
Episode #16
Autonomous Driving: "Safer Than Humans" Is the Wrong Story - Phil Koopman - S2E16
In this episode of Automotive Storytelling, the only podcast examining mobility through storytelling, Stéphane Lagresle sits down with Philip Koopman, Carnegie Mellon Faculty Emeritus and originator of the UL 4600 autonomous safety standard, to unpack why the industry's robotaxi safety narrative keeps breaking and what it would take to rebuild public confidence. The AV industry has built its case on one metric: fewer collisions. Phil argues that the metric is doing very little work, and that the gap between what companies claim and what the public sees is widening rather than closing. KEY INSIGHTS A NARRATIVE REWRITTEN TWICE, AND BREAKING A THIRD TIME: Autonomous driving began as a way to increase highway capacity. It became "computers don't drive drunk, so they will save lives." Today it is "the statistics prove we are safer." Phil traces each rewrite back to the moment the previous story collided with reality. STATISTICS LOSE TO VIDEO: Safety is decided by extremely rare events, often called "edge cases". While the public sees a robotaxi drive into floodwater or over lit fireworks, companies talk about millions of miles driven. SAFE IS A BETTER GOAL THAN SAFER: Robotaxi companies sell on being safer than human drivers. Phil makes the case for a different bar, closer to the UK's "careful and competent driver" standard: no red lights, no wrong-way driving, no passing stopped school buses. Earn that, and the statistics can settle later. THE HUMAN NEVER LEFT THE LOOP: Remote assistance is what makes robotaxis viable today, yet the industry frames those operators as having no safety role. PRACTICAL TAKEAWAYS LEAD WITH BEHAVIOUR. Audiences judge safety by how a vehicle drives, not by null hypotheses and confidence intervals. WIDEN YOUR SAFETY DEFINITION. Blocked fire trucks and riders pulled off trains never show up in fatalities per mile, and they still count. EXPLAIN THE FIX, NOT JUST THE FACT OF ONE. "We fixed it, trust us" collapses the moment the same failure repeats. BUILD THE HUMAN INTO THE STORY. Remote operators and supervising drivers are part of the safety case, so stop treating them as a footnote. ABOUT OUR GUEST Philip Koopman is Faculty Emeritus at Carnegie Mellon University, where he spent nearly thirty years as tenured faculty researching embedded systems, software safety and autonomous vehicle safety. He is the originator of UL 4600, the safety standard for autonomous products, and the author of "How Safe Is Safe Enough? Measuring and Predicting Autonomous Vehicle Safety" and "The UL 4600 Guidebook". Before academia he served as a US Navy submarine officer during the Cold War, designed embedded CPUs at Harris Semiconductor, and has since conducted more than 200 industry design reviews of embedded software projects. He writes on AV safety on Substack and LinkedIn. AND ONE LAST THING Our closing question is always the same: what is your number one storytelling advice to our audience? Discover Phil's answer at the end of this incredible episode. Subscribe on YouTube, Spotify and Apple Podcasts, and follow us on LinkedIn. "Under the Hood: Automotive Storytelling" uncovers the human narratives driving innovation in an industry that impacts lives worldwide. Hosted by Stéphane Lagresle, this podcast explores how the power of storytelling ultimately connects technology to humans.






