlist databases that weren't intended to be used in real-time. Our deployment of Flock curbed more innocent vehicles than actual stolen cars, because Illinois LEADS isn't reliably updated, and so pings on vehicles that were reported stolen (whether or not they actually turned out to have been stolen as opposed to borrowed by a family member or something) weeks ago and recovered.

▲

conductr 2 hours ago | parent | next [-]

My car was reported stolen mistakenly, long story but was cleared up within a few hours (an officer came out to confirm vehicle was in my possession). Then a few days later Flock identified my vehicle driving and notified the cops. It was me driving my normal commute and I was pulled over at gunpoint. When I finally explained the story, they were like “oh yeah, we see that in the system but sometimes there’s a lag between databases.” Really? Wtf guys

▲

MichaelZuo 2 hours ago | parent | prev [-]

That sounds like a much more foundational and serious issue than any number of ALPR systems?

If Illionis LEADS lacks credibility?

▲

tptacek 2 hours ago | parent [-]

It has plenty of credibility at the task for which it was designed. It wasn't designed as a backend event system for real-time sensors.

▲

MichaelZuo 2 hours ago | parent [-]

Huh?

You just said it “isn't reliably updated, and so pings on vehicles that were reported stolen…” are weeks out of date?

Even if all ALPRs vanished from the Earth tomorrow, that still indicates a lack of credibility in the pings?

▲

BobaFloutist 2 hours ago | parent [-]

If I had to guess, it's like when an expensive medical test gets more accessible and they have to update their model.

If the cost to get an MRI means everyone that gets one has a combination of symptoms and risk factors raising the pre-test probability, then it makes sense to treat MRI findings aggressively. If they become cheaper and start using them as screenings, they need to update their approach.

Similarly, if license plates are scanned when cops are already pulling someone over for moving violations (or the car is accumulating a ton of parking tickets, having been dumped), it might be ok if their status isn't updated that frequently, and it still might make sense for cops to approach the car with the idea that it might be stolen (something a drivers license check against registration can quickly clear up, which shouldn't matter too much if they were getting pulled over anyway).

If the system is being used to justify pulling people over in the first place, it needs different parameters.

	▲	MichaelZuo an hour ago \| parent [-]
		Clearly in the real world Illinois, there are multiple systems that are “being used to justify pulling people over in the first place”. But how does that relate to the credibility of pings?