Comment by notmyuserlogin | Hacker News Reader

Comment by notmyuserlogin | Hacker News Reader

Parent Full thread

It isn't just a problem with Android. I volunteer for a small fire department. We respond to about 500 calls a year. Since January I can think of three times the automatic crash detection on iOS devices has called us out by mistake. 1) A person left their phone on their car and it fell off. Being a small town one of the volunteers was able to find the owner and bring them the phone. 2) A gps location in the middle of a lake. The best we figure is one of the people on a jet ski or wake boarding. 3) Some people jumping on a trampoline.

Each of these means 2-6 volunteers responding from home to the station and then spending 30-60 minutes driving around in large trucks looking for non-existent emergencies. Each call also gets an ambulance staffed with career paramedics.

On the other hand someone's Apple watch did call us and we found he had fallen and gotten stuck down in some bushes and did need our help.

There is lots of promise, but also the tax payers are footing the bill for the false positives, not to mention the added risk to responders.

For everyone blaming modern tech: The only time police have ever come to my house from a 911 call was back in the 90s. Some combination of a noisy phone line and a broken 900MHz cordless phone managed to call 911 and they followed up. They said not to worry, it happens all the time and was a notable portion of their calls.

These types of false calls have been happening for a long, long time. We should get more data and fix the system for sure -- but this isn't a new dynamic and the historic baseline before smartphones isn't zero.

Back in 1996, I was living in Almaden Valley (South San Jose) and we had underground utilities. We also lived on top of an underground stream.

After a rainstorm, water got in and intermittently shorted out the phone line. It was clicking like crazy!

I was on my cool new Motorola StarTAC talking with Pacific Bell to report the problem. Then I heard a loud knock on the door: "San Jose Police. Open up!"

I asked the officers what the problem was and they said "We got a 911 call with no one on the line. We tried to call you back, but no one answered. So we had to come out and investigate."

I invited them in and said, "I think I know what happened." They followed me over to the landline speakerphone in the kitchen and listened to the clicking.

Then I explained, "You remember the old rotary dial phones? They worked by making and breaking the circuit, just like this clicking. Even if we all have touch-tone phones these days, the phone lines are still compatible with the rotary dial. So somewhere in the midst of all this clicking, there were nine fast clicks in a row, and then one click, and one more. And that dialed 911. Sorry about that!"

I have to register my building's alarm with the county and pay a fine after 3 false alarms, or not register and the police won't respond. I also pay an annual registration fee.

I wonder if too many "smart" false alarms will lead to similar regulation.

In my city, monitored alarms aren't worthwhile for non-commercial properties. For the police to respond, your alarm has to be registered, and registration requires authorized first responders with keys to confirm that an alarm is false or valid before police are called/dispatched.

Some alarm vendors will offer "video verification" which typically satisfies that requirement in most markets that require it. Many Alarm.com-based vendors (like Surety Home) can enable it on your system.

I’ve read through the city bylaw here, and there doesn’t seem to be any affordance to “video verification”. The only thing that’s easier with a professional company is that if you have a monitored alarm without security staff in the city, you must provide the police with contact info of two authorized keyholders, while you only need to give them the contact info of a single alarm company that’s capable of responding to alarms on a 24/7 basis.

The only notable exception I can see is that monitors of alarms at financial institutions are allowed to contact police directly without sending someone else to investigate first.

It generally feels like the city/police here has decided that alarms are not a thing for the police to deal with, unless it concerns banks.

I've had my iPhone call 911 twice.

Once I was tucking to pick up speed and I must have accidentally held the side button down in my pocket. I didn't notice anything had happened until I got a call back asking if I was okay. (I did hear the countdown alarm that plays, but misattributed it to a snowmobile or other equipment at the ski resort.)

The second time I was also skiing and did actually fall. I was unhurt but I guess going fast enough to trigger the call. Unfortunately I couldn't get myself situated enough (gloves, zippered pocket, super steep hill) to cancel before the call went through.

The dispatchers were great in both cases. Asked me a few questions to make sure nobody in the area needed help and nothing else happened.

> A gps location in the middle of a lake. The best we figure is one of the people on a jet ski or wake boarding

How do you rule out BUI/drowning? It would suck to be given up on. Is there ever any information indicating the call was placed by a device?

A classmate of mine accidentally drove into a lake and drowned when GPS/E911 conflict dispatched responders to the wrong location. It's not a perfect system to begin with, and made worse by automatic dialers undermining responder trust.

I'm happy to foot a bill for false positives if the true positive rate is 1/4. Seems like money well spent.

Are these false positive data points logged in a database?

Are your records shared or aggregated up to a central agency?

Wondeing how we might calculate the cost/benefit analysis.

I hate Apple as much as any other large tech company, but 1/4 true emergency rate seems like a pretty good start in cases where a person's life may be at risk!

Emergency services in my city (the one you call via phone) have a "true emergency" rate of about 1 in 10 according to emergency personnel I talked to, so it is always a matter of balancing the false positive/false negative rate.

The boy who cried wolf had a better rate 1/3.

911 is not well funded in the majority of the US. In many rural areas, one false call that requires EMS sent out could cause another person with a legitimate call to wait a hour or more.

Another way to look at it - 1/4 of the time people who needed emergency responders had to wait because they were busy looking into false alarms.

Seconds literally matter for many emergencies.

This conclusion doesn't follow, it only makes sense if they are 100% busy all the time. I volunteered in an emergency ambulance a little bit and most of the time we waited in the waiting station.

Another way to look at it - 1/4 of the time people who needed emergency responders had to wait because they were busy looking into false alarms.

Your position makes the assumption that the rest of the emergency services infrastructure is at maximum use at all times.

The OP was talking about a place where they use volunteers, so it's not likely that they're constantly in use.

While you are correct that seconds can sometimes matter, it's not always true. Not every emergency call is life-or-death. Not every emergency call even requires a response.

Imaging a hypothetical world where every call is a true emergency, and emergency services are at 100% utilization 100% of the time is arguing just for the sake of arguing.

I live in a place where emergency services is over-taxed. But I'd rather have actual lives saved with a certain number of false alarms than have people die because someone decided that perfection is the only option.

"1/4 of the time people who needed emergency responders had to wait because they were busy looking into false alarms."

That would be even better, as that would mean 3/4 of the alarms are hits/non false-positives. I argue that even a 1/4 hit rate, i.e. 3/4 false positive rate, is a good start.

I don’t think that follows from their example at all.

Reply on news.ycombinator.com