Saturday, February 26, 2005

GMWIMW - Give Me What I Might Want

With all the talk about search engines, personalization, tracking, histories, etc., there is a little too much focus on trying to give the user results that their past history suggests that they would want. Maybe it's just me, but I have a different interest than merely wanting to see stuff similar or related to what I've seen in the past or what people similar to me are interested in. I'm always searching for new stuff, so what I would most like the computer to do is to "Give Me What I Might Want" or GMWIMW.

This is actually the opposite of using my past history to predict what I might be interested in. Rather than take my history and moving delta to similar topics that correlate well with my past interests (or even new results of people similar to me), I want to make a quantum leap in some unexpected direction and get results that will likely have the lowest possible correlation with my past interests (or the results selected by people similar to me).

This is what I want the computer to do. Whether this is feasible, is another matter.

Actually, I do know for sure one technique that at least offers the possibility of showing me results that I might want: randomly select an item of information that I've never seen before. Now of course that will frequently (usually) give me all sorts of uninteresting stuff that I have absolutely no interest in. That's okay. Just give me a little button so that I can signal topics that should be semi-permanently crossed off my potential interest list. I say semi-permanently, because even then, the computer might periodically query me as to whether some of those topics should really stay on my "do not show" list. It could do this by displaying closely related results (to the results I've expressed an extreme disinterest in) on the off chance that there was simply some superficial detail that discouraged me. In any case, after a short while, the computer would have quite an impressive library of topics and sub-topics that can be weeded out of even a random GMWIMW process.

I'm not suggesting that GMWIMW should be a random process, but at least there is some hope that GMWIMW could conceivably be implemented.

To me, this is a "growth-oriented" search strategy. One that seeks new paths. One that seeks new horizons. One that seeks enlightenment. One that seeks inspiration. One that seeks innovation. One that almosts makes the computer seem to have something like intuition.

On the other hand, I don't presume for one moment that my interests in GMWIMW coincide with those of the average search user.

Still, almost everyone has moments when all the traditional, methodical, and even heuristic strategies and techniques for making incremental forward progress are not getting you anywhere. Those are precisely the times when GMWIMW is the optimal search strategy.


Post a Comment

Subscribe to Post Comments [Atom]

Links to this post:

Create a Link

<< Home