On Culpability: When AI Goes Wrong, Who is at Fault?

As an educational technologist, I spend a lot of time thinking about how the rise of generative artificial intelligence is impacting teaching and learning. But AI has been developing and creeping into many areas of our lives for years now, long before you ever heard of ChatGPT. Personal digital assistants like Siri and Alexa help us with minor tasks. AI-powered recommendation engines on every streaming service help us find new music or movies. Predictive text and autocorrect are useful AI supports when we’re texting. Navigation apps use big data and AI algorithms to help us plan our routes. And even mundane word processer functions like grammar check and spellcheck use rather rudimentary artificial intelligence to help us improve our writing. 

All of these are relatively routine examples of artificial intelligence that we likely don’t even think about. Perhaps these once were wonders to us, but no longer. There are so many ways technology oozes into the nooks and crannies of our lives, these sorts of technologies might seem—dare I say it?—boring?

We live in a culture that prizes novelty, after all. And technological developments often bring novel situations we must grapple with.

For example, my wife and I were in Phoenix recently, and we noticed something that surprised us. Among the throngs of vehicles, the roads in Phoenix are heavily populated by Waymos. Waymo, if you are not familiar, is a company that develops self-driving cars. Formerly a project of Google, Waymo has been developing autonomous vehicles since 2009. They have made all kinds of breakthroughs in terms of cameras, audio sensors, lidar, computer modeling, and artificial intelligence algorithms to manage all the data the vehicle needs and make decisions that allow for safe driving. Phoenix was ground zero for Waymo’s launch of their self-driving fleet of ride-share vehicles, sort of like Uber…but with no human behind the wheel.

I confess, I was fascinated! But I wasn’t quite ready to hop in and trust the machine to do the driving for me.

Why not? 

The short answer: the novelty of this situation brings up some anxiety for me. I am not (yet?) confident that the machines will be able to handle the complexities of something like driving out in the wilds of the real world. Artificial intelligence is powerful…but am I ready to let go of my own agency and trust the machine? When AI goes wrong, who is at fault? The programmers? The users? Can we assign guilt to the machine itself?

My experience watching Waymos driving all around Phoenix was definitely on my mind when I recently read Bruce Holsinger’s unsettling novel, Culpability. The book was published in 2025, and as someone who regularly works with and studies AI systems, I found it a gripping, very contemporary story. The book is equal parts family drama and techno-mystery, exploring the nature of responsibility (or culpability) when it comes to situations where both humans and machines are actors in an event.

The story opens with a devastating highway collision. The Cassidy-Shaw family is traveling in their high-end autonomous minivan when the vehicle crashes into an oncoming car, killing its elderly occupants. While the AI system was navigating, seventeen-year-old Charlie—a star athlete bound for college on a lacrosse scholarship—was in the driver's seat. 

This tragic event shatters the family’s polished veneer, transforming their lives into a complex web of legal peril and psychological unraveling. Charlie is crushed by the trauma of the accident and under investigation by a dogged detective, and begins to spiral. His father, Noah Cassidy, a protective defense attorney from a working-class background, acts as the family's emotional anchor, working to fiercely guard his children from police interrogation. Meanwhile, the pre-teen daughters, Alice and Izzy, deal with their physical and emotional trauma in isolation; Alice increasingly withdraws into furtive interactions with an AI chatbot. 

The wife and mother of this family, Lorelei Shaw, was also in the vehicle, and was also impacted by the accident physically, emotionally, and even professionally: Lorelei is a brilliant, world-renowned AI ethicist whose academic research on the morality of algorithmic accountability. In the wake of the accident, Lorelei's behavior becomes increasingly erratic, which raises Noah's suspicions that she might know more than she is initially letting on. 

To escape the pressure of the investigation, the family retreats to a vacation home on the Chesapeake Bay, only to find their rental house landing them next door to the compound of a ruthless tech billionaire, Daniel Monet. When a reckless, high-stakes romance sparks between increasingly emotionally troubled Charlie and Monet’s impulsive daughter, Eurydice, a second crisis forces the family to confront a chilling reality: every member of the Cassidy-Shaw family is harboring a secret that directly implicates them in the initial crash. The rest of the story unfolds in dramatic fashion, leading to an ultimately satisfying and realistically-complex conclusion.

The thing that really resonated with me is the idea that AI is never neutral; choosing to use artificial intelligence means choosing to allow something other than ourselves to shape our perceptions, actions, and perhaps even our self-justification. The central characters (including Mom the AI ethicist, Dad the lawyer, and three kids who all have their own relationships with phones and people) all illustrate different viewpoints on what AI is and how it is shaping contemporary society.

The fascinating question raised throughout the book—both in the crash, and through the subsequent events—is “Who is to blame when things go sideways in a situation with AI?” In the novel, no single actor, human or machine, can fully own the consequences of the crash, yet no one is entirely absolved either. Holsinger captures the modern temptation to offload responsibility onto systems we design, while also insisting that design itself is a moral act, and one that has consequences that we must grapple with if we are going to use powerful artificial intelligences in our day-to-day lives.

What lingers for me after reading is Holsinger’s refusal to offer clean answers. I found the book so engrossing precisely because it forced me to sit in the discomfort of distributed blame, especially as someone who uses AI regularly. It isn’t all doom and gloom in this book, but it isn’t a rosy picture of a tech utopia either—sort of like this brave new world we find ourselves in. 

It’s a fast-paced, fascinating book. There is some strong language and minor “adult” content, so let the reader beware. But if you are interested in expanding your thinking about the ways artificial intelligence is already influencing your life through a fictional-but-realistic lens, I recommend this book to you. It is an intriguing way to help us explore our own thinking about culpability in the age of AI.

I’m still curious about Waymo, and perhaps I’ll take a ride in one the next time I’m in Phoenix. But after reading Culpability, I have different questions than I did before about what I’m agreeing to by hopping into the backseat of an autonomous vehicle. 

Dr. Dave Mulder will lead a session on artificial intelligence and its implications for our work in the world at the At Work in the Garden conference at Dordt University on September 17 - 19. This conference centers around reclaiming a robust theology of work as God’s good design, entrepreneurship, free enterprise, and vocational calling.