At least where I am in Europe, insurance premiums and speeding tickets are completely separate. It makes sense that people who get caught have to pay higher premiums though, as I assume they are more likely to be involved in accidents.

Tickets for all violations result in points. Speeding points depend on the amount over the limit. Total points in last 3 years are one factor that goes into insurance pricing.

It's a social anti-credit system.

It is a little dream of mine to mine videos for crazy traffic behavior and sell the license plate numbers to insurers. Should be pretty simple, the uploaders even do the tagging for you. How to do that while complying with GDPR is a little nugget to be solved. I'd need a method for take down and a method for checking the vehicle hasn't been sold (that's public at least).

Probably not going to work for the same reason that the police generally won't go around and arrest someone based purely on dashcam footage: who is to say that the video Person A has of Person B driving illegally is actually real? It is all too easy to doctor videos, and if insurers/police just blindly trusted videos people find on the internet/record on their phone/dashcam/etc then it would be pretty easy for Person A to frame Person B for crimes they did not commit.

Innocent until proven guilty, not Innocent until some grainy youtube footage that kinda maybe looks like you were speeding. Police equipment is calibrated and the evidence stored appropriately for a reason.

That said, at least the in the UK, the police do use video footage from the public to a certain degree. I believe this is often used to go and "have a quiet word" with the driver in question, i.e.: <unexpected knock knock on the front door> "Was this you sir/madam? <shows video on smartphone of sir/madam driving dangerously>" and then give them a warning if they own up to it (...and this sort of intervention is probably enough to on its own without having to go any further, i.e. having a police officer standing on your doorstep with video footage of you driving like a prick and being able to "get away with" just a warning/telling off), but I do not believe that the footage on its own is enough evidence on its own since it is so easy to fake.

I agree. Thing is, most actuarial risk factors are really bad predictors individually. Combined though, they work enough for a functioning market. This would be just another factor. Don't exclude any of those license plates straight up, just ask 5 or 10% higher premiums. Some of these people will stay and pay some more, some will leave for other insurers and stop cutting into your bottom line. No such thing as a bad risk, only wrong premiums.

You can make secure dashcams that add some hashsum/crypto signatures based on all GPS signals received at moment.

That's quite general problem around deepfakes - how to generate video that's guaranteed real. Some form of DRM or blackchain is probably needed, not to anyones liking.

Guaranteeing that the video is genuine still doesn’t solve the problem that [person] != [car] != [license plate]

In my country [person] != [car] is one of the exceptions on innocent proven guilty. You are responsible for who as access to your car and it is on you to reasonably prove that it was not you driving.

Typically happens with family members.

>who is to say that the video Person A has of Person B driving illegally is actually real?

Is this ever a real concern with CCTV footage?

But sure, a dashcam won't necessarily ID the driver, just the car.

I've suggested this to an insurance company. Unfortunately keeping a list of dangerous drivers (or dangerous license plates) is a bit illegal re: GDPR and using it for pricing mandatory traffic insurance is also not doable. Pricing is only allowed to use factors that can be demonstrated to correlate with the risk the company sees in actual claims.

But one can dream :) A good thought technology for reducing anger towards reckless drivers is to assume the crazy BMW driver just had a bad case of diarrhea and needed to get to a bathroom very fast.

At least in NL the privacy authority (under GDPR) and courts (under former law) afaik have ruled a license plate is not personal information. Filming in public is legal. Insurance contracts are two party contracts where there is basically freedom to accept or not. Also no specific rules on actuarial factor, outside of 'illegal discrimination' (all pricing is discrimination). So all those flags are still green in my book. (I work in insurance.)

Thing is, doing this will get you in a shit storm so that might be the simple reason no insurtech has tried it yet. Perhaps some smaller insurers are doing it and keeping their mouth shut. No problem if nobody knows where the license plates on the exclusion list are based upon.

And a method to verify the identity of the person actually doing the driving at the time.

In NL we (mostly) insure cars, not drivers. My wife, my kids, even my neighbor or bookkeeper can drive my car and be insured. If they crash it's my premium that gets adjusted. So if someone catches them doing crazy stuff with my car, I find it quite right morally to adapt my premium. Deep fakes are a problem though.

This is the case in my country as well, it make sense that the car owner should be responsible for whatever actions that car takes if the driver cannot be identified.

Even if it is stolen it works, as you would obviously file a police report for the theft and so that would indemnify you of any crimes committed from when it was stolen.

The only issue is what happens if your bookkeeper gets drunk, borrows your car (because they know where you keep the keys) and then commits a crime?

There are more companies doing that than you think.

You can test it in countries that aren't limited by the GDPR.

> If you give people a choice, the super wealthy mostly will pay a relatively small fee and go about their day, and you'll end up temporarily incarcerating the very poor for the same social malfeasance. (With some distribution of behaviors in the middle.) Is that better?

I think this pilot is giving people a choice, but any actual implementation would not.

It’s probably a choice because of restrictions upon this “innovation unit”. If deployed for real, itnprobably wouldn’t be a choice.

> If you give people a choice,

I would be curious to see how/if it brings down speeding if they don't get a choice at all but just have to wait. It's not very practical but I do think it would work in principle.

that's how I read that. this experiment specifically gave a choice to study which penalty is more popular, which is probably inversely related to its effectiveness.

that is if the study is on the speeders, I'm not sure the average person wants lighter penalties on speeding.

How do the insurers get to know about those tickets?

In the US at least, it's public information. The ticket is usually a civil infraction that results in a legal judgement against you.

For the average person, this info is probably stored in a basement near a beware of the leopard sign but insurance companies will go through the trouble to get it.