@ SxSW: Bryan Caplan: Questioning The Wisdom Of Crowds

We see tons of startups predicated on the so-called “wisdom of crowds”, an idea that’s been so thoroughly internalized by people, the actual premise, that crowds are wise, are rarely questioned. But maybe when the individuals within the crowd don’t know anything, the crowd doesn’t either. Bryan Caplan, an econ prof at George Mason University, said he was invited to speak at SxSW Interactive to provide a counterpoint to James Suriowiecki’s popular book The Wisdom of Crowds. Caplan’s own book, The Myth of The Rational Voter argues that typical voting schemes often result in bad policies. He makes a somewhat elitist argument, which he’ll acknowledge, but coming from an economist, there’s a lot more depth to it than, say, similar arguments from Andrew Keen, whose viewpoint can pretty much be boiled down to “experts rule and the plebes drool.” While Caplan’s presence here may seem a bit far afield form everything else, it’s not hard to make the connection.

The Mirage of Aggregation: The idea that voting works is based on the so-called “Miracle of aggregation”, which Caplan likens to a form of alchemy. You take some uninformed opinions, mesh them together, and shake it all up and voila, it’s informed.” While Surowiecki’s book cited a number of examples of crowd wisdom (groups of people collaboratively guessing the number of jellybeans in a jar or guessing the weight of a cow), Caplan criticized what he saw was a “(weak) hunt for counter-examples.” His one-off counter-example: In polls asking voters what percentage of the budget is spent on foreign aid, averages come around 10-20 percent, when in fact it’s actually just 1.2 percent. Ultimately, in Caplan’s view: “The miracle of aggregation fails and it fails very directly.” As that relates to democracy: “What’s interesting about this miracle of aggregations is it gives people a way to believe democracy works despite the public’s deficiencies.”

Predictions markets and the web: I asked during the Q&A what the implication of his work on stuff like prediction markets and other wisdom of crowds-based startups. As he argued, there’s a huge difference between public opinion polling and betting markets. For one thing, there’s a financial penalty for being wrong. So ignorance gets blinded down, so to speak. The problem is when there’s no policy for being wrong, and no way to keep out the ignorant: “On almost any question, if the betting market says something I believe it.” He didn’t address it directly, but the question for a site like, say, Digg, is to determine whether there’s an adequate penalty for being wrong. Typically, for a non-financial site, the currency is reputation.

Bottom line: Caplan’s presentation and work should be though-provoking from a web perspective, but it’s not ultimately damning. Nor is he in as much contradiction with Surowiecki as he suggests. He argues persuasively that you can’t bake wisdom by cooking with ignorance, but it doesn’t mean that things like prediction markets, or other collaborative filtering services (Last.fm, Netflix (NSDQ: NFLX), etc.) can’t work, when people voluntarily.