Autonomous Driving Systems (ADS, aka driverless) are continuing their slow march toward adoption. Just last month, the National Highway Traffic Safety Administration released their long awaited proposed voluntary national framework for the evaluation and oversight of vehicles equipped with ADS. The program lays out a two-step program to improve industry transparency and assist in the evaluation of the safety of this technology. While federal leadership on the issue of ADS safety is welcomed, their slow and dogmatic approach could leave us following, not leading, the world in the development of ADS.

Much has been made of the potential safety benefits of ADS. Sadly, much more has been made about the potential safety impacts of ADS mistakes. The implicit messages have been that if an ADS system makes one mistake or is involved in one crash then the technology is an unreliable danger to the motoring public. While the goal of zero deaths is laudable, year after year we’re 40,000+ fatalities away from this goal. For those of us that have committed our careers to improving highway safety, fairly evaluating the safety improvement potential of new programs and technologies by comparing them to the status quo is critical to understanding the practicality of deployment.

That’s why a recent study sponsored by insurer Swiss Re and Waymo and conducted by researchers from MIT and Stanford researchers piqued our interest. The study compared property damage and bodily injury claims of Waymo driverless cars with those of insureds operating in the same zip codes.

The study analyzed 25.3 million fully autonomous miles (no driver in the driver’s seat) with that of 40 billion miles of exposure traveled by human insureds. The study found that driverless vehicles experienced an 88% reduction in property damage claims and a 92% reduction in bodily injury claims.

Of course, the technical details of Waymo’s system likely differ in significant ways to that of its competitors so it would be improper to attribute the same magnitude of safety improvement to all vehicles. That said, it’s clear the potential is real.

Unfortunately, ADS development is often met with skepticism and adverse events are pointed to as evidence that the technology is a public danger. It’s the reptile theory applied in the public square, where each incident is misattributed to some kind of systemic problem. This ignores important context and forgoes potentially massive safety improvements.

We often measure safety gains in terms of incidents avoided and lives saved. In the evaluation of ADS, we should do the same. This will help us avoid missing the forest for the trees by avoiding myopic thinking and zooming out to see the potential for massive safety gains.