What self-driving cars taught me about trust

I’ve now ridden in a Waymo, a Cruise, and a Tesla with FSD beta. Three different approaches to the same idea. Three different feelings in the passenger seat.

And each time, there was a moment.

The moment

It happens about 5-10 minutes in. Your shoulders drop. Your jaw unclenches. Your hands, which have been gripping the door handle or the seat or your own knees, relax. Your eyes stop darting between the road and the dashboard. You stop anticipating the car’s mistakes and start watching the world go by.

That’s the moment you trust the car.

I didn’t decide to trust it. My body decided. Some deep part of my nervous system ran its own calculations on the car’s behavior, observed enough correct decisions, and downgraded the threat level from “you might die” to “this is probably fine.”

Probably fine. Not definitely fine. My body kept one finger on the alert. I never fully relaxed the way I do in the back of an Uber with a human driver. The trust was provisional. Conditional. A lease, not a purchase.

The three experiences

The Waymo felt like riding with a very careful grandparent. Slow to accelerate. Conservative at intersections. Wide turns. It always took the cautious option. I trusted it quickly because it felt like it was trying not to scare me. The moment came at about 3 minutes.

The Cruise felt like riding with a confident teenager. Smoother than the Waymo. Quicker to accelerate. Closer to how a human would drive, which made some maneuvers feel natural and others feel too aggressive. The moment came at about 7 minutes, and was interrupted once when the car braked hard for a plastic bag.

The Tesla FSD was different because I was in the front passenger seat, watching the driver’s hands hover over the wheel. The car drove well on highways. On city streets, it was less predictable. Hesitant at some intersections, aggressive at others. The visualization on the screen showing what the car “sees” was both reassuring and terrifying. It saw everything. But seeing isn’t the same as understanding. The moment came at about 12 minutes and was partial. I never fully trusted it.

What trust actually is

I’ve been thinking about trust since those rides. It’s not a decision. It’s a process. You don’t decide to trust someone or something. You observe. You gather evidence. Your nervous system integrates the evidence without asking your conscious mind for permission. And at some point, a threshold is crossed and your body relaxes.

Trust is also not binary. It’s not on or off. It exists on a spectrum. I trust an elevator more than a rope bridge. I trust a commercial pilot more than a Cessna pilot. I trust a human driver more than a self-driving car, but less than I used to, because the data on human error is pretty damning.

The interesting thing about self-driving cars is that they’re asking us to trust a new kind of entity. Not a person. Not a tool (tools don’t make decisions). Something in between. A thing that acts autonomously but isn’t alive. A thing that can be wrong in ways that are different from how a person is wrong.

When a person makes a driving mistake, we understand it. Distraction. Fatigue. Anger. Human causes for human errors. When a machine makes a driving mistake, it’s alien. It misclassified an object. It encountered an edge case not in its training data. The steering wheel turns for reasons we can’t intuit.

Trusting something you can’t understand is a specific kind of trust. It’s the kind of trust you have in a bridge. You don’t understand the structural engineering. You just observe that it holds, day after day, and eventually you stop thinking about it.

Maybe that’s where self-driving is headed. Not understanding. Just observation. Enough days of it working, and the moment comes for everyone. Not a decision. A threshold.

I’m not there yet. But I’m closer than I was.

Related thinking: