Hacker News Reader

Comment by bertil | Hacker News Reader

I’m surprised that all this web of data hasn’t led to the most meaningful and significant improvement we still need in the car industry: have people who drive dangerous pay more for their insurance. It takes very little time near a road to notice that some people present an order of magnitude more risk than others, and no one has ever tried to confront them about it — at least successfully.

With cars and their drivers killing more than a million people every year, a little constructive feedback would be a major help to avoid so many tragedies.

View on HN

> present an order of magnitude more risk than others

The data really doesn't bear this point out, or the category of drivers your considering are such a small part of the total that changing their behavior will have almost no noticeable impact on the total.

Further.. at least in the US, the majority of fatal accidents are single vehicle accidents where the driver was impaired either by alcohol or other drugs. You don't really need to mine data from the car to figure out who and who isn't the problem here.

> With cars and their drivers killing more than a million people every year,

That's uncharitable. Bad road design and failure to make protected pedestrian paths (16% of all fatalities in the US are pedestrians) definitely deserve some credit here too.

> a little constructive feedback would be a major help to avoid so many tragedies.

Based on US data: If you drink and drive you should be revoked for 10 years. It should be illegal to give people under 24 vehicles with more than 250hp, or any power level with a turbo.

Commercial, fleet insurance works exactly like this.

Back in the 80's there were already such solutions that would monitor speed and location based on cell tower. the data would be chirped back periodically. The price of the insurance would depend on driving speed and postal code for the cumulative information of the entire truck fleet.

Today, this is not even a question. It is the de facto way of charging fleet insurance.

At least three of the companies listed in the article offer apps that track your driving and give an insurance discount if you sign up. This could be seen as effectively the same thing; safer drivers paying less is perhaps equivalent to dangerous drivers paying more. Yes it’s voluntary so this isn’t perfect, but I’d speculate there is some presumption that in general safer drivers are the people signing up for the discount. This unfortunately comes in the form of a privacy trace-off, but if having dangerous drivers pay more is the goal, I’m not sure there’s a way to have that without some monitoring.

Another way your wish already exists partially is that people who cause accidents have higher insurance rates. This isn’t 100% effective, but some of the people who prove themselves more dangerous really do pay higher insurance already.

> safer drivers paying less is perhaps equivalent to dangerous drivers paying more

That’s not what happens in practice: drivers concerned about their privacy don’t use those apps, not those who drive the most carefully. Subscribers remain a minority. This is a shame because careless driving requires very little information, nothing that is genuinely affecting privacy.

How do you know what happens in practice? I have no idea who’s signing up for discounts. I’d speculate wildly that specific monetary discounts win over generic privacy concerns more often than not, but I have no idea.

But if we’re to have dangerous drivers pay more, without it being a voluntary opt-in system, then someone needs to be able to monitor all drivers, right? What information are you thinking of that isn’t considered private? You could have the cars reporting only speed & steering & accel/decel telemetry, but that might be easily hackable. Having GPS to compare against is much more trustworthy. What if primary components of safe driving are where and when you drive? Choice of roads and time of day may matter for some drivers as much as speed. Maybe the behavior in the proximity of other cars is a primary factor, I wonder how that could be reported - how often you pass, how much room and time you leave when changing lanes, how closely you follow, etc.

I wonder what it would really take to identify dangerous driving. The largest factors identified by the NHTSA are: drinking, speeding, being “distracted” (using a cell phone), and driving tired. Speeding might be the easiest, while monitoring for drinking and tired and cell phone use seem more invasive.

> How do you know what happens in practice?

I’ve done consulting work for that industry.

> What information are you thinking of that isn’t considered private?

Statistical distribution of the absolute jerk. People who race, and distracted drivers have to correct at the last minute both have sudden changes in acceleration.

That seems pretty reasonable as one valid data point, but unlikely to capture dangerous driving broadly and accurately, no? Certainly location and traffic and speed matter, and jerk might not tell you much about people who drink or drive tired; certainly a large percentage of accidents happen without sufficient deceleration prior to collision. I could be wrong though, maybe the accelerometer data over time is reliable at identifying bad drivers, it’d be interesting to see how well it does.

My kids use the insurance company apps and they are pretty awful in terms of accuracy. The apps nit pick the turning and braking based on acceleration data, and I’ve ridden with them and watched it call out safe driving as bad. One downside of this is that neither my kids or my wife and I trust the insurance company app to understand safe acceleration. I’m a little bit worried about what happens to this data and to the insurance company’s conclusions about what stops and turns were safe or not. It would be bad IMO if this record follows people around informing law enforcement using poorly decided thresholds for safety. The crappy app, of course, does not mean that the insurance company can’t reliably identify dangerous drivers, but there’s no indication to me that they’re using the data in a way I’d want or agree with... even if I’m completely on board with your suggestion to identify dangerous driving and charge for it.

The industry is helping distracted drivers by adding stuff like lane-keeping assistance, AEB and what not. The legislator, by requiring these systems in vehicle, helps people drive while distracted. OTOH, those systems do prevent accidents.

Go after the insurance companies instead. They are highly motivated to pass on the costs to their customers. I don't know exactly what would be an appropriate metric though.

Because that’s not their problem. They want all their customers pay more, because people can change their driving habits. It shall never be the reason for profits to fall through.

Such check has been implemented many years ago by insurance companies through data collection from GPS devices.

It's even in the commercials in the US.

edit: and the last four entries on the list from tfa.

Reply on news.ycombinator.com