The Sh*t L*st

I read the first two chapters of James Surowiecki's new book, The Wisdom of Crowds, at the campus book store yesterday. I enjoy the neat counterintuitive moralisms of his New Yorker columns and I'm enjoying the book, too. I'm looking forward to getting back to it next time I'm at the book store.

But he seems, thus far, to be trafficking a form of enlightened mysticism. I notice that he slips in a number of qualifiers in tossing out his claims regarding the magic of markets: a lot of generally's and usually's. He's hedging. Also the anecdotal quality of a lot of his early examples raises questions. Not that the anecdotes aren't accurate or even somewhat representative. But they do hint at a selection bias (the kind of thing, ironically, an intelligent market transcends.)

According to the examples he provides, crowds are good at making predictions related to crowd behavior or, say, in the case of jelly beans or ox weight predictions, intuitive acts that are so intuitive, we may may not even recognize them for what they are. Predicting the number of jelly beans in a jar or the weight of an ox is not so much a demonstration of some kind of esoteric collective insight into the mysteries of candies or livestock as it is an exercise in correlating visual perceptions with some kind of standardized unit of measurement -- something we all have some ability to do instinctively. The deviations over and under the correct result, being effectively random, cancel one another out. It's an example of groupthink -- to use a term that's been popular lately and that I heard Surowiecki himself on the radio expostulating about this week -- as much as anything.

Here's a falsifying case: get your huge, diverse crowd and have them predict the sequence of the genome. You could have them work on that last 10 percent that scientists with the genome project haven't worked out and is going to be the most expensive part to sequence. The problem is not just that the group won't do better than say 25-35% in their predictions, but that the information as a whole is usless unless it's over 99% accurate.

Surowiecki suggests markets as a new tool for business and science. But it would probably involve as much energy and expense figuring out for what situations, beyond the obvious cases, this would be appropriate as just hiring the good old experts and letting them take apart the problem. Ultimately you'd be relying on informed elites -- with some heavy number-crunching of past performance -- to make the decisions about what novel situations in which to apply this. Is there a meta-market for figuring out where the wisdom of markets applies?

But, sure, if you want to know how people will vote or what movie they will see this
weekend or what type of car will be most popular next year, any large group is going to be better than a single expert. The aggregate in this case accumulates a lot of relevant information that the expert can't really get at except by tapping the aggregate. I suspect that the book will end up being an argument for the superiority of powerful polling and database technologies (like Google) over traditional human-grounded market research presented in the guise of a radical new thesis about society.