Insurers are weighing up the differing risks of traditional and autonomous driving. Morag Cuddeford-Jones reports

Asmore vehicles rise through the levels of autonomy the real concern of the insurance industry lies in the mix of autonomy and manual drivers on the road.

In an £8M UK government-funded trial of autonomous vehicles led by the Transport Research Laboratory (TRL) in 2017, it was discovered that human drivers might alter their behaviour if they knew autonomous vehicles were present. Not only will the risk profiles pertaining to different drivers and cars change but risks without precedent will begin to emerge, particularly from the new behaviours identified in the TRL’s research.

Additionally, insurers are going to have to mitigate for differing levels of data available to them and create a framework to ensure fair dealing with customers at all levels of machine-assistance.

The impact of new driving behaviours

The TRL research determined that it wasn’t so much the autonomous cars that could pose a problem but ‘standard’ drivers reacting to them. In one simulation drivers tended to pull out into smaller gaps when a large proportion of autonomous vehicles were on the road.

Roger Lanctot, associate director of Strategy Analytics’ global automotive practice, doesn’t mince his words: “We have seen the manifestation of the old saying that ‘the safer cars are the less safe people will drive’.  Tesla opened the door to a wide range of irresponsible, negligent and dangerous driving behaviours.  But when the requirement of a hand on the wheel was restored, much of this went away – and more corrective action occurred after the implementation of geo-fencing of the autopilot function along with limiting the speed.”

Marcus Rothoff, Volvo Cars’ autonomous driving programme director, agrees: “The risk profile is changing. There will be gradual steps and the first will be in creating a safer environment for everyday driving. It’s the boring part of driving where people get distracted and we currently don’t train people to be good at this. We’re really good at driving but really bad at doing nothing.” He cites the importance of functions such as automated lane control as one of those features to help people drive more safely.

Rothoff suggests insurers can be involved in the creation of advanced support systems. “Cars need to be able to support drivers with both collision avoidance and helping the them stay alert. Manufacturers can work with insurers to understand how small incidents are happening and help drivers stay on top of things.”

The impact this will have on insurers’ ability to model driving risk is, as yet, unsure. Lanctot adds: “Mixing autonomous cars with non-autonomous cars will likely create new driving scenarios where there are greater deltas between the vehicles driving on the same road.  The variability of driving speed and behaviour – with some influenced by computer-based decision-making – is likely to increase claims.  There is insufficient data to model these scenarios today.  UBI and ADAS data are only for individual cars and do not reflect the broader driving environment.”

Assessing claims in a semi-autonomous environment

The mix of man and machine becomes problematic when it comes down to claims reporting and assessment. Right from first notification of loss (FNOL) there is the presumption that any data from an autonomous vehicle is more accurate and less open to interpretation. However, this doesn’t mean they are infallible.

Lanctot explains that: “While event data recorders and cameras will record crash events – attempting to understand precisely the nature of the decision-making process behind the autonomous driving vehicle will be nearly impossible. The human driver, on the other hand, can give testimony as to his or her motivation and intentions.”

Understanding how to process claims involving assisted and non-assisted drivers will rely not just on data from the time of the incident but also a whole layer of data learning acquired during the driving lifetime. David Karlberg, strategic product director, Wireless Car, warns: “There are discussions with regards to how much data should be shared. Obviously more data is better to figure things out but there is also the need for security and privacy.”

Currently there is more data regarding the human in a claims event including smartphone use, driver monitoring systems and event data recorder.  The autonomous vehicle will have been simply ‘doing what it was told’.

Ultimately, this difference in risk factor between fully autonomous vehicles and various levels of driver assistance could lend itself to splitting the elements insured. Karlberg notes: “Google has been focusing on not having a steering wheel and others are looking at systems that just enhance and support the driver. There’s currently a lot of discussion as to whether Tesla versions are safe or not and there will be different legislation to cover these questions. We’re in an exploratory phase. It’s a huge change in how society works.”

Lanctot concludes that, in an insurance and claims environment, the capabilities and behaviours of autonomous versus human drivers may be so different, they become in essence separate entities: “The latest insurance industry research initiatives are built around insuring the car and the driver separately.  SwissRe is a leader here but it is too soon to say what the result of the company’s research will produce.”

[Ins.Cuddeford-Jones.2017.01.16]