Big auto and tech companies are confronting obstacles that hinder deployment of self-driving passenger vehicles, particularly how to handle liability when accidents happen.

Consider the conundrums that self-driving vehicles pose for insurance regarding who is at fault and how restitution is handled. Under the current system for dealing with accidents, investigators apportion blame by determining drivers' negligence or determining whether they failed to exercise care. Drivers are generally required to have insurance, and their carriers pay damages accordingly.

It's not a perfect system, but it works. But what happens when there's no driver?

The insurance industry lives and dies on actuarial statistics and currently lacks the data needed to make prudent decisions about insuring autonomous vehicles. How many accidents will there be? What will cause them? How big will the payouts be? And who will be held responsible? Without the right data, autonomous vehicles could face a chicken-and-egg problem: No data means no insurance, but no insurance stifles the potential to generate data.

A lack of insurance also represents a problem for getting autonomous vehicles fully on the road and widely available. Figuring out a solution might sound difficult, but insurance already deals with some of these problems and is actually getting better at it.

In fact, if you pay close attention to how AI and digital technologies are challenging an entrenched industry like insurance, you can see the outline of how AI and digital confronting other industries can work out.

Related: 7 steps forward in the mainstreaming of autonomous vehicles

|

Autonomous vehicles are like hurricanes

When it comes to coverage, the insurance industry is not new to big and abrupt changes. Case in point — catastrophic commercial insurance related to climate change. Scientists are pointing to an increase in violent and unpredictable weather, and past events are falling short as benchmarks to help prepare. Even the most detailed satellite images and simulations were never going to prepare people on the ground for what was coming.

And the problem isn't limited to hurricanes. There is a new and unpredictable “normal” for wildfires, as we've seen in California. Wildfire season is now lasting longer around the world, with the number of major wildfires in the Great Plains tripling in the last three years. Structures lost to wildfire in the American west has increased by 300%.

All of this is happening as the need for rigorous planning and agile systems have increased as more and more channels spout data. It's overwhelming even for the biggest companies with well-resourced risk management teams to determine what damage was caused by wind or flooding or was aided and abetted by poor property management.

|

An outline to success

Amid these changes to weather and information consumption, insurance is searching for solutions not just in terms of how to rework models, but how to get facts fast and streamline response. A key question is, how do you recalculate and address risk on the fly?

The answers give an outline to think about the unpredictability of autonomous cars, and the other changes artificial intelligence likely has in store for us. It starts on the edge where incident meets insurance policy. In the business this is called first notice of loss:

Get the facts right the first time. A strong contact center team is key, but not sufficient. The technology is available to make sure that the human interactions that are such a big part of novel incidents are more accurate, efficient and effective. Getting anything right starts with getting the facts right at the outset.

Move information fast and accurately.  Advanced dissemination and escalation engines are indispensable. Bad information spread over social media can exacerbate an incident. The same is true regarding conflicting accounts from onboard sensors, people and external sources of information (e.g., cameras and road sensors). Sometimes more data actually produces confusion, and it's important to triage and move it right at the beginning edge of the process into either human or algorithmic hands that can distinguish between signal and noise.

Things can happen all at once. One thing we've learned about the automation of fast-moving systems is that when something goes wrong, it can go wrong completely and quickly. Think about flash crashes on the stock market where robotic traders all start making the same move at the same time faster than humans can intervene. Whether it's AI driving cars, or in other areas of risk, insurance intake systems need to start fast and then keep up with a rush of information.

Companies can't always know what's coming their way, so they need systems that can be set up overnight. Being able to fight robots with robots on the front end, or at least robot-aided people at the start, can help. The past happens fast. Since we don't have past data to draw from on how autonomous vehicles will behave, we'll need to make the best use of the data as it comes in and be able to adjust with fluidity.

Insurance systems designed to plod through data, where a black swan may show up occasionally, will need to adjust and readjust as black swans come in flocks until trends settle out. Dynamic rules-based intake scripts are not only essential at the outset at the edge of the insurance system, when the claims first come in, they allow for an intake process capable of adjusting to changing circumstances.

Related: Top insurance technology issues nagging at industry leaders

|

Insurance can aid positive change

It's important at this point to note that insurance shouldn't be a barrier to autonomous vehicles and the beneficial side of artificial intelligence.

There can be little doubt self-driving vehicles have the potential to deliver enormous benefits. Many planners envision autonomous vehicles being deployed as an on-demand service — like Lyfts that arrive when you need them — rather than being owned by individuals. This would go a long way toward reducing urban congestion and solving parking problems. The public would be relieved of the burden of fueling and caring for a large hunk of steel in the driveway. Driveways and even garages might even someday be things of the past.

Most importantly, our current transportation scheme is causing a public safety crisis of epic proportions. There are 6 million traffic accidents in the U.S. each year, injuring 3 million people and killing 40,000 — the equivalent of two passenger jet crashes each week. Approximately 94% are caused by bad driver decisions. By eliminating drivers, self-driving cars could go a long way toward alleviating this crisis.

|

And then there's government

Advanced automation and digital technologies can make insurance more resilient and mitigate risk on the fly. But this works better when government is ready to sort out public policy. Currently, the government is trying to catch up with data privacy and find a framework to guard against risk.

One option for lawmakers is to put the blame on manufacturers, with accidents becoming product-liability cases. This would discourage innovation among the companies that we are relying on to solve the current highway-carnage crisis. Also, product-liability cases are notoriously messy and take years to solve.

You can imagine the legal rathole: The victim or their estate sues the big-pocketed manufacturer. The manufacturer — eager to avoid a judgement that could have cascading effects for other accidents — in turn sues the municipality that constructed the roads, or a parts supplier whose technology may have played a role. In this scenario, the lawyers do well, but compensation for victims is slow, and concerns over risk hinder progress.

A better option might be a no-fault type scheme in which potentially responsible parties —either passengers, owners or manufacturers — pay premiums that would cover damages without apportioning blame. There are complications to this approach, however: By socializing the cost of accidents, companies have less incentive to invest in safety. And, currently, regulations are not compatible with such a system.

|

Looking ahead at autonomous vehicles in the P&C world

For the moment, it's not a slam dunk that autonomous vehicles will achieve their potential. They could still flounder in the marketplace, like Google Glass, Segway or QR codes, or never make it past the research phase. Technology is a big part of the test — and it's still too early to tell whether the inevitable system imperfections will turn out to be bigger problems than driver error.

So far, four fatalities have been attributed to autonomous vehicles. Tesla, which leads the industry in vehicle miles driven, contends that its Autopilot feature reduces accidents per miles driven by 40%. Others dispute this, but for now, there's not enough data to draw meaningful conclusions. The technology is in its earliest phase, and the industry is confident that it can improve public safety. In fact, a recent survey by law firm Perkins Coie found that the industry believes that reducing traffic accidents is the greatest benefit that autonomous vehicles will deliver. Insurance and government can play a role, by catalyzing the infrastructure needed for smart cities.

According to that survey, executives believe liability, and with it, the insurance implications, are the top obstacles to bringing autonomous vehicles to market. In fact, the survey found that the industry is more optimistic that it can keep people safe and attract investment than it is about handling the financial fallout when accidents do occur.

But insurance has a way to handle this. As risk gets faster, our only choice is to get faster too, and fight data agility with data agility right at the event horizon where the first claims come in.

Haywood Marsh is General Manager of NetClaim, an insurance claims reporting and distribution management business. He can be reached at [email protected].