Making guesses isn’t a problem. The ideal information software requires no interaction at all! You open your maps app in the morning, and it instantly brings up how to get to your next calendar appointment. If you’re making a long drive, it automatically suggests a gas station along the route. The more our software can infer, the better!

Bret Victor has a great essay on building information software called “Magic Ink” [1]:

> Information software, by contrast, mimics the experience of reading, not working. It is used for achieving an understanding—constructing a model within the mind. Thus, the user must listen to the software and think about what it says… but any manipulation happens mentally. Except possibly for signaling a decision, such as clicking a “buy” button, but that concludes, not constitutes, a session. The only reason to complete the full interaction cycle and speak is to explicitly provide some context that the software can’t otherwise infer—that is, to indicate a relevant subset of information. For information software, all interaction is essentially navigation around a data space.

Of course, guessing poorly is a problem, but that’s an issue with execution.

[1] http://worrydream.com/MagicInk/#interactivity_considered_har...