NYC Surveillance Tech on Shootings Gives False Alarms 87% of Time | Hacker News Reader

NYC Surveillance Tech on Shootings Gives False Alarms 87% of Time

bloomberg.com

25 pointsloteck6 comments

A high false positive rate doesn't mean much on its own. The real question is what is the point of detecting gunshots? If you are trying to decrease response time to real events, you want to be minimizing your false negatives, and increasing false positives to do so is likely a good strategy. Similarly if you want to use it in conjunction with other data collection, it's probably better to have false positives than false negatives. Of course a more accurate system, with both fewer false positives and false negatives, is always desirable, but the perfect is the enemy of the good enough.

The only reason false positives would be an issue is if you had some better method of identifying where resources ought to be allocated, and instead you were going after these false leads. If the cops had a better system they were being prevented from using, I doubt they'd be arguing to to keep this one in place, and the critics would probably mention it.

I'm not saying this system is worth whatever it costs, indeed the value prop of detecting gunshots seems rather low to me, but the particular argument made in this article is poor.

False positives are a problem if the response is itself potentially dangerous.

I suspect smoke alarms have even higher rates of error.

Smoke alarms in a domestic setting might, but not systems installed in large buildings.

Think of it as complimentary and/ or a confirmation, not automated or human-replacement.

A 911 call plus a matching shotspotter report has a pretty strong value. Much greater than the sum of the parts alone.

I had police at my door in under 2 mins from calling 911 because a shotspotter report had dispatch and the local officers already 'aware' and 'on-lookout' for a possible source minutes before I had assessed/secured things and called.

They were already nearby, had slightly diverted to circling the neighborhood & checking things out, and just needed a final street address plus confirmation of a need to escalate the response.

In other words... it worked perfectly.

But for reasons... my call ultimately ended up a 'false-positive' in the system, despite the system having worked correctly, both technically and systematically: the shots heard/detected were actually fired and it was a situation where police were ultimately called/summoned and arrived quicker than otherwise would have been possible without the headstart.

So beyond the assessing the 'value' of these systems, you also have to assess what exactly is getting measured, and what the precise meanings of false-positive and false-negatives might be in various use-cases.

This tech has proliferated across cities in the US by claiming to be a "force multiplier". That's supposed to mean it makes police more effective at their mission without actually adding any additional headcount.

But if 70-90% of the time the tech is sending police on goose chases that end with no findings, it seems like "force multiplier" falls into one of those marketing buckets where the truth is the exact opposite. The tech actually divides police from the mission.

Many, many cities are siphoning off public taxpayer dollars and sending them to this company.

Reply on news.ycombinator.com