Backseat driving in the time of autonomous vehicles

Netstar

At some stage in the last decade cars have learned how to drive autonomously, meaning almost completely by themselves. In the future, there will be no need for a driver’s seat, and further down the line, little need for insurance policies as we know it.

Automated vehicles are deeply unnerving, likely due to them being such an obvious and relatable example of robots infiltrating what has always been the domain of humans.

Traditionally, a human shifts behind the steering wheel of a car pushes down on the accelerator with their foot, and the vehicle moves off in a certain direction. A few years ago, a car warning a driver of an impending collision with, say, a dustbin was reason enough for jaws hitting the floor. Then cars started parking themselves, some ghost in the machine handling the wheel… Worrying to some, cars have come a long way since then.

The Society of Automotive Engineers (SAE) created specific definitions describing advanced driver assistance system (ADAS) functions, rating these on a scale of 0 to 5. Levels 0 through 3 being where the human driver is still very much in control of the car’s movements. Levels 4 and 5 describe a system in which “You are not driving when these automated features are engaged – even if you are seated in the driver’s seat,” according to the SAE J3016 guidelines.

Features such as automatic emergency braking, blind spot warning, lane centering and adaptive cruise control fall under Levels 1-3, where Levels 4-5 include functions such as traffic jam chauffeur, local driverless taxis, and pedals/steering wheel that may or may not be installed. Defining how the system should work is the easy part, but at the moment these are mere definitions.

The SAE’s trademarked rating system is far from perfect, as noted in an article in Popular Science. “This has the implication that the higher the level, the better, or more advanced the automated system is,” Mahmood Hikmet, head of research and development at New Zealand automated shuttle company Ohmio, is quoted in the article. “But that’s not true. The only thing SAE J3016 levels tell you about a system is about the responsibilities of the human or the automated system while executing the driving task,” he says. “That’s it. Nothing about the operational design domain, capabilities of the system, or how advanced the systems are. Just what the human and the automated system are respectively responsible for doing during operation.”

CNN journalist Michael Ballaban recently tested a Tesla installed with “full self-driving” beta software. Ballaban, being chauffeured, and driving, around a busy street in Brooklyn, describes the car as “technically, sort of driving itself.” In one instance, he wrestles back control of the wheel from the computer as a motorcycle driver cuts him off and the car veers, wants to, at least, straight into a delivery truck in oncoming traffic. It’s the first of many nervous moments during the test. Driving down a straight road the car seems to be doing fine but things get hairy in challenging situations – “stuff that experience teaches you,” he says. “It’s a little like teaching a teenager how to drive. You’re always watching, always waiting. You never know when it might try something new.”

Teaching computers what is tantamount to human experience is an ethical and mathematical minefield. Consider the following thought experiment: An automated vehicle is cruising along the highway and must contend with an animal directly in its path. Does it swerve to save the animal and put the passengers in danger? What if it were a human in its way? What if there were more humans in its path than passengers in the vehicle? And what if it were children in the road and adults in the vehicle? It’s these types of ethical dilemmas that have, for the most part, kept fully autonomous vehicles in the labs and off the roads. As Ballan’s Tesla test shows, the so-called fully self-driving cars today tend to deal with potential collisions by steering straight into another one. It’s far from ready for public use.

Self-driving vehicle makers Waymo and Navya already operate and sell Level 4 vehicles that operate in the US. Waymo’s autonomous taxi service, Waymo One, has been delivering services in Arizona since the end of 2018, and was recently introduced in San Francisco on a trial basis.

Most people agree that self-driving vehicles will happen but differ on when we’ll see the technology scale to global use. When it does, however, it will have huge, massive, metamorphic implications for various industries, including telematics and insurance. The former industry being positioned better in the market than the latter…

Nelson Mills, writing for Forbes, explains that driver liability will likely turn into product liability when computers are fully in charge of driving. “The vehicle OEM or whoever supplied the automation system and/or its components may be liable,” he says. “While portions of the liability pool shift from personal to commercial in nature (i.e., product liability versus personal auto) the overall size of the risk may not change dramatically. Rather, insurers will need to understand the risk posed by two drivers; one human and one robotic,” he explains.

Deloitte, in its report Autonomous vehicles and the insurance industry, says insurers can prepare for the self-driving evolution by:

• Recognising that the rate of change by geography and age may vary
• Developing more technical underwriting capabilities
• Preparing for incremental changes to cost structures
• Navigating with insufficient or incomplete data, and exploit emerging sources
• Establishing advanced analytics capabilities
• Planning for product and business-line shifts—including offering driverless car insurance
• Retraining claims adjusters to interpret intricacies of shared driving
• Recognising the threat of non-traditional competitors

“With sensor-loaded cars poised to reduce accidents by 90 percent, and ridesharing/ride-hailing trends pointing toward decreased vehicle ownership, auto insurers could be challenged to compensate for the apparent inevitability of falling premium rates and perhaps a substantial volume of business,” the Deloitte report notes. “As a result, insurers should consider developing transformational strategies to remain relevant and create value for consumers, yet still be profitable in this emerging environment.”

Clifford de Wit, Chief Technology Officer at Netstar, a subsidiary of Altron, says the technology is at a point where autonomous driving is already a reality. For telematics companies, the roadmap may be clearer than for insurers, as telematics sensors and related technologies function as the neural network of these vehicles, telling them where to go and how to behave based on mapping technology.

“Vehicle manufacturers have a vested interest in providing the autonomous driving experience,” says De Wit. “Telematics providers are embedded into their manufacturing stack.” To roll out autonomous driving requires an inordinate amount of data for the mapping of roads and road conditions, but also in keeping track of changing environments in particular areas. Without telematics, autonomous vehicles would be driving blind. It’s one of the main reasons why developed nations have been at the forefront of mapping their environments, explains De Wit.

He says that telematics providers have a very good idea of what’s happening on the roads, and how drivers behave in these environments. “As an industry, the telematics providers play a key role in providing the insights that will enable autonomous vehicles to drive on our roads,” he says. For telematics companies, there’s an obvious opportunity to partner with autonomous vehicle providers in the future. “As an industry, we localise well,” he says.

BDO listed the following considerations for insurers, which also applies to brokers and their clients:

• Which levels of automated vehicles is the insurer prepared to cover?
• If humans have less and less control behind the wheel, how will the insurer determine premiums? What new or additional risks will need to be insured?
• Where does the insurance company fit in the new ecosystem of ride-sharing, considering the ultimate buyer of insurance products will change over time?
• How will the insurer adjust its underwriting and pricing models as the current model of driver-centric insurance products eventually ceases to exist? What impact will this have on the company’s existing distribution channels?
• Is the insurer’s systems and technologies equipped to handle the new autonomous vehicle ecosystem?
• How will the new automated technologies affect the cost of claims including frequency and severity of losses over time?
• If a self-driving car crashes, who will be found liable for damages? What if the crash is caused by a weather event, or the self-driven car itself?

Netstar has an OEM relationship with one of the global leaders in auto manufacturing in providing telematics devices in each vehicle that rolls off the assembly line. It enables the vehicle manufacturer to globalise down to an individual country level by using Netstar’s local knowledge of dialing into the telecommunications infrastructure. “There’s going to be a point where these partnerships involve the incorporation of autonomous driving features,” De Wit says. Globally, this is already happening, but not yet in South Africa.

So, what will happen to the insurance industry if computers take over as drivers? “At some point, we are going to have humans and autonomous vehicles sharing the roads, and we must think about what that means for insurance and how telematics data can help insurers figure out who’s at fault in the event of an accident,” he says. This is widely in use today.

Back in 2016, KPMG, in its Automobile insurance in the age of the autonomous vehicle report, noted that a decline in accidents will drive a drop in industry loss costs and, subsequently, premiums, with a fall starting when cars start converting to self-driving models. “The mix of insurance will also likely change, as commercial and product liability lines expand. Within 25 years, our models suggest a scenario where the personal automobile insurance sector is less than 40% of its current size.” While the predictions may have a few years, decades even, added to the timeframe, given the slower-than-expected rate of development, the authors warned that the elimination of excess underwriting capacity could have a severe impact on the market, further exacerbated by changing business models and new competitors. “No-one has a crystal ball, but we are convinced that a period of unprecedented change has begun,” the authors warn.

Data management will be critical for insurers, seeing that autonomous driving requires and generates data, which will only increase as the internet of things (IoT), and the “web of information”, increases between the road ecosystems. “In this environment, data management – integrity, storage, analytics, and security – becomes critical,” according to Deloitte.

But this new burgeoning industry can also hold opportunities for insurers, says De Wit. With electronic vehicles, there’s an opportunity to insure various expensive parts such as the battery. In terms of driving risk, everybody’s asking who would be liable for an accident involving these vehicles. There are ongoing legal trials today in which OEMs have clearly stated the driver behind the wheel is the liable party, even if the vehicle was driven autonomously. From an insurance perspective, the onus still rests with the driver. But how does the insurance industry approach a situation in which the client is liable for the risk but it’s the vehicle making the decisions? “It’s an issue that some very smart actuarial scientists will have to consider.”

De Wit says it will be an autonomous vehicle evolution and not an all-out revolution, given the speed at which change is happening in the mature markets. This transition will be multifaceted, as noted in the report by Deloitte. The authors posit four future states of mobility, including personally owned driver-driven vehicles, shared-driver driven vehicles, personally owned autonomous vehicles, and shared autonomous vehicles. These are expected to evolve at various rates and will share the roads for the foreseeable future. The success of new technologies, regulatory uncertainty, and consumer attitudes about giving over control to self-driving vehicles are the factors that will result in uneven adoption. “Our research indicates that the pace of these trends may vary among age groups and geographies, but compelling economic and societal benefits suggest that the future will increasingly be defined by shared self-driving vehicles,” the authors note.

Insurers thus have time to prepare. “The better insurers can digitise their business, becoming more agile in doing so, the better prepared they will be,” De Wit explains. “The best way to prepare for the eventual digitisation of the vehicle industry is to digitise the insurance business.”