This piece was published in the October 2018 American Planning Association Northern News. Here I was considering the response to the effect of autonomous vehicles on the pedestrian environment.
Photo by the author, captioned: Pedestrians have close encounters with vehicles all the time. How will that change as AVs become more common?
Autonomous vehicles (AVs) have a pretty good safety record already, and we can reasonably expect that they’ll be more reliable than cars driven by humans. AVs will not become distracted, sleepy, bored, angry, or intoxicated. Their sensors will see in all directions and their reflexes will be fast. When they tailgate, it will be called platooning and it will save space on the road and energy.
So it was a jolt when an AV being tested in Arizona struck a pedestrian this year. We could rationalize it as an indication of immature or flawed technology or manufacturing, like a bolt that snaps and brings down a bridge. But that was not entirely the case. Maybe more surprising than the crash was that the car saw the woman before it killed her.
The AVs are coming
Members of the planning and urban design profession are thinking about the many potential effects of autonomous vehicles on our cities — positive and negative, large and small. AVs might induce sprawl, reduce the need for parking, exacerbate air pollution, create congestion, reduce transit usage, and impact equity.
That said, amidst an iPhone-like technological optimism and while occupied with the other continuing demands on our professional attention, we are largely letting the technology firms and car companies drive us toward the looming AV future. For most of us, AVs will be the first physical robots with which we interact, and we really don’t know what to expect. It’s common to imagine being inside an AV, watching movies, catching up on our reading, eating, or sleeping. Very little is being suggested about what it will be like walking or cycling, facing an AV at the crosswalk. As it’s coming toward you, is the AV seeing you (think Arizona)? What calculations are being made in its electronic brain?
Why did the AV hit a person it saw?
Imagine the process of a computer driving a car as being similar to a smart phone’s autocorrect function rather than to a calculator solving an equation. The AV is reacting to conditions on the fly and with imperfect information. To prevent the car from stopping at every drifting shopping bag, it is programmed to ignore objects that have a lower probability of being human — which leads us to Arizona (https://bit.ly/2wn3175). As the technology improves, misidentification will happen less frequently, but there will always be uncertainty. And safety will never be the only concern for AVs, any more than it is for human drivers, who may speed in a school zone because they’re late for a meeting. Will there be something to stop a ridehailing company from dialing down the safety to trim a few seconds from each ride if it can save money and raise the share price?
Is there a correct response to AVs?
In my work, we are considering the design issues posed by AVs at the neighborhood, city, and regional scale (Chicago example, https://bit.ly/2oa0FkW). But no matter how the city is designed, if AVs operate carelessly, aggressively, or unpredictably, they will diminish our public spaces: Pedestrians fought it out with cars and trucks on the streets of the early 1900s. By the late 1920s, with the invention and outlawing of jaywalking, the cars had won (https://bit.ly/2FAuSkQ).
Unsurprisingly, we’re in the early rounds of a similar battle as technologists call for the control of pedestrians to meet the needs of AVs (https://bloom.bg/2LGVAdl; https://bit.ly/2onaKeC). This time we need to start with a set of rules — something that works for everyone and establishes how robot drivers must behave on our streets — if we are to protect both our sense of safety and our actual safety. Traffic laws may punish lawbreaking, as we saw this year in San Francisco when an AV was ticketed for allegedly failing to respect a pedestrian’s right of way (https://bit.ly/2ws7X7m). But we need something more foundational, more akin to Asimov’s Laws of Robotics about the relationship of robots with humans (http://bit.ly/2LewZMB).
I suggest these five principles as a starting point:
An autonomous vehicle must conform its behavior to the safety, comfort, and expectations of people outside the vehicle.
Humans must be made aware when a vehicle is under autonomous control.
Before it may move at any speed, an autonomous vehicle must be a minimum of five feet from any outside human.
An autonomous vehicle must signal its intentions to people outside the vehicle but must not command them in any way.
A non-occupant must be able to control an autonomous vehicle, at a minimum to cause it to stop.
(For background on this list, see my article in The Urbanist [Seattle], https://bit.ly/2woMqMV.)
Etiquette for robots
Whether or not you agree with these particular rules, we need standards for AVs beyond the laws that now apply to vehicular movement and traffic safety. If every AV manufacturer or operator has its own rules and its own expectations of pedestrian and bicycle behavior, we on the street will never know what to expect. We do not want a world where we need to know what brand of AV is approaching to know whether it’s safe to cross the road.
Generally, the federal government regulates vehicle safety and the states register vehicles and license drivers. As both the vehicle and driver, the AV can fall through the cracks. The current federal administration has taken a hands-off stance. According to the U.S. Department of Transportation, “the Department’s preference is for regulations that are non-prescriptive, performance-based, and seek to enhance safety whenever possible” (emphasis in the DOT original, https://bit.ly/2oj9LvQ). Is “whenever possible” good enough? Given the locations of many of the companies involved and much of the testing, it seems that the essential work will happen at the state and city levels, perhaps specifically in California and the Bay Area. We will need city officials, traffic engineers, pedestrian and bicycling activists, health experts, psychologists, and equity advocates to assist in setting the rules. This is not because of the ethical concerns about tech companies acting in their own interests, but because they have different goals than do the many important groups in society, and those groups need to be at the table. We need a public conversation about the rules on our near-future streets, and we need an entity — one with the ability to make the rules — to convene that conversation.
Early efforts to form cooperative relationships with AV companies have had mixed results (See CityLab, https://bit.ly/2Fj3OGg). Though there are efforts underway like the Autonomous Vehicles Perspective Paper by MTC and ABAG (https://bit.ly/2LzoAU4) that seek to address AV issues, the focus is too broad to address the fundamentals of behavior and safety. A fragmented local response could well lead to federal preemption that, in turn, may serve the corporations more than the most vulnerable users of our streets.
A future history
With AVs running in the streets, will the planners of 2070 regret our inaction? Will we be like the city builders and officials who enabled the proliferation of automobiles in the early 20th Century but failed to see how the auto would diminish our cities, our environment, our health, and our public spaces? (See http://bit.ly/2BzbuXo) Or will the denizens of 2070 congratulate us on our foresight? Let’s not wait for more tragedies like Arizona. Let’s not wait to work out the terms of our relationship to AVs after they’re ubiquitous. Let’s find a way to come together and develop a structure for this important relationship among humans, streets, and AVs.
