Computerized Adaptive Teaching?

As I continue to think about how to make the best new teaching thing on the internet, I finally got around to reading up about Computerized Adaptive Testing, famous as used for the GRE. The math used for rating questions and so on is pretty neat. There's a whole thing called Item Response Theory, which is fancier than the chess ranking thing that I was thinking of, after seeing it used in The Social Network. The two are similar, and similar to what I was thinking of, in that they consider questions as ranked in the same way as people, but almost always along just one dimension: goodness-at-chess, or goodness-at-math, or even just "intelligence" or whatever. That seems like the main drawback for adapting such ideas for a more useful automatic evaluation and/or instruction system. A good teacher knows what you're good at and what you need to work at, not just "how smart you are" or some similar unidimensional measure.

And because I was on wiki, I also went and read some about the Likert scale (fancy name for "do you agree? choose 1, 2, 3, 4, or 5). I was scarred for life when somebody mentioned this Likert thing like I should know what it was, and I only knew it as that thing, where you choose 1-5. There's also a whole wiki page just for "rating scale" - for when you need to be particularly scientific about whether she's an 8 or a 9, I guess.