Published on December 14, 2023
The rise of self-driving cars has been a hot topic recently, with many companies investing in the technology. While some see self-driving cars as a solution to traffic congestion and a safer alternative to human drivers, others are skeptical of their effectiveness and the potential impact on employment.
In this article, we delve into the experience of Prof. Dr. Markus Hackenfort, Professor of Human Factors Psychology at the ZHAW School of Applied Psychology, who spent several weeks in San Francisco over the past months. During his time here, he gained firsthand experience with the reality of self-driving cars in the city and combined it with his research perspective.
He also examines the potential benefits and drawbacks of self-driving cars and how they might shape the future of urban transportation.
Last year, I had a watershed moment. I attended a traffic psychology conference in Gothenburg for three days, where one major topic was self-driving cars. Sessions about how they should be connected and designed, the intentions of traffic participants and their acceptance of automated vehicles, pedestrian-vehicle communication, and much more revealed the importance of this subject. Among other methods, researchers intended to shed light on human behavior in these self-driving cars by using sophisticated Wizard of Oz experiments – where participants are under the illusion of sitting in a driverless car.
Then I came to San Francisco, and with that, my watershed moment. It was like, suddenly, all the future we were discussing before became a reality – but in a far more pragmatic way than many seemed to have anticipated.
Of course, I had heard about robotaxis in San Francisco before. Every now and then, I encountered people in Europe with more or less abstract knowledge about the technical development of something extraordinary like self-driving cars. But I didn’t expect an army of free-floating, casually looking cars on the city roads with many additional sensors but no driver.
Of course, we do have a notion of self-driving cars in Switzerland, based on experiments in Sion or Zug – but not only did some of them end up early in museums, they had to be accompanied by safety drivers, hardly exceeding the speed limit of 12 miles per hour, and were sometimes hopelessly incapacitated by trash bags, for example.
Compared to these experiences, the picture in the Bay Area was quite different. But not because the streets in San Francisco are more accessible and easier for them to drive on. Self-driving cars learned quickly how to ride in nearly the entire city area, up and down the hills and spanning an area roughly the size of 64 square miles. Driverless cars seem to be swift learners. Cruise robotaxis, for example, quickly changed their behavior when news stories came out that they were driving over unsecured fire hoses in the streets, to the anger of firefighters working on the scene. After a software update, they stopped and waited until they were cleared to continue.
The Future is Now: How Self-driving Cars Shape the Urban Future
During a hearing in San Francisco on August 10, 2023, about a full-day charge fare permit for robotaxis, many advocates underlined the prospect of mobility for everyone that would come with robotaxis. Today, very few people use a cab to get from their home to a bus station. But some people would likely use robotaxis for those short trips, which would improve access to public transportation since, unlike cab or Uber drivers, a robotaxi doesn’t care if a trip is short.
Those who feel uncomfortable walking the streets at night or by bike may also embrace the option to take a robotaxis to get home from their bus or train station. In terms of security, a robotaxi seems like a very safe option. But also in terms of accidents, robotaxis were touted as doing much better than humans because they always abide by the rules and are never tired or distracted.
The crash rate benchmark that Cruise and the University of Michigan Transportation Research Institute, and General Motors developed for ride-hailing cars in San Francisco supports these claims. According to this data, Cruise’s self-driving cars reported 65% fewer accidents overall and even 94% fewer collisions, where Cruise’s self-driving car was found to be the leading cause.
But it is worth knowing that this safety record is based on just one million miles. Accidents are considered statistically rare, making these findings even more preliminary. This was tragically documented by an accident on October 2, 2023 in San Francisco when a human-driven car hit a woman who was crossing a street. She was then hit by a driverless Cruise car that stopped immediately and probably even faster than any human-controlled vehicle ever could.
Although the robotaxi stopped immediately, the woman was hit. But the severity of the injuries she suffered might have been worsened by the driverless car after it stopped since it started to drive again, dragging the woman for about 20 feet while she was pinned underneath. It is suspected that the driverless car started moving again due to its “Collision Detection Subsystem,” which is supposed to help traffic since it tells the car to move out of the way. Cruise later admitted to a “Software Flaw” that could have led the driverless car to move again to get out of traffic.
The way the situation was handled by Cruise – especially concerning an alleged accusation of preventing immediate full insight into what was happening – hit the self-driving vehicle concept right at its core. Public trust was damaged and, in an immediate reaction, led to an initially enforced, later voluntarily extended suspension of unaccompanied rides by Cruise.
That also demonstrates that it’s not just about pure objective accident rates that concern the public. I have learned that human drivers consider themselves safer and better than others while assuming to have fewer accidents than their peers – even if the opposite is true. As long as we think we can handle the situation by bicycle or car, we often feel much safer than we sometimes are. Conversely, in some cases, we are worried about our safety, although nothing suggests an unsafe condition.
Both sides of this continuum are known by psychologists, from the “illusion of control” to the fear of flying due to not having any control of the situation. Applied to the present context, it remains to be seen if passengers in robotaxis really experience trust and a more pronounced sense of safety given the higher level of safety indicated. The principle of machine learning implies that unforeseeable events with driverless cars will occur for a long time and billions of miles. These incidents are rare and certainly less frequent than among human drivers, but they will appear, and the more the technology and the miles driven increase, the more they will happen. The crucial questions are: how safe is safe enough for us, and on which variables do we base this?
Another question is how humans will embrace robotaxis in traffic. Resources with limited availability – as public space on streets is – evoke competitive behavior in humans; countless psychological experiments have proven that in different contexts. Humans might consider their machine counterpart a contestant, even more so as it is programmed to be law-abiding, reserved, friendly, defensive, and compliant.
Are we Prepared? Governance is Now
Not long before the incident on October 2, Cruise announced its expansion course over the coming months, mapping and deploying its services in 16 cities. In San Francisco alone, at some point, more than 500 robotaxis roamed the streets owned by just two companies. According to California’s Department of Motor Vehicles, as of November 1, 2023, as many as 39 companies own an Autonomous Vehicle Testing Permit (DMV CA, 2023).
Even though one company is cutting back its capacities significantly at the moment – if any of these companies deployed only 200 driverless vehicles each in San Francisco, 7,600 cars would float the streets, which is far more than the number of e-scooters on the streets (SFMTA, 2023). The impact on the environment and life in cities will be huge – regarding a possible change in mobility behavior away from public transportation to more individualized mobility or in terms of how scattered the living and housing will be in the distant future.
Society and their governments must decide how many robotaxis are acceptable and under which condition – shared, pooled, or individually used. Even if currently halted, in the works is the Cruise Origin, an easily accessible minibus, which opens the competition with public transport. On the one hand, it could mean a more flexible transportation system that works around the clock in every corner of a city and could operate considerably cheaper because of the absent driver. On the other hand, that could massively change how we know public transportation and how it works. However, the biggest threat to the environment would be a shift to smaller-scale transportation with more vehicles on the road than today.
All these changes have taken on a breathtaking, likely more than exponential speed. We need governance that recognizes the signs of the times, anticipates the latest developments, and integrates them as a complement – and not unilaterally as a threat to what already exists. That could mean, for instance, combining the technology into the existing public transportation systems through public-private partnerships.
The Bottom Line
Robotaxis are on our doorsteps. Despite the recent setbacks, this technology will continue to develop with great strides – with or without regulation. The benefits of driverless cars in our cities are too significant to ignore. How humans deal with them both as fellow traffic participants and service providers is completely uncertain, but it determines the technology’s success or failure and its environmental impact.
Research – itself facing the challenge of keeping up with such rapid changes – should be able to accompany this process closely and by considering the reality as it is, like currently in San Francisco. Politicians and administrators, in turn, would be well advised to keep a very close eye on the developments, which are taking place in faster and larger steps, and to set the course for their society as early as possible.
Prof. Dr. Markus Hackenfort is a Professor for Human Factors Psychology at the University of Applied Sciences in Zurich (ZHAW), and has spent several weeks in San Francisco over the past months on a research sabbatical.
Hackenfort and his team investigate human behavior with a focus on risk perception and prevention in real traffic and with the help of both a 360°-bicycle and a dynamic car simulator.