I’ve been reading a lot of different books recently that talk about the problem (well, sometimes they’re good, but sometimes they’re a problem) of information cascades. And that got me thinking about Google searches.
See, in the post below, I wanted to hyperlink the term reservation price, so non-econogeek types who might not recognize the phrase could quickly see what it meant. So I punched it in to my Google toolbar, hit search, and clicked through to the first result. A quick skim confirmed that it was about what I’d had in mind, so I used that. (I did about the same with information cascades above.)
Here’s the thing: Which result was first was determined by Google’s PageRank algorithm. Which, to simplify a bit, means that each inbound link to a page counts as a “vote” for it, pushing it up the rankings. Logical enough from the outset. But what happens when people routinely decide which pages to link by following the heuristic I did—see who Google ranks first (or in the first few for a search term, if the first result isn’t quite what I want) and use that? You get a kind of lock-in, where pages that rank high at the outset become increasingly unlikely to be displaced by newer pages as people use the Google ranking to determine which pages they ought to link.
It’s still only a heuristic, of course; if the first page of results seem quite inadequate, or the fifth result looks, offhand, like it’s of higher quality than the first, people don’t follow the rule slavishly. But it does suggest that maybe the longer Google’s around, the more it’ll be the case that it’s very good at satisficing searches, but not so good at optimizing results.