Monday, April 24, 2006

The Talent Show

I don't believe Google's PageRank can be the final solution as far search rankings go. The reason for this belief is similar to my reason for believing oligarchies are unlikely to produce the fittest rulers (not a good sampling, not meritocratic, etc.)

Much has been said about the role of power law distributions on the Net. Although things begin democratic and meritocratic with PageRank, over time, I believe things eventually become oligarchic, and the resulting system is one in which the rich get richer at a rate that far more often than not outpaces challengers.

Lately, Google's rankings have seemed more erratic than usual. Results and PageRanks seem much less correlated than they used to be. I wonder if that means I'm seeing shapes in the search result clouds, or if it means Google's been jiggering its algorithms more than usual.

A problem with PageRank is that once the power structures and establishment are in place, we're left with the question of what happens when the best results to a particular search have a really low PageRank.

In such cases, if PageRank is strictly followed, the best results will probably never be found; likewise, the probability of anyone linking to them will be much lower than the probability of someone linking to someone who's already at the top.

No matter how I look at it, PageRank seems to be necessarily a suboptimal strategy, if the objective is providing users with the best search results possible.

One obvious corrective to me is a search engine equivalent of a talent show; i.e., algorithms that give search result unknowns a chance to compete for the big time. Such a strategy would involve mixing the usual results up a bit with low ranking candidates that appear to have potential. If one of these candidates is selected by users and it seems to satisfy them (i.e., they don't come back for more), the rank of the candidate could be increased.

Update (4/26/2006): Wattakoinkidink! This just popped up in my RSS reader:
Shuffling a Stacked Deck: The Case for Partially Randomized Ranking of Search Engine Results


"In-degree, PageRank, number of visits and other measures of Web page popularity significantly influence the ranking of search results by modern search engines. The assumption is that popularity is closely correlated with quality, a more elusive concept that is difficult to measure directly. Unfortunately, the correlation between popularity and quality is very weak for newly-created pages that have yet to receive many visits and/or in-links. Worse, since discovery of new content is largely done by querying search engines, and because users usually focus their attention on the top few results, newly-created but high-quality pages are effectively ``shut out,'' and it can take a very long time before they become popular."


Post a Comment

Subscribe to Post Comments [Atom]

<< Home