Straight talking from the Editor’s keyboard;
The problem with Tesla’s automated collision avoidance braking system activating by itself has been around for some time, but recently the mainstream media has reported on it in more detail, as the US driver safety body, the NHTSA has opened an investigation. So called `phantom braking’ incidents have occurred, to the extent where about 750 users have complained to the US driver safety body, up from about 350 which the BBC reported on back in February this year. Tesla Model Y and Model 3 cars are being looked at and Tesla is expected to formally respond later this month.
But it isn’t just braking suddenly, according to CleanTechnica, a website which showcases `green’ tech such as battery powered cars, Full Self Driving mode testing resulted in some phantom swerving. Basically the car decided to avoid hazards which weren’t there, according to the website author’s own regular commuting experience.
DATA MUST BE RELIABLE TO PRICE RISK
For the insurance industry the problems that Tesla encounters in pushing forwards with fully autonomous cars is that ADAS data MUST be 100% accurate, so that underwriters can price risk. This is a particular problem perhaps for Tesla owners, since the company runs its own insurance company, plus they are at the forefront of autonomous car development. So Tesla risk is priced in real time and based on the data being sent from vehicles which showcase the very latest driver assistance technologies.
Let’s be clear, I’m a big futurist and admire Tesla for forging its own path on insurance, but if drivers are penalised for `poor’ driving when it’s actually the car’s fault, then the trust that has to exist between insurer and proposer evaporates. There is an evangelical aspect to Tesla ownership and if you expect your buyers to be test drivers in some respects, then the insurance T&Cs need to reflect the possibility that ADAS data could be less than 100% reliable.
In a highly litigious country like the USA, that is a bold road to drive down, since you risk lawsuits based on rogue risk pricing. It could happen, in the same way that VW-Audi has settled on diesel mpg claims for example, Tesla owners could argue that their insurance was artificially high due to faulty software. A very similar argument, as I’m sure attorney Lionel Hutz would agree.
But let’s move on.
THE BIGGER ADAS PICTURE
Of course Tesla isn’t alone in using more Automated Driver Assistance Systems. Even humble city cars like the VW Polo now have features like main beam permanently on, with the car dipping automatically, plus adaptive cruise and lane change alert, while Travel Assist keeps the Polo a safe distance from the vehicle in front.
Companies like LexisNexis and Cazoo Data Services are using ADAS feature databases, so that insurers and claims specialists can see which features have been fitted to cars, and more importantly, which ADAS tech has been activated – or over-ridden recently.
This stuff matetrs because it is ultimately about pricing risk more fairly for car insurance customers. If your driving score, or ADAS history is good, then your quote, or renewal, is likely to be lower than your neighbour. Data can free insurers from the 20th century practice of pricing car risks on asset value, repair costs, driver occupations and postcodes. In short, it can blend the real time ADAS data from connected cars, into a monthly subscription policy, which is truly flexible and directly reflects the actual risk day-to-day, or even by the hour for delivery drivers and couriers.
Now take this further; if we are going to develop actuarial tables in the future, which are managed by AI, then the data being fed into an automated insurance underwriting platform simply HAS to be 100% truth. No BS, no phantom swerving, braking or car park donuts on your driver scoresheet. The contract between insurer and consumer can only be fair if the technology that underpins risk pricing is working effectively.
This is why partnership between insurtechs and car makers is essential in the future, the worst case scenario is ring-fenced manufacturer data which turns out to be unreliable and inaccurate, being fed into insurer and broker platforms as gospel truth.
LESSONS FROM THE PERIODIC TABLE
In Primo Levi’s brilliant anthology, The Periodic Table, the elements of chemistry are used as metaphors for the nature of mankind and the connections which decide outcomes. Incidents from WWII, and Levi’s time in the Nazi death camps, highlight both the many evils and the immutable, eternal beauty of life.
One of the many wonderful quotes from the book is this; “Matter is matter, neither noble nor vile, infinitely transformable, and its proximate origin is of no importance whatsoever.”
The same wise approach needs to be taken with connected car data. It cannot be assigned a cartoon-like character, like good or bad, or twisted for political ends. It must remain neutral and therefore able to be transformed by over-layering and mixing data streams, so that more complex chemistry and balanced equations can take place at the point of quote. Although the origin of the data does matter, in terms of location, that factor is surely outweighed by the question of accuracy. For without an accurate baseline on ADAS deployment, no insurance brand, even Tesla’s, can honestly process insurance claims, or price future risk by driver behaviour.
In the end, data has be like DNA in court. It must be the final word, the immutable proof.