[andrewla]

Doesn't this assume that you're advancing a state machine from some known initial state? If you are in the middle of a navigation and you are interrupted, in order to figure out what to do next you need to recover context. The more your muscle memory remembers the sequence ( 1 -> 2 -> 2 -> 3/4 for raising/lowering the fan speed ) the less you'll be able to access it consciously and the more you'll have to look at the options to figure out what to do.

The lack of tactile feedback on touchscreens is an issue when using it, but I'm not convinced that it's any better or worse for partial-attention tasks. For full attention tasks (e.g. navigating a settings menu) it's far superior to a button-based approach.