Hacker News Reader

Comment by bootlooped | Hacker News Reader

My insurance company had me install their app to rate my driving for a couple of weeks. I asked them how they knew I was the one driving the car, they basically said "we can tell with technology and science". This was an unsatisfactory answer to me, because I really don't see how they can tell the difference between when my phone is in my pocket when I'm driving, and when my phone is in my pocket and I'm the passenger.

If they can't tell whether you are the drive or the passenger, it seems like their data on distracted driving would be extremely tainted.

View on HN

FWIW, it's actually possible to determine driver vs passenger side of the car based off accelerometer data (even in pocket). I actually prototyped an app that used accelerometer data to detect phone usage while driving, and that came up during my investigative research. Sadly, my company never followed through on plans to file patents and develop it (this was back in 2012).

The data and interpretations are pretty cool though. We came up with signatures for texting, talking (both phone to ear and holding in hand for speaker), and a few other common activities.

Not saying they are doing anything like this, just that it is possible.

That you could detect that in pocket is kind of mind blowing. I guess minor differences in movement related to operating the vehicle versus sitting passively produces an identifiable signature?

The difference is actually from movement of getting into the vehicle. I don't remember the details of that specific research but I did find it really cool at the time. IIRC, there was also (separate) research that could determine front seat/back seat, based off movement generated from the car turning.

Consider the arc of a typical turning radius from vehicles of varying widths. There would be slight variances in 3d space between a sensor on the left and right sides, obviously, depending upon the direction of the turn. It will be sharper on the sensor if it's on the side of the direction of the turn. Inferring the direction of a turn in an accelerometer is trivial, and inferring the side of the vehicle, and distance from the wheels (front vs back seat) should also be possible. Again, throw a sensor into a vehicle once, and it's hard to tell, but lots of times, you'll get enough data to easily tell.

More so, especially if you know the type of vehicle someone owns. Without knowing the type of vehicle, you would need data from both sides to differentiate between passenger and driver. But it should be trivial to detect the gradient (slope) of the sensor curve while the vehicle is turning either direction to determine whether a passenger is on the left or the right side. Doubly so with more than two dimensions of measurements available (vehicles normally sway a bit on one side or the other while turning one direction or the other). Simple physics that every suspension will follow... Your body will also sway a bit in three dimensions differently if the phone is on your person. Using these principles, you could also determine whether someone was siting in the front or the back, too. Knowing the type of vehicle, or given enough variance in measurements (more data points is better), it should be possible to infer the position of the sensor due to the gradient of the arc in the turning radius.

Just spit-balling here... lol

Guess all this data gets shafted if you live in the UK but drive a left-hand side vehicle though!

Nah,should be trivial to determine your location from cell network and infer which side the driver is on if they know what kind of vehicle you have insured. ;)

Obviously, it's inference though, it's not 100% accurate, and it doesn't have to be.

I don't think their is such thing as a typical turning radius. On a left turn some people go wide and clip the turning lane of the perpendicular street, others go very sharp and turn at the last minute to the point of overshooting the left slightly. Sometimes it changes if there's another car in that perpendicular turning lane, or if there is traffic and you have to gun it on green to get through, or if the box is blocked, or if there's a box in the road you have to dodge, or a rideshare blocking half the lane dealing with a passenger, etc. etc.

I often put my phone on the passenger side when it is charging. And someone mentioned women's handbags on passenger side.

Although presumably that is correlated with not using phone at same time!

Yes, there were clear signatures for phone in pocket / in purse / other non-usage actions.

"signatures", but how good were they? I'd not be surprised if you could do much better than random guesswork, but I'd need to see a lot more data to accept that this is good enough to be scaled up population-wide.

Chances are the data is being fed into a black-box neural network and they don't have damned idea why the black box says your driving is good or bad.

"we can tell with technology and science" ie. "magic".

I would imagine bluetooth would most likely be connected if you're the driver.

If they can get the bluetooth id of the car you're paired with, maybe they could say "this is the common car driven".

And if they had location information, they might be able to match it offline to other data, like a license plate as you drove past a license plate reader.

and if they don't have any of that, maybe I should shut up because now they have ideas.

> I would imagine bluetooth would most likely be connected if you're the driver.

I would think the opposite. When I'm the passenger, I connect to Bluetooth so I can control the music and navigation. It makes no sense for the driver to do that.

It may not matter if it is tainted or not, if it gives them an advantage in pricing insurance policies on average.

Knowing that you do or do not not fiddle with your phone on the road too much as a passenger or driver might be enough of a signal.

Reply on news.ycombinator.com