As I mentioned several weeks ago I saw a presentation by Professor Schneiderman of UMD. One of the high level themes of his presentation (and his book, Leonardo’s Laptop) is that programmers should not so much seek to make computers magically find the answers for us (and indeed, many of the “promises” of AI have never come to fruition, but should instead strive to make tools that support human creativity.
How to do that? One great application domain is sifting through mountains of data to spot hidden trends and relationships. If a computer isn’t going to do this well, then it should facilitate the user’s having the “aha” moment. People think with their eyes (indeed, the image is processed quite a bit before getting on the optic nerve) and are amazing at spotting things that don’t “fit” or belong – so show them the data in a way that isn’t row after row of numbers or text, and then let them interactively sort and filter. (and “drill-down” and “roll-up” aren’t adequate info-vis tools).
Two cool tools in this vein were demonstrated- The Hierarchical Clustering Explorer, a prototype out of the HCIL at UMD, the other, Spotfire is a commercial application. Both let you simultaneously view several different “slices” of a mountain of data, and changes you make to one view (filtering for example) are instantly reflected in the other views. One of the cool things about both tools is the capability to effortlessly stretch and shrink axes, to zoom in on subsets of the data – because, as Prof. Schneiderman said well, “A pixel is a terrible thing to waste”.