Humans cause most self-driving car accidents

Since 2014, there have only been 34 reported accidents involving self-driving cars on California roads, according to state incident reports — and most happened when a human-driven car rear-ended or bumped into a self-driving car stopped at a red light or stop sign, or driving at low speed.

Why it matters: A major benefit to self-driving cars is the potential to reduce traffic accidents caused by human error. While it's a small set of data, the low rate of accidents caused by self-driving cars underscores the technology's enhanced safety.

Data: State of California Department of Motor Vehicles; Icon: Guilhem/The Noun Project; Chart: Andrew Witherspoon/Axios

But humans will continue to be a problem for the foreseeable future. A closer look at those accident reports reveals stark differences between how self-driving cars interpret the rules of the road and how humans behave behind the wheel. For example, human drivers make sudden lane changes or run red lights — not the way self-driving cars are taught to behave on the road. These awkward interactions between self-driving and human-driven cars will probably result in more fender-benders as more autonomous vehicles arrive on the roads.

The self-driving cars were at fault in only four incidents, and in autonomous mode in only one of those four. In six out of the 10 incidents in which the cars were in manual mode (with human drivers in control) at the time of the collision, the cars were previously in autonomous mode until drivers took over for safety reasons.

In context: It's important to keep in mind how long the cars are on the road. Waymo, for example, filed 13 accident reports in 2016, but its cars also drove 635,868 miles in autonomous mode during that period, or just about 1 for every 50,000 miles. To date, 36 companies have permits to test self-driving cars in California.

Data note: In making the chart above, Axios had to make assumptions in a few cases, including the car's general speed and which driver was at fault, due to information gaps in the accident reports.