When Will We Trust Self-Driving Cars?

Autonomous vehicles (AVs) have moved from test tracks to public roads at a turbo-charged pace. Over the next two years as on-road pilot programs and testing increase, these vehicles are likely to go from novelty to reality, and more road users across the country will encounter them. If regulators establish guidelines that ensure AVs are identifiable, consistent in their actions during common on-road scenarios, and act to reduce risk when human error could contribute to a crash, human trust in AVs could flourish as all road users learn to accept and manage the added risk from maneuvering alongside new technologies that may fail. However, AAA’s annual automated vehicle survey reveals the public remains fundamentally anxious about the safety of self-driving cars. Their fears are not unfounded, and policymakers should take note. 

According to the results of the 2021 survey, which was conducted in collaboration with the Technology and Public Purpose (TAPP) Project at Harvard Kennedy School’s Belfer Center for Science and International Affairs, nearly half of US drivers report they would feel less safe sharing the roads with self-driving cars, whether major freeways (47%) or local streets (44%). Only one-third (34%) of drivers favor testing AVs on public thoroughfares. The rest — opposed or unsure — ranked concern about the safety of sharing the road with self-driving vehicles as their top concern, along with issues regarding who would be responsible if they had an accident with a self-driving test vehicle. 

Following the 3-year anniversary of Elaine Herzberg’s death in March 2018, when she became the first known pedestrian to be struck and killed by an AV, policymakers should be reminded that the public’s trust in these vehicles is fragile. A AAA survey — conducted before the tragic 2018 Uber AV test vehicle crash in Tempe, AZ — found that Americans were beginning to warm up to AVs, with 63% saying they feared riding in a self-driving vehicle, a considerable drop from the year before when that number was 78%. The same survey conducted a month following the crash found that fear had jumped back up, with 73% saying they would fear riding in a self-driving vehicle. 

Upon investigating the Uber crash, the National Transportation Safety Board criticized state and federal safety officials for their hands-off approach to testing on public roads. Perhaps it is time they consider a new approach. Based on the research findings, the public needs four elements of acceptance addressed in order to trust AVs: 

• AVs demonstrate safe and predictable behavior, as defined by regulators and the public, on roads and highways; 

• AV technology will operate as designed reliably; 

• Road users understand who would be responsible in the event of a crash with a self-driving vehicle; and 

• Humans can identify when a vehicle in operation is an AV. 

If public testing is meant to help AV developers design more reliable vehicles to establish trust in the technology, regulators should use these elements to provide adequate oversight of AV testing, ensuring that real-world failures will result in improved software that governs AV behavior. The sheer complexities of navigating busy streets make additional fatal crashes all but inevitable as AVs rack up more miles driven. This means that when Americans have questions about the value of testing AVs on public roads after crashes occur, policymakers like US Transportation Secretary Pete Buttigieg should be able to explain the purpose of public AV testing. The four elements of public acceptance are a great place to start when public conversations occur. 

Over the coming years, as AVs grow in number across the nation’s transportation network, the capabilities of these vehicles should be shaped by the expectations of road users who have established behaviors that self-driving cars need to master in order to gain the trust of humans. Until AVs are proven to be reliable and safe, regulators will have to bridge the gap and ensure that both the hardware and software can achieve expected road behavior. There may be a wide spectrum of behaviors believed to be necessary for AVs to demonstrate, and regulators should encourage that discussion. Buy-in from all road users on what constitutes the driving behaviors for these cars to emulate will help establish the public’s baseline understanding of how the vehicles operate, giving all road users a place to start when assessing how to interact with, instead of fear, a self-driving car. Only after the average road user knows what to expect from a self-driving vehicle will they truly trust the decisions made by it.