Looking Harder at the Arizona AV-Pedestrian Fatality

Note: This is a guest post from Robbie Webber, a Senior Associate at the State Smart Transportation Initiative, based at the UW-Madison. She is a former Bike Fed staff person and board member. She is now on the board of Madison Bikes. This blog is based on the best info available as of posting.

You will probably see reports and discussions about the fact that a pedestrian walking a bicycle died when hit by an autonomous vehicle (AV) in Tempe, Arizona. But maybe we are concentrating on the wrong design problem. Here’s the money quote from Slate:

“You can expect that Uber, local regulators, and tech evangelists will make much of the Tempe police report that the woman was outside a crosswalk, although North Mill Avenue—the eight-lane road the victim was attempting to cross—has only one crosswalk in nearly two miles of road, making jaywalking a requirement of the urban design.” (emphasis added)

Robbie Webber

I looked at a satellite view of the location where she was killed, and even before reading the Slate article, that was my first thought: How are you supposed to cross this street safely, regardless of who is behind the wheel?

From a Google Streetview of the area where the woman was killed (not sure of the exact location, but the general area as indicated in media reports), I found two things of note: 1. That paved area sure looks like a place where they are inviting people to cross the median of this divided highway. 2. Note the “No Ped Crossing” sign, as if they KNOW people might want to cross there. (To clarify, if the media reports are correct, she was probably 300-600 feet south of a crosswalk, which is less than a mile, but still a long walk at 10 pm, and a ridiculous distance in the summer in Phoenix.)

I always think that, if you have to put up a sign telling people not to cross somewhere, maybe you should think about making that a place where crossing is easy and logical, not making them take a long hike on a hot Arizona day to just get across the street.


For anyone who knows this area, it is just north of a river and a freeway, and a pretty un-friendly area for anyone not using a car. At the same time, it is a short physical distance from the campus of Arizona State University. The street she was crossing turns into the bar/restaurant/entertainment strip for the campus/downtown less than a mile away and is quite pedestrian friendly. But what a difference that one mile makes in safety for human-powered travelers!

The human “safety driver” said that the first indication that someone was in the street was when the car hit her. We don’t know if a human driver could have avoided this crash or not, so I am hesitant to blame AV technology. After all, humans don’t have great safety records when it comes to operating around pedestrians and bicyclists.

However, this still doesn’t address the general lack of pedestrian (and bicycle) facilities in the area. The road and whole area is built for fast travel by motorists, and it is very hard to cross, even at actual crosswalks — which are widely separated. It’s just not a good place to be a human unprotected by a metal box, and that’s not good for humans in general.

Maybe it’s not about the automated vehicle, but about the city in general. The Phoenix/Mesa/Scottsdale metro area, is not safe for people walking. Just last week the Arizona Republic ran an article highlighting the terrible record of the metro area when it comes to pedestrian safety.

“In the past week, there have been 10 pedestrian traffic fatalities in the Phoenix metro area reported by The Arizona Republic.

“Two more people were left in critical condition during that period.”

Robbie Webber is a Senior Associate at the State Smart Transportation Initiative, based at the UW-Madison. She is a former Bike Fed staff person and board member. She is now on the board of Madison Bikes.

2 thoughts on “Looking Harder at the Arizona AV-Pedestrian Fatality

  1. After reading some recent articles about the crash (with information released since Robbie’s blog post), I think it is a combination of streets not designed for pedestrians and bicyclists AND serious deficiencies in AVs current technology/learning. This arsTechnica article makes the case that Uber’s AVs are particularly bad at the things humans learn to do to avoid crashes with other vehicles, pedestrians, and bicyclists who may be acting erratically or not following the law. If streets are not built for pedestrians and bicyclists (we know they aren’t), then those users are probably less likely to follow road rules. Therefore, Uber AVs, as they currently stand and combined with the way we build roads, may be particularly dangerous towards pedestrians and bicyclists.

    Action steps, of course, as always, include Complete Streets. But I think we should also push for regulations and better data collection on the development of AVs to ensure that they are as safe or safer than human-driven cars. Currently, Uber is pushing for fewer regulations so that they can compete with more advanced AV companies in Europe (see this WSJ article and this CNET article for more details on that). This is unacceptable when peoples’ lives are at stake.

  2. People’s opinions have been rather divided: many seem focused on the pedestrians actions (whether she “caused” the crash, etc.), while others treat this as a referendum on any and all auto-pilot vehicles. While the circumstances seem unique in this case, I contend that an unexpected object in the roadway is a rather common occurrence: persons, deer, or otherwise. Likewise the technology of self-driving cars is not going away. What concerns me most is the failure of the vehicle to engage the brakes and alert its occupant in any way. In almost every other industry that uses computer automation and puts human life at stake, there are redundancies and fail-safes built-in to the designs. The current state of automobiles (as evidenced by the long list of recall notices) does not have adequate fail-safes and redundancy in their design to handle these situations with a high degree of reliability. We’ve long accepted a lack of mechanical redundancy in our vehicles in the name of making them more efficient and less expensive, despite the occasional tow back to the shop. We also give a pass to human negligence and poor judgment when collisions occur. But when setting the standard for computer-piloted vehicles, we ought not accept the status quo. If the cost-savings are so compelling, then why not invest in more reliable technology from the start?!

Leave a Reply

Your email address will not be published. Required fields are marked *