The process of conveying information through visual means is too often thought of in purely spatial terms. We try to map concepts into space and then present that space. This works wonderfully for those people blessed with strong spatial reasoning skills, but fails badly when used on those not so blessed.
There is no reason for such narrowness of reach. If we think in terms of visualization rather than spatialization, we can break away from the rut in which we're caught. The trick is to search for other modes of human cognition that are nonspatial. Two in particular strike me as worthy of further development: emotional, as visualized through faces, and linguistic, as visualized through the written word.
The presentation of human faces in software is woefully inadequate.
Whether we like it or not, people anthropomorphize their computers and relate to them in emotional fashion. We can convey important information to the user through facial displays. For example, various error messages can be accompanied by faces showing varying degrees of alarm, to communicate the seriousness of the problem. Skeptical faces, inquisitive faces, intent faces can all be used to communicate important information to the user. Most important, good facial displays can do wonders to help a user navigate through the kind of complex problems that now bedevil many users. The stupid paper clip used in some Microsoft software has cast doubt on the concept, but remember that a good concept can always be unfairly discredited by a poor implementation.
The use of linguistic structures seems a throwback to the bad old days of command-line interfaces, parsers, and all that nonsense. But if we approach language as not such a sequence of characters, but a grammar for combining words into meaningful expressions, we can see the value of Linguistic User Interfaces (LUIs). The trick is to use a subset of a natural language that is OBVIOUSLY a subset; if the user ever thinks that we are offering true natural language, then s/he will surely end up frustrated with our presentation. If instead we offer a kind of "baby talk" with the computer, we can greatly expand the working vocabulary of our users and offer large feature sets unhandicapped by clumsy user interfaces. Remember a LUI is distinguished from natural language in that it is interactive. Interactivity makes it possible for the computer to interact directly with the user during the framing of his/her expression, thereby greatly simplifying the interaction.