I think maybe the principle is that we want to be able to transfer as many bits/second of data as possible. We have the computer->human interface down, because

I think maybe the principle is that we want to be able to transfer as many bits/second of data as possible. We have the computer->human interface down, because screens can already emit more information than humans can process. But human->computer we can only go so fast. This is partly limited by how many fingers we use to do the input. I can type faster on a keyboard than a screen. But I think the main issue people are worried about is that humans can only think so fast. So if the interface requires asking humans for input too much, it will take forever. So instead you try and guess what the humans want. And if you do this well enough, then the input method doesn't matter. Some people say they can type better on their phones because they can use swiftkey. So I think most visions of the future are thinking about AI, and intelligence, and asking "what will we be able to guess without asking the human first?" And this is where they think interaction design is going. Another point of fact. Such a machine requires less bps from the human. That is, it requires them to do less thinking, which most people will prefer. And if you're trying to predict the future, "machines will require less thinking (but be more mass market)" is a pretty safe bet. Sandwiches are very fancy things, as everyone knows. But I expend a lot of mental effort getting the cheese slices in the right thickness. I'm sitting there focusing on making sure I don't deflect the knife too deep or too shallow. This is a very fancy task. But many people prefer specialized cheese slicing devices, because they makes more precise slices, and because they require less mental effort. So expect UI to follow the same trend, from general tools requiring complex input to particular tools requiring simple input.