There is a persistent hum of confusion whenever people talk about autonomous vehicles. Are they actually safer, or is the data just dressed up to look better? According to the latest findings on Misryoum, we’ve got a clearer picture, though it’s not as simple as checking a box. The benchmarks, pulled from recent studies like Scanlon et al. (2024) and the follow-up work from Kusano et al. (2025), focus on state police reports and total miles driven. We’re talking Phoenix, San Francisco, Los Angeles, and Austin—major hubs where Waymo is actually out there doing the work.
Here is the thing about comparing robot drivers to humans: you have to compare apples to apples. The researchers had to scrub the data to only include passenger vehicles on road types where Waymo actually operates, which means skipping freeways entirely. They also applied a 32% underreporting correction for minor injuries—because, let’s be honest, humans don’t report every single fender bender to the police. Actually, they often don’t report most of them at all.
But wait, what about the tricky streets? Not every road is created equal. If Waymo sticks to the tougher, high-traffic corridors while the average human driver might be cruising through quiet suburbs, a direct comparison breaks down. That’s why Chen et al. (2024) came up with a spatial reweighting method. It adjusts the benchmarks to match where Waymo is driving most often.
It’s a complicated dance of numbers, really.
These adjustments, part of what they’re calling the RAVE—or Retrospective Automated Vehicle Evaluation—best practices, aim to align the datasets more tightly. It’s an attempt to stop us from comparing city-wide averages to specific, localized driving behavior. It makes the data more representative, though I suppose we’ll see if it holds up as they expand into more complex areas of the city. Or maybe it’s just the best we can do for now with the records we have.
Ultimately, the goal is transparency, even if the math makes your head spin a little. It’s not just about saying one is better than the other; it’s about making sure we’re looking at the same reality. I caught a faint smell of burnt ozone walking past one of the sensor-laden vans in Phoenix last week—a reminder that this isn’t just theory anymore. It’s happening, and the data is just trying to catch up to the road.