Hacker News Reader

Comment by blackaspen | Hacker News Reader

I don't think that braking slower, or accelerating less (or even limiting turning g-forces) is really an indicator of "Safe Driving". Yes, it is definitely a signal, and obviously a good one otherwise insurance companies wouldn't be using it. But I think that the mark of good driving is knowing what you and your car can do, and actively paying attention to the road around you (several hundred feet behind and in front), and thinking about piloting the 2 ton piece of metal you're in as opposed to blindly doing it. A better, more strict drivers licensure program is the number one way to increase safety on roads and thus lower the cost of insurance -- but it has to be all in or nothing.

On the flip side, I would find it somewhat amusing to put one of the trackers in my track car and see what they think of that...

View on HN

> I don't think that braking slower, or accelerating less (or even limiting turning g-forces) is really an indicator of "Safe Driving". Yes, it is definitely a signal, and obviously a good one otherwise insurance companies wouldn't be using it. But I think that the mark of good driving is knowing what you and your car can do, and actively paying attention to the road around you (several hundred feet behind and in front), and thinking about piloting the 2 ton piece of metal you're in as opposed to blindly doing it.

If it is a "good signal" then it's an indicator pretty much by definition. The thing is, it doesn't matter what is the poetic mark of "good driving"; what counts are accident statistics going down. One constant about all humans is that you can't ask them to think, or care, or be nice, and expect them to change behaviour. You need to make them change, by changing their habits and/or incentive structures around them, of which, in the case of cars, the insurance bill is a very important part. People behave in a deterministic way, especially when you look at big groups (such as drivers). You don't yell at a river and expect it to change it's way, you force it to do it.

> A better, more strict drivers licensure program is the number one way to increase safety on roads and thus lower the cost of insurance

That would most certainly help, but I wouldn't call it the number one thing (the number one thing would be banning human drivers and replacing them with self-driving cars). I'd love to see driving licenses requiring training comparable to pilot licenses.

So I guess it has accelerometers because GPS doesn't refresh quickly enough to measure acceleration.

So, who's going to design/build a nice shock-dampening mount for the devices?

> So I guess it has accelerometers because GPS doesn't refresh quickly enough to measure acceleration.

True, at least not for anything less than a second or two.

> So, who's going to design/build a nice shock-dampening mount for the devices?

Android devices have accelerometers built in that are cheap and would provide adequate resolution for this kind of measurement, and that can't be easily broken. No shock mount required.

I think the idea of the parent comment was to build a mount that would hide periods of high acceleration. I imagine that's only possible if your "shock mount" is significantly longer than the car itself.

> I think the idea of the parent comment was to build a mount that would hide periods of high acceleration.

Ah. Missed that.

> I imagine that's only possible if your "shock mount" is significantly longer than the car itself.

Yes, true -- a little calculation shows it would be nearly impossible to hide a car's jackrabbit start from within the car.

Darn. Maybe someone will come up with something that sits on the I2C or SPI bus of the accelerometer chip and does ???.

How about not fighting the measures created to help drivers avoid killing people?

How about refusing to create spying devices of any kind on ethical grounds?

You'd have to refuse to create computational devices of all sorts. It is the case of the Great Web of Causality, that things are heavily interconnected with each other. Everything you do inflicts consequences on everything else within your light cone, and back-tracing effects to what caused them is only the matter of enough data and computing power.

To put it in a less grand/mythological way: they can always find another way to get the data they need. It's only a matter of computing power and having a little nut of imagination in one's head. Think of all the side-channel attacks talked about recently - reading speech data from laser, reading speech data from high-speed video, reading crypto keys from power consumption patterns, reading crypto keys from voltage variations on Ethernet cable, reading data from screen reflections off the wall, reading data from RF emissions of CRT displays, reading keypresses from phone accelerometer, reading speech from phone accelerometer, etc. etd.

There's always a way.

That doesn't mean I will be the one to build it. Many undesirable things are imminently possible, easy even, but very few people do those things. With enough solid arguments and education, mass surveillance, even ostensibly for good purposes, can be added to that list.

I don't know... The more I think of it, the more I start to believe some form of technological determinism is driving humanity. That the new technologies determine what will be taught to people and what they will argue about, not the other way around. I mean, do we have a single example of a new technology that could be useful to some while harmful to others, that we collectively agreed not to use or pursue, by means of arguments and educations? Nuclear weapons don't really count, because their use seems dictated solely by game-theoretic reasons (MAD that saved us from global destruction, some nations like Russia apparently pushing their luck by declaring readiness to use tactical nuclear weapons to support their conventional forces, etc.).

what counts are accident statistics going down

Kind of. What counts to the insurance companies is having to pay out as little as possible, while still taking as much of people's money as they'll give. So they do want fewer accidents, but they also want more paying customers that won't file claims (uninsured/undocumented driver's fault) or have claims filed on them (when it's their own fault). Using incentives to manipulate (manage) customers' behavior achieves this, with the PR goal of "lowering accident statistics."

Fair... I don't yell at rivers.

But by your same logic the number one thing isn't banning drivers and replacing them with self driving cars (have you ever written perfectly, 1000% bug-free code?), it's banning all movement faster than walking speed -- and even then requiring people to wear crash helmets and blinkers and "blind sport information systems"...

There is no perfect approach. But I don't think that insurance companies doing this is even on the 'better' spectrum. It doesn't solve people: not turning on their lights, not merging correctly, talking on the phone, keeping their cars in safe operating condition...

Yes, paying attention to a few numbers and alerting people helps -- but if you want real improvements, you have to tackle the source.

Much like trying to drive on a highway of people who's sole purpose is to "achieve the best MPG EVER!!!!" this encourages hyper-focus in the wrong direction.

> Fair... I don't yell at rivers.

So don't yell at drivers. They won't listen, they won't change.

> But by your same logic...

A small digression: that's why I hate "logic". Logic is so 1200s. We have this thing called probability theory and we really can deal with numbers other than 0 and 1.

It's all about cost-benefit analysis. Cars get people killed (random statistics time! It's about Hiroshima + Nagasaki every two months.). Pedestrians not wearing crash-helmets rarely so (I'm not counting idiots who think they can go on a highway on a bike at night without any light or HV clothing; but there it is again, they fail at basic risk/reward calculation and you need to force them somehow to behave). Self-driving cars don't need to be "perfectly, 1000% bug-free", they just have to drive better than humans, which doesn't seem too high a standard and I'm confident it will be reached in few years.

> There is no perfect approach.

Couldn't agree more.

> But I don't think that insurance companies doing this is even on the 'better' spectrum. It doesn't solve people: not turning on their lights, not merging correctly, talking on the phone, keeping their cars in safe operating condition... (...) but if you want real improvements, you have to tackle the source.

Well, I honestly don't know of any better way. Education doesn't work. Most of the drivers I know don't accept the fact that their reckless driving is a part of the problem (I do know exceptions, but sadly, those are mostly exceptions, and they still seem to care about insurance costs more). Social expectations are really hard to manufacture, and the current one worldwide seems to be "driving is a right, drivers are the best, lawmakers and traffic controllers impose bullshit restrictions on speed to make money on us". So forcing people financially to behave seems like the only real option left.

> this encourages hyper-focus in the wrong direction.

I think I get your point - you're afraid that our values (safety) and insurance company's values (profit) are not well-aligned. I agree. But I believe they're aligned enough to push us in the right direction some more in an effective way. But I do believe that at some point those values will fall out of alignment to the point that we might be forced to ditch dependence on insurance for behaviour forming. The question is when.

Reply on news.ycombinator.com