The only metric that matters is how fast you can get attention back on the road.

They're dangerous for controls the driver would reasonably need to operate while driving the car. They're fine for more-complex at-rest configuration, or stuff a passener would care about.

The article directly links to a study that shows this is also true of physical buttons. Regardless of the fact that buttons are tactile, people don't go feeling up their radio without looking, even if they can. Furthermore, the vast majority of infotainment input today is into phone mirroring systems like carplay.

This whole thing is compounded by the fact that Mazda's knob solution was actually worse while being marketed as better. While a touchscreen needs to be looked at to find a button, a cursor controlled by a knob needs to be watched in whole to navigate to the button. Your fine motor skills as a human allows you to directly press a button, physical or not, without looking at your arm to get near it.

With the knob, you have a few issues:

1. iOS made the focus border on UI elements very faint. So it’s hard to tell where the knob is at rotating through all the UI elements.

2. Because zoom is kind of a sub feature, you have to rotate through like 10 buttons to get to the right thing, click, then get into a submenu.

3. Because not many apps design around the knob… the active “cursor” can get trapped in a submenu where the knob just rotates between a few buttons and can’t escape back to the root of the app.

Basically, it takes active attention to zoom in/out. Touch screen, I could probably do it without looking.