Another way is to redefine the problem so its solution can have a simple mental model. This is similar to the subset approach of the first case (Economical De

Another way is to redefine the problem so its solution can have a simple mental model. This is similar to the subset approach of the first case (Economical Design), but it might not be a subset (i.e. it might overlap with other uses), and it might not serve only one user scenario, perhaps not even the most common case. It requires a wider perspective. For this to actually be useful, it needs to interact with other solutions to other problems. That is, it refactors the set of problems, shifting different aspects to different places, in such a way that it changes the information required from the user, to make a simpler mental model possible. IMHO the key questions are: - What information is needed from the user? - How can we arrange things to minimize that? Surprisingly often, it can be reduced to zero - automation. BTW: The iPhone doesn't have a "small number of affordances". You can touch the display in 1,000's of distinct places, with many simultaneous touches registered distinctly. What the iPhone has is flexibility and power: instead of a specific grammar of interaction, it enables the specification of many arbitrary grammars of interaction. IOW it's programmable.