My son got one of those twenty questions games — the ones where you answer yes/no questions from the little plastic ball and it guesses the thing you were thinking of — for his birthday. Internally, there’s not a lot to these; they’re basically a fairly simple expert system and a rom full of facts about everyday things. What’s interesting to me though is the way that they get close to the right answer, even when they get it wrong. For example, when we tried to get it to guess “green bean” it came up with “asparagus”, which is not a green bean, but another thin, green, vegetable — not bad.
So I got to thinking about graphics design tools for beginners. Imagine a tool that had an internal expert system, which encoded a bunch of the standard patterns used by web sites (at several different levels (i.e. everything from nav bars and rollovers at one end to blogs, wikis, e-commerce at the other)). Rather than just picking 1 of N templates as the basis for your website, you would answer 20 questions about the capabilities it should provide. As you progressed through the questions, it would use its internal predictions about the result to suggest other questions to narrow things down, eventually getting to the point where it was dealing with the stylistic elements. Even if the site it gave you at the end wasn’t exactly what you wanted, it would probably be close enough.
In any case, it seems like this would be a more flexible way to approach the problem than I’ve seen in most of the newbie-level web tools (like, iWeb) but it’s still a metaphor that new users have no trouble with. Of course, it all depends on how well you can tune the questions. My favourite question from the toy is “Does it weigh more than a pound of butter?”. You want the questions to have that feel: clear examplars that partition the space. (“Should it be simple like Google, or dense like Yahoo!?”)