In order to psychologically approach interfaces, one might always need to contextualise them, as cognitive processes are affected by the environment they’re immerse into.
Some complex systems require complex interfaces and steep learning curves for users to acquire an adequate state of mind (air controllers, train controllers and operators, jet pilots), and simplifying them might “seduce users into shallow cognitive behaviours,” but this might not be the case at all levels, where sometimes there’s no need for any learning and fulfilling an activity might suffice.
So for example, the process of sending an email might be enhanced by a step by step wizard experience (reminding the user to select a sender, write a message, append an attachment, etc.) while the driving of a train or an airplane might benefit from a steep learning curve that assures and enhances the attention and modal cognitive approach of the user, as train operators and air controllers can prove.
Learning an interface is a process affected and mediated by context, and so is any cognitive approach, I would argue.