> A counter example is a physical button on manufacturing equipment. It’s huge, tactile, and labeled. I know I could press it, but that doesn’t necessarily help me understand what happens when I click it.
On the other hand, you can be very sure you won't press it unintentionally if you stay away from it, and the difficulty in pressing it provides a hint of how big of a deal it is to press it. Membrane buttons, on the other hand, can very well implement flat or hidden buttons that one might press by accident.
Relevant to your insight (see "The Design of Everyday Things" by Don Norman):
1. The gulf of execution: how do I do something? (i.e. what can be clicked/tapped/pressed?)
2. The gulf of evaluation: how do I know/understand what happened as a result (of clicking/tapping/pressing something?)
The example of the touch feedback cuts both ways: those buttons exist in context, and one can certainly provide touch feedback without a border/shadows. Text-only buttons also exist in context and the language can do a lot to explain what happens.
A counter example is a physical button on manufacturing equipment. It’s huge, tactile, and labeled. I know I could press it, but that doesn’t necessarily help me understand what happens when I click it.
Everything on a screen /could/ be clicked or tapped, it’s inherent to the medium, making something “look clickable” doesn’t necessarily help clarify if/when I should click it, or what the result will be