Hacker News Reader

> valuable data to insurance companies

Great idea!

But on the other hand I'm sorry to inform you that your insurance has gone up, because we found that drivers in your neighbourhood are in general 2498% more likely to have accidents. Maybe you're next. We'll just make sure that rates have already been adjusted to account for that.

Also we saw that you were driving (very smoothly, congrats on that!) around some rather shady parts of town. This will lead to another rise in your fees, we just want to make sure that we got you covered in case your car gets stolen...

> But on the other hand I'm sorry to inform you that your insurance has gone up, because we found that drivers in your neighbourhood are in general 2498% more likely to have accidents.

These types of statistical analyses are exactly why insurance is as cheap as it is.

I'm no fan of all the discounts insurance companies hand out for handing over private data like driving behaviour or healthcare stuff, but insurance companies with no risk management can't compete in any kind of open market.

I would argue that in terms of broader accessibility of insurance, these microtargeting approaches may actually be an improvement for many. If you live in a troublesome neighbourhood you may be paying through the nose for getting categorised as "one of the bad ones", but these individual data points can help prove you're a responsible driver and escape some of the cost of having a shitty start in life.

I'm still against these data collection practises ("you want privacy? Live in a rich neighbourhood and get a college degree!", plus a lack of any kind of law preventing these companies from sharing or selling that data) but there are solid arguments in favour of feeding insurance companies more data.

As the modelling gets better, are we ultimately heading for a future where the only people who are able to get insurance are people who don't need it, because they aren't going to make a claim? If insurers get too good at predicting the future then it breaks their entire business model.

Insurance is for unpredictable outcomes such as a lightning strike hitting your car or even just another driver having an undiagnosed medical event that causes them to crash into your car.

Why would you want insurance for predictable events when you can just put the money to one side yourself and not feed the insurance companies' profits?

Insurance is for hedging against a risk. It has nothing to do with predictability. In fact, the bedrock of insurance is predictability.

You have a 1% chance of sustaining $1MM in damage. Do you carry around a million dollars or do you get insurance?

What about a billion? What about a thousand? Congratulations, you now know what you’re willing to pay for a deductible and what you’re going to look for in an insurance policy max.

I'm not understanding your point. The 1% chance means that you are uncertain whether or not you are going to be in that 1%. Otherwise, some people would have 100% chance of sustaining $1MM in damages and others would have zero% - that's predictability.

Similarly, insurance companies frown on people getting insurance when the person knows that they have a claim to make - when there is certainty, then insurance is the wrong product and may be considered insurance fraud.

I think you're mixing up predictable events with probabilistic events. Just because you know the average likelihood of an event occurring does not mean that it is predictable. (e.g. the chance of getting red or black on a roulette table has known odds, but you're only going to be able to win if you can predict the outcome)

The un in unpredictable got autocorrected out of my original post (It has nothing to do with UNpredictability.)

My point was that insurance isn’t about unpredictability, it’s about cash flow and risk hedging based on probabilities.

> My point was that insurance isn’t about unpredictability, it’s about cash flow and risk hedging based on probabilities.

I disagree. If you can predict that your house will never be burnt down, then you wouldn't need buildings insurance - it's the unpredictable nature of these events that makes insurance useful. Yes, the costs of these unpredictable events are spread out across the insurers so that they can turn an unpredictable event that causes a huge expense into a predictable insurance premium instead.

We are looking at this from different points of view.

You define predictable as: I can predict if and when this will happen to me.

I define predictable as: I can predict that this will happen one in one hundred times.

Both are correct definitions of predictability. In the insurance context, your view is the one of the insured and mine is the one of the insurer. The switch between the two occurs at some risk-adjusted monetary threshold. In other words, whether you are an insuree or an insurer depends on your risk tolerance. For example, most home owners are insurees for the purposes of fire. Most home owners are also (self-)insurers for the purposes of, say, a leaky faucet. The decision to procure insurance is therefore not predicated on the unpredictability of the event but on a risk adjusted monetary basis (ie, cash flow and risk hedging).

I'm using the standard dictionary definition of "able to be known, seen or declared in advance" (there's also "behaving in a way that is expected" which is not relevant). We are discussing events that are not known in advance, so by the dictionary definition, they are not predictable. Knowing the probability is not the same as you will not be able to say "it's going to happen this time".

Sorry, but you're using a bizarre definition of predictability. If I were to say that the chances of anything coming from Mars are a million to one, then you'd be considering it "predictable".

I don't know what it's like in USA, but where I'm from insurance premiums change a lot from location to location, so much that some people change their residence to a cheaper city.

They change a lot in the US, too, and different companies have different methods of spreading risk across geographical areas, which can lead to shockingly different prices from different insurers depending on your location.

> ... your insurance has gone up, because we found that drivers in your neighbourhood ...

That's what happens now. This would be "your insurance has gone up because you are more likely to have an accident. If you drive more carefully you can get a discount."

It's actually a thing already here in the UK. Many insurers let you get a black box and if it shows that you drive carefully you get a discount on your premium.

I believe insurance companies already look at your location when deciding premiums, so this would just give them greater granularity on your actual driving habits.

In the limit of perfect knowledge being transferred to your insurer, will an insurance stop being one, instead turning into an obligation to pay a middleman for all your claims?

Only if that perfect knowledge includes knowledge of future events such as a tree falling on your car or it being struck by lightning etc.

What you describe is actually becoming common.

What I want is something in cars that makes it impossible to go over X inside city bounds etc. The tech for knowing the speed limit isn't good enough yet to be used for limiting, but some kind of max limit in an area a bit over the highest speed limit would help a lot.

Electric scooters have geofencing and speed limiters. My drone has gps zones it wont fly. Electric bikes have speed limiters. But you can just take your car and blast 150+ down a residential street if you'd like. Just insane. It's a solved problem, just need enforcement.

> But you can just take your car and blast 150+ down a residential street if you'd like. Just insane. It's a solved problem, just need enforcement.

There are already a solutions, it is called police, speed traps, fines, and license revocation. Not everything has to be done with preventative measures by a nanny state.

And how well has that worked so far, you feel?

Having limits on things doesn't automatically make things a "nanny state". Don't argue like that. Is it a nanny-state that there are speed limits?

> Is it a nanny-state that there are speed limits?

No, speed limits are fine. Its enforcing them with technological tools that are not under control of the citizen that is not fine.

Also, at least in my country, the code says that I am allowed to break the speed limit if there is a good enough reason (e.g. I am trying to escape from somebody that wants to kill me, I am bringing somebody urgently to the hospital because he may otherwise die).

For example one of my friends was born during the Oil Crisis of the 70s, and at the time car circulation was limited (odd plates one day, even plates the other). Both the parents had the "wrong" plate that day, but they decided to use the car anyway because it was an emergency. They got stopped by the police, and once understanding the issue, the police escorted them to the hospital with siren and lights.

How do you do that with externally imposed limit ? You don't (sorry your kid didn't make it, but that's the law) ? You prescribe a button that says "I'm in emergency" ? (no-one drunk enough will ever use it, I am sure, and neither will criminals using a stolen car).

Also in general there is the issue that in a healthy society most people will respect the law, and only a few break it. Putting restrictions on otherwise law-abiding citizens tells them the state doesn't trust them, and in the long term citizens will answer in kind.

edit: I will add that even if it were against the law, in some cases breaking the law and paying the consequences (paying fine, losing license) may still be worth it if it is an actual emergency.

You're still going to need traditional enforcement. The people doing those sorts of speeds modify their cars will just push a new tune to remove it (remember, most cars today are limited by the factory well before that to match tire speed ratings). So there's already evidence that limits set today do nothing to prevent this sort of abuse.

I guess gun control doesn't work either then.

In the UK, when looking at google drive maps I can already see the speed limit correctly in 99% of situations. This is occasionally incorrect due to recent changes or roadworks, but not very often. It also warns of upcoming fixed and mobile speed cameras. This is ironic as such trackers used to be illegal in Europe but are now built into European cars/mapping apps.

I don't really know why cars are allowed to be sold to go such ludicrous speeds in the first place. Sure, there's no speed limit on the Autobahn, but there aren't many countries where you can legally drive that fast anywhere.

Short bursts of acceleration may be required for evasion sometimes, even at high speeds, but you don't need to drive 160 km/h for more than a minute, and you won't be hitting the Autobahn if your car is sold in South America.

There are legitimate use cases. Some people go to the track, emergencies in sparely populated states when emergency services are far away, etc.

The other thing to realize is the people who go those sorts of speed also tend to modify their cars. They'll just push a new tune that removes the limit. So you still need traditional enforcement.

> It's a solved problem, just need enforcement.

In reality it's a political problem. Here in the UK, certain right-wing media outlets and personalities are often complaining about the "War on Motorists" which is ironic as when you look at RTC data, it's the motorists that are massacring pedestrians and other road users.

>However, I've long thought that motor vehicles should have some kind of black box that monitors various driving metrics such as how "smooth" the driving is. This could provide valuable data to insurance companies and they should provide a sizable discount for drivers who agree to provide that data and have good driving scores.

This already exists, at least in the UK market. A lot of insurers will give you a discount if you let them fit such a "telematics" box to your car. It's still a privacy nightmare.

They have been doing this in the US, for some time.

Fleet cars tend to have these systems, as do trucks. I have a friend that drives 18-wheelers, and often complains about getting nagged by Safety, when he does something out of bounds, or, more often, when drivers around him drive out of bounds.

It’s definitely an insurance thing, with his company. They probably save millions per year, using that system. I explained to him that’s why they’ll never make an “exception” for him, because he’s skilled and experienced. The system is there for all the low-rent knuckleheads they hire.

> A lot of insurers will give you a discount if you let them fit such a "telematics" box to your car.

And for bonus points (pun not intended) will suspend your insurance because they think you're speeding even when you're not [0], with not a lot of recourse.

Driving without insurance could get you a £300 fine and 6 points on your license - enough to cancel it in one go if qualified less than 2 years, which is usually the target audience of these boxes.

https://www.bbc.co.uk/news/uk-wales-67335868

That's obviously a fault with that insurance company and the recourse is to show that the particular road had a 60mph maximum limit. I'd expect you could sue them for breach of contract too if they suspend your insurance for a false reason.

> When the comprehensive part of his cover was suspended on Friday for all the speeding alerts, Connor took a day off work to call the company and try to find a solution.

> Sterling Insurance made Angela and her son go to the locations of his alleged speeding infractions and take pictures of the road signs next to a digital map on a phone.

Yeah, just show them that their data is wrong - no big deal right? I guess in an ideal world such litigation would happen cheaply and quickly too.

If a company doesn't react rationally, then taking them to court is the solution. Ideally you should choose to do business with a company that isn't so obstinate that you have to resort to courts to remedy issues.

Certainly, if you get into legal trouble for not having valid insurance, you could easily demonstrate the facts as to why the company illegally revoked your insurance and how they had no right to do so.

Reminds of when tomtom sat nav data was used to find locations where users where speeding to set up speed traps

https://www.theguardian.com/technology/2011/apr/28/tomtom-sa...

Well, I'm in the UK but not a driver (cyclist by choice). My point is that the black boxes should be almost mandatory (or at least fitted by default) in order to detect health problems which I don't think the insurance company provided boxes do. Have it be a standardised device and provide the data to the owner so that they can decide on whether or not to provide that data to third parties (probably worth having anonymised data available to the government too). It could be a very useful early warning of a number of conditions as well as providing an incentive for drivers to pay attention.

What sort of early warnings are you thinking about exactly? Also, lol, no, governments are not gonna bother propely anonymising the data too. Plus if it's anonymous what's the point of this data?

Early warnings of dementia or even eyesight problems - both of those are likely to lead to a change in driving style with more pronounced braking as the driver reacts slower to events.

It would make sense to anonymise the data before sending it to the government and the point of the data is to identify road congestion and problematic junctions. It could also feed into better tracking of traffic lights timings etc. None of those uses have a requirement to know whether it's Mr Jones in his red Corsa or Mrs Malaprop in her SUV.

location data is basically impossible to anonymise: if you have a record of someone's movements over any more than a day you can almost certainly work out where they live and work, for example. When it comes to personal data you can't just wave an anonymisation wand and say it's definitely OK now, you need to think carefully about the exact case, and especially the more data that is collected the more difficult it is to actually keep things private.

Fully against all remote monitoring, especially having insurance companies having access, along with no kill switches.

Remote monitoring, aside from the incredible security risks, provides no context as to why.

Frankly, I think it should be illegal for cars to have any form of wireless networked access. I personally disable all of that in any car I buy, typically by removing antennas and replacing with resistors.

> However, I've long thought that motor vehicles should have some kind of black box that monitors various driving metrics such as how "smooth" the driving is. This could provide valuable data to insurance companies and they should provide a sizable discount for drivers who agree to provide that data and have good driving scores.

This has been available for years from some insurance companies.

There’s no reason it needs to be built into the car or to automatically report it against the driver’s will. It’s an opt-in thing.

My coworker got it when they first came out because she thought she was a safe driver and wanted a discount. She got a monthly report of how many sudden stops it recorded and the number was much higher than she would have guessed. She never could figure out what was triggering it, but began to suspect it was the stop signs between her house and the office. So instead of coming to a complete stop she started slowing down and then rolling through (which isn’t legal in our state). Unintended consequences strike again.

This unfortunately already exist in europe. Young drivers can get a black box installed in return for a discount on their insurance. This is because those costs for young drivers can be astronomical.

I would never trust a private for profit company with this much data.

> However, I've long thought that motor vehicles should have some kind of black box that monitors various driving metrics such as how "smooth" the driving is. This could provide valuable data to insurance companies and they should provide a sizable discount for drivers who agree to provide that data and have good driving scores.

That just opens the door to the driving version of "your grandma had cancer, so you won't get health insurance". Unfortunately, it also already exists.

> That just opens the door to the driving version of "your grandma had cancer, so you won't get health insurance".

I think that highlights the stupidity of a privatised health insurance system. Not wanting to know about health problems due to a financial disincentive is not a sensible way of managing a populations' health, when typically prevention is much cheaper and more effective than waiting until obvious symptoms show up. I suppose it depends on whether profit is more important than the population's health.

This already exists and has stopped people I know personally from driving like idiots.

https://www.boxymo.ie

Sometimes sudden braking is due to other drivers, animals, etc.

They already have data readers that plug into your OBD2 port for insurance purposes.

> Sometimes sudden braking is due to other drivers, animals, etc.

Those events will be much rarer than the type of late braking that poor drivers will make due to lack of observation. It'd be trivial to separate out the emergency stops from the everyday driving data.

I highly doubt it will be trivial. I rarely have to brake hard. But when I do, it's usuallynin an area where the other drivers are terrible and hard braking is simply more common (as are running lights and stop signs, eg Philly and parts of Jersey).

So yes, in a population of good drivers it would be simple to select people with higher rates of hard braking. But punishing them without context could lead to false punishment (similar to the cases of nurses that just happened to have higher rates of patient deaths - the prosecutor fallacy I think). Once you surpass a threshold of bad drivers in the population, it's more likely in my mind that the hard braking people may actually be the good drivers avoiding the ones running lights.