Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

ShotSpotter: listening in on the neighborhood

computer.rip

Post reply on HN
Print view View original

261–270 of 443 posts

Re: ShotSpotter: listening in on the neighborhood

#261

Post by sbierwagen »

9935c101ab17a66 wrote: post #246
sbierwagen wrote: post #224

Earlier quoted context omitted.

He was setting off fireworks on January 25? Is there a holiday on that day I'm not aware of?

Can I ask why you’re concerned about the legitimacy of the kids reasoning for setting off fireworks? Like, if he was shot and killed on a holiday, it’s not his fault, but it’s perfectly reasonable to be shot and killed by a police officer when a opaque surveillance system reports gunshots on non-holidays?

Cops are Bayesian reasoners.

If you set off fireworks inside the White House, right now, would you be surprised to suddenly meet the acquaintance of many armed men? Fireworks don't often go off in the Oval Office. Loud banging noises are therefore assumed to be gunfire, not fireworks.

If you set off fireworks on the South Pole on July 4th, would you meet police? No, and for two reasons: July 4th is a well known day for setting off fireworks, and the South Pole is far away from basically everything.

If you set off fireworks shortly before midnight, on a random Thursday in January, in a city that's famous for having large numbers of murders, should policemen assume it's fireworks, or gunshots?

Re: ShotSpotter: listening in on the neighborhood

#262

Post by nonrandomstring »

jcrawfordor wrote: post #125
jcrawfordor wrote: post #71

Hello all, occasionally I write what I consider "Albuquerque content" and I do not expect it to become broadly popular. This article is something I put together very quickly and it probably assumes a certain degree of familiarity with the political context around policing in Albuquerque (which either side will tell you is very contentious) and, more broadly, policing and civil justice. Even without the currently evol…

Let me follow up, and maybe I should write this up on CAB but I don't really want to get too much into police reform there, what I know factually about the ShotSpotter system. Most of this comes from discussions with APD leadership and officers in the context of the police oversight role I used to hold, and they are very much limited in what they can say. Some of this is specific to APD SOP and other police departmen…

Previously done a lot of research into firearm and explosive acoustics (and the DSP to locate and categorise) and can say this is a _hard_ problem. Military versions for use in wide open spaces have a much better chance than in urban areas.

Less than five percent seems unlikely, but I wouldn't be surprised to hear "less than half the time".

Re: ShotSpotter: listening in on the neighborhood

#263

Post by qingcharles »

delichon wrote: post #7

> It's not perfect, but the distance from your house to a ShotSpotter sensor correlates fairly well with your household income. The wealthier you are, the less surveilled you are. To give ABQ police the benefit of the doubt, that pattern could also be compatible with more gun crime equaling more surveillance. It would be nice to have enough gun crime and sensor location data to see how true that is. When the sensors…

There is a ShotSpotter on the lightpole outside my door. My household income is about $2400/yr. My hood is a warzone though (South Side Chicago), so not unexpected lol

Re: ShotSpotter: listening in on the neighborhood

#264

Post by Tabular-Iceberg »

scheme271 wrote: post #219
Aurornis wrote: post #197

Earlier quoted context omitted.

> The more you monitor an area, the more samples you get, biasing the stats So by monitoring an area with high crime rate, you catch more crime, which results in more policing for the area with a confirmed high crime rate? And that's a bad thing?

It's biasing the sample. For example, drug usage among teens are roughly the same between white, hispanic and african americans but african americans are more likely to be arrested and charged for drug possession or usage. Here the difference is probably because african americans are stopped and searched more even though their usage is about the same as whites.

It’s probably a cultural factor. Most criminal enterprises want to be discreet in order to avoid detection by law enforcement and rival enterprises. Whereas the black criminal community has made an entire industry of advertising their criminality through music and fashion.

Re: ShotSpotter: listening in on the neighborhood

#265

Post by nceqs3 »

throwaway9917 wrote: post #248
jcrawfordor wrote: post #71

Hello all, occasionally I write what I consider "Albuquerque content" and I do not expect it to become broadly popular. This article is something I put together very quickly and it probably assumes a certain degree of familiarity with the political context around policing in Albuquerque (which either side will tell you is very contentious) and, more broadly, policing and civil justice. Even without the currently evol…

The tone of your article makes me honestly, really angry. You know damn well that the reason the sensors are in those neighborhoods are because that's where people are getting shot. You even talk about how a school where some little kid got shot has a sensor, as if it's some sort of punishment for the lower income people there. Perhaps it's because the police and the city government want to deter or solve murders tha…

I wholeheartedly agree.

Re: ShotSpotter: listening in on the neighborhood

#266

Post by BHSPitMonkey »

delichon wrote: post #7

> It's not perfect, but the distance from your house to a ShotSpotter sensor correlates fairly well with your household income. The wealthier you are, the less surveilled you are. To give ABQ police the benefit of the doubt, that pattern could also be compatible with more gun crime equaling more surveillance. It would be nice to have enough gun crime and sensor location data to see how true that is. When the sensors…

> To give ABQ police the benefit of the doubt, that pattern could also be compatible with more gun crime equaling more surveillance. It would be nice to have enough gun crime and sensor location data to see how true that is.

The very point of the article is that, absent leaks like this one, there could be no way for anybody to independently study or verify the fairness of the sensor distribution, or even the real efficacy of the reports the system produces—which is a troubling situation to be in when the state has an outsized amount of power to prosecute people based on potential junk science that will be hard for defendants to challenge in court.

The answer isn't to give police departments the benefit of the doubt (which they so rarely earn), but to demand better transparency and citizen oversight of the technology poised to be used against us.

Re: ShotSpotter: listening in on the neighborhood

#267

Post by WalterBright »

jakelazaroff wrote: post #249
WalterBright wrote: post #247

Earlier quoted context omitted.

Some crimes, like shoplifting, are hard to measure because it often goes unreported because the defunded police don't respond to them. But the murder rate is easy to measure. When people put bars on their windows on the ground floor, it is not the police causing them to be willing to spend the money on that. Crime can also cause poverty. For example, if the family breadwinner goes to jail, the family slips into pover…

Is the murder rate easy to measure? As best we can tell, “stand your ground” laws in states like Florida result in hundreds of deaths each year. How many of those would be considered murder in other states? Police killed over 1,300 people last year. How many of those people truly posed an immediate danger to others, and how many were murdered by a trigger-happy cop who was not held accountable by his colleagues? Crim…

> Is the murder rate easy to measure?

Yes. The medical examiner decides if a death is homicide or not.

> result in hundreds of deaths each year

Lacks context - the number of homicides in the same year.

> How many of those would be considered murder in other states?

They're still considered homicide.

> Police killed over 1,300 people last year

That figure is for all deaths where police were involved. A subset of that would be the police killing. Furthermore, if someone points a gun at a policeman, and the policeman kills him in response, that is self-defense, not murder. There are officers killed while on duty, too:

https://en.wikipedia.org/wiki/List_of_law_enforcement_office...

Less than 10% of homicides result from police action.

> how many were murdered by a trigger-happy cop who was not held accountable by his colleagues?

In Washington state, each death from a police encounter are investigated by law, and charges get filed if the officer broke the law. That would include being trigger-happy.

> It’s not objective

It's objective enough. My larger point is there aren't a lot of (or even any) homicides that go undetected in wealthy communities. Furthermore, your figures lack context as you didn't compare with the total amount of homicides. Your figures are not enough to claim that the higher homicide rates in poor communities are the result of police murders.

Re: ShotSpotter: listening in on the neighborhood

#268

Post by hcfman »

nonrandomstring wrote: post #262
jcrawfordor wrote: post #125

Earlier quoted context omitted.

Let me follow up, and maybe I should write this up on CAB but I don't really want to get too much into police reform there, what I know factually about the ShotSpotter system. Most of this comes from discussions with APD leadership and officers in the context of the police oversight role I used to hold, and they are very much limited in what they can say. Some of this is specific to APD SOP and other police departmen…

Previously done a lot of research into firearm and explosive acoustics (and the DSP to locate and categorise) and can say this is a _hard_ problem. Military versions for use in wide open spaces have a much better chance than in urban areas. Less than five percent seems unlikely, but I wouldn't be surprised to hear "less than half the time".

Categorisation will certainly be error prone as there is so much similarity because fireworks and gunshots. The main difference I expect is the speed of the shockwave resulting in steeper spectrograms for gunshots. However, a steep spectrogram is essentially showing a large amount of high as well as low frequencies. The higher frequencies degrade very quickly over distance reducing the difference of a gunshot from a firework the more this degrades.

Not finding evidence is a different problem, it could be as simple as using a revolver rather than a semi-auto that ejects the cartridges.

In any case, it would be great to see independent testing of this problem by the sort of people in this forum and you can use the software I developed last year (https://github.com/hcfman/sbts-aru) to do so. That software sets the time on the Raspberry Pi to have less than 1 microsecond of error, which is more than enough for any validation efforts.

Re: ShotSpotter: listening in on the neighborhood

#269

Post by hcfman »

reaperman wrote: post #119
sixothree wrote: post #60

Earlier quoted context omitted.

This is entirely grotesque. But the systematic deception is especially appalling.

ShotSpotter has (used to?) manually changed the location of detected gunfire to placate police who needed the evidence to point to their primary suspect. Then they could go to court and say “this system showed the gunshots came from the suspect’s location”. But “the system” didn't show that, until ShotSpotter employees manually changed the data. They spent a lot of effort hiding that as well.

Which is why we need public counter validation. (https://github.com/hcfman/sbts-aru) :-)

Re: ShotSpotter: listening in on the neighborhood

#270

Post by hcfman »

photochemsyn wrote: post #131
Aurornis wrote: post #65

Earlier quoted context omitted.

> This system detects door slams and popped volleyballs just as well as it detects gun fire (which is to say, imperfectly on all counts). Sound classification moved far beyond what you’re claiming years ago. If you set the threshold at “imperfectly” then no system will ever meet your bar, because perfect is an unreasonable threshold. However, the claim that we’re unable to differentiate between gunshots and car doors…

Sound classification under fairly well-controlled conditions could be very different from a real world streetcorner situation, where sound can bounce off surfaces and generate different signals depending on direction and distance of origin relative to the collecting microphone. High-quality omni-directional microphones can be expensive and I imagine there would be theft concerns if they used really good ones.

I recommend primo EM272 microphone capsules for use with https://github.com/hcfman/sbts-aru. They are high quality, very sensitive with high signal to noise ratio, lauded for nature recording use cases. They can be bought assembled for around 65 euros in the Netherlands. However these capsules are often found in much more expensive equipment.
Post reply on HN