Hacker News

I got one of those dongles from my insurance company that plugged into the ODB2 port and reported my driving habits.

I was a bad driver. It would frequently beep at me to let me know that I had braked too hard. I was mystified. "What should I have done differently," I'd think, as I raged at the objective machine that judged me so.

The next time my brother came to visit, he called mom. "Oh, and presidentender is a good driver now." I didn't put the pieces together right away, but it turned out that the dongle had actually trained me, like a dog's shock collar.

The reason for my too-frequent hard-braking events wasn't speed, although that would be a contributing factor. It was a lack of appropriate following distance. Because I'd follow the drivers in front of me too closely I'd have to brake hard if they did... Or if they drive normally and happened to have a turn coming up.

Over the period I had the insurance spy box in my truck I learned without thinking about it to increase my following distance, which meant that riding with me as a passenger was more comfortable and it beeped less often. Of course since I'd been so naughty early during the evaluation they didn't decrease my rates, but I think the training probably did make me statistically less likely to crash.

Aim to make the road laminar. Every time you hard brake, you're causing the milk jug to glug, making a ripple of entropy as momentum turns to heat from your brakes and those behind you, sometimes in perpetuity. I learned this while doing a 1.5hr daily commute in a Subaru with a clapped out manual transmission. I wanted to conserve energy shifting, but realized I was now participating in large choreographed dance of "smooth" with other drivers who already knew this. There are many of us. And we all glare at the driver blinking their red lights on the interstate indicating that they're loud and proud of introducing turbulence to an otherwise peaceful system.

Nice research. This is fairly well known in insurance circles. Most auto insurers that do telematics consider hard braking the strongest indicator of risk. One of the things that we do at work (Cambridge Mobile Telematics) is build tools to deal with this risk. We have apps that monitor driving and we play a tone to indicate that a hard braking event was detected. Simply letting people know that they had a hard braking event is an effective mechanism for behavior change (other companies have similar tech)

This type of research is highly valuable but too rare; this is generally because of how we view Road Accidents at a core level:

- Road Accidents: "A driver caused this, let's determine who, and find them at fault."

- With Air Accidents: "The system caused this, let's determine which elements came together that ultimately lead to this event."

The first is essentially simplifying a complex series of events into something black and white. Easy to digest. We'll then keep doing it over and over again because we never changed the circumstances.

The second approach is holistic, for example even if the pilot made a mistake, why did they make a mistake, and what can we do to prevent that mistake (e.g. training, culture, etc)? But maybe other elements also played a part like mechanical, software, airport lightning, communications, etc.

I bet everyone reading this knows of a road near them that is an accident hotspot and I bet they can explain WHY it is. I certainly do/can, and I see cops with crashed cars there on a weekly basis. Zero changes have been made to the conditions.

this google research is a fascinating pivot from the usual driver-centric data we look at in insurance risk modeling. usually we use hard braking as a proxy for how safe an individual driver is. but using it to identify specific road segments or intersections with bad geometry is huge. it basically flips the script from individual liability to infrastructure-level risk assessment.

Too bad there's no map with such indicators, I'd definitely use it for my route planning, especially in unknown area. I usually know pretty well dangerous parts if I drive there frequently.

In unknown roads/highways I can predict hard bumps/gaps by seeing dark oil spots in the middle of each lane.

I'd love to have a danger heat map displayed on a HUD while driving. Say a default green banner that goes red near a hot spot or even animates near a current hazard. Mostly it could use these same stats, but then be strident if anything unpredictable is detected nearby.

Against inferential statistics: https://hdl.handle.net/2381/37564

PDF download: https://iase-pub.org/ojs/SERJ/article/download/215/119/726

I'm really curious what their data looks like at the various racetracks and circuits. Fun fact; most raceways have accurate street-level indicators (including that they are one-way, but sadly they are not the best racing line) on most online maps, and my car did complain to me in its weekly report about a lot of hard turns, quick acceleration, and hard braking with helpful pins on e.g. Laguna Seca or Thunderhill corners.

In theory, the most dangerous turns would probably have higher variance on hard braking data.

This is a great use of this technology. In aggregate, these hard braking events _do_ tell us about road design issues. They also tell us about problematic drivers, in aggregate.

I'll never use one of these dongles, though, because I don't want my every move second-guessed. There's nothing _inherently_ dangerous about isolated hard braking or cornering or acceleration events. It all depends on context. Am I braking hard to avoid an obstacle or mistake by another driver? Is there someone behind me that's likely to rear-end me, or am I in the middle of a highway in the desert? Did I just replace my brake pads and I'm bedding in the new pads?

I don't want to have to worry about whether I've used up my invisible quota before the algorithm decides I should be moved into a more expensive insurance bracket.

Hard braking could be detected externally; you can tell when vehicles are braking hard from the deceleration and suspension effects, without any surveillance equipment installed in them.

That's not gonna be something Google would research, of course, due to next to no alignment with their interests.

There’s a section of I-15 in Utah’s Salt Lake County which reliably has a crash on weekdays at 6pm. It was unfortunately at a pinch point in the mountains with no good alternate route… very annoying.

In a similar way that Google Maps shows eco routes, it’d be fun for them to show “safest” routes which avoid areas with common crashes. (Not always possible, but valuable knowledge when it is.)

When we worked at a p2p car sharing company it was well understood what a treasure trove that past accelerometer data was as good input to frequency prediction of a claim resulting from a particular rental.

Not surprising but it is nice to have these data streams to explore locations that could potentially be remediated. I think anyone who drives interstates in metro areas would agree cloverleaf interchange are generally terrible with any significant traffic. Add in the general proclivity to drive much higher than the posted speed limit and these become dangerous due to the speed differentials and we've known this for 50 years.

"A 1974 study by Hall and Dickinson showed that speed differences contributed to crashes, primarily rear end and lane change collisions"

Hall, J. W. and L. V. Dickinson. An Operational Evaluation of Truck Speeds on Interstate Highways, Department of Civil Engineering, University of Maryland, February, 1974.

What is the actual use of this?

This research team used Google's first-party location data to identify San Jose's Interstate 880/US 101 interchange as a site with statistically extreme amounts of hard braking by Android Auto users.

But you don't need machine learning to know that... San Jose Mercury News readers voted that exact location as the worst interchange in the entire Bay Area in a 2018 reader poll [1]

It's not a lack of knowledge by Caltrans or Santa Clara County's congestion management agency that is keeping that interchange as-is. Rather, it's the physical constraints of a nearby airport (so no room for flyovers), a nearby river (so probably no tunneling), and surrounding private landowners and train tracks.

Leaving aside the specifics of the 880/101 interchange, the Google blog post suggests that they'll use this worst-case scenario on a limited access freeway to inform their future machine-learning analyses of other roads around the country, including ones where presumably there are also pedestrians and cyclists.

No doubt some state departments of transportation will line up to buy these new "insights" from Google (forgetting that they actually already buy similar products from TomTom, Inrix, StreetLight, et al.) [2]

While I genuinely see the value in data-informed decision making for transportation and urban planning, it's not a lack of data that's causing problems at this particular freeway intersection. This blog post is an underbaked advertisement.

[1] https://www.mercurynews.com/2018/04/13/101-880-ranks-as-bay-...

[2] https://www.tomtom.com/products/traffic-stats/ https://inrix.com/products/ai-traffic/ https://www.streetlightdata.com/traffic-planning/

OK, now that you have this data, give me a "prefer safer routes" option in Google Maps navigation!

While you're at it, give me an option to avoid unprotected left turns and to avoid making a left turn across a busy road where cross traffic does not stop. (But only during heavy traffic; it's fine when nobody is on the road.) Not only are these more dangerous, they're also more stressful and they also introduce annoying variation into my travel time.

How long until my insurance company can figure out my commute route and adjust my rates based on the collective risk of the segments?

Dude the fucking 101S/808S connector is atrocious on so many levels.

Like another poster said, this is very well known already. It's one of the reasons why municipalities purchase this data from data brokers.