I have heard (but have no insider knowledge) that it's not just the cost of parts, but what parts do to the development lifecycle.
With physical parts, the development process is highly sequential. Pick the look, design how it fits, engineer what parts are used, manufacture tooling etc etc in a waterfall. If a revision needs to be made, the whole process needs to be re-started adding a huge amount of delays.
With a touchscreen, the physical touchscreen and the software that runs on it are parallel threads. You can make most UI changes without impacting the manufacturing/design pipeline at all. You don't even need to have planned what the interface looks like before you finalize the parts needed.
I suspect that another thing that's expensive, besides analog controls, is creating and testing a new, better interface for driving ICE cars using digital controls. So the only thing old our old-tech car companies do is add ad-hoc extensions of the usual car controls. And it is economical to package all these in a touch screen along with all the standard controls they can get away with.
I feel like you could solve that with some constraints defined early in the process—the controls must fit in dimensions XxYxZ, and have K-input wires, and draw at most 5V of power, or whatever. Then the control designer can go off and design whatever they want within those constraints, and the car designer can go off and design the interior of the car knowing what the head unit will require.
It's a good argument except the basic controls needed for a car haven't changed in decades: We all need the same climate controls, media controls, etc.
For deep car settings, those have been screen controls for over a decade now and that makes sense.