Wednesday, October 22, 2008

Crowds – Wisdom or Madness?

What are we to make of The Wisdom of Crowds (2005) by New Yorker business columnist James Surowiecki?  After all, we were brought up on Charles MacKay’s Extraordinary Popular Delusions and the Madness of Crowds. The two books are at odds with each other.

First published in 1841 and popular ever since, MacKay’s work has stood the test of time.  It introduced the notion of mob psychology, noting that cool rationality is much more likely to show itself in individuals – and disappear in larger groups. MacKay put forth the Dutch tulip bubble and the Salem witch hunts as examples.

Surowiecki uses contrary examples – including the Who Wants to Be a Millionaire audience polls – to show that crowdsourcing can work.  The idea has caught on.  Companies now crowdsource their new product development! (Look here and here for examples.)  

Well, MacKay and Surowiecki can’t both be right, can they?

The Founding Fathers distrusted crowds.  That’s why we live in a republic and not a pure democracy. The idea that crowds could be right seems to stem from John Brunner's 1975 novel Shockwave Rider and is widely implemented today in what economists call “prediction markets.” 

This somewhat interesting notion got a black eye in 2003 when DARPA rolled out its “Policy Analysis Market,” which appeared to allow anonymous people to bet on terrorist attacks. Widespread public disgust was the main result.

Moreover, if markets (that is, mobs) could predict, then a horse with nine-to-four odds at the track would actually win about four of every thirteen races against a similar field, and this is demonstrably not true. (The historical database for horse races is far vaster than any for 21st-century artificial predictive markets!) 

Did Surowiecki and the economists forget that Shockwave was a work of fiction? Ah, but careful readers see Surowiecki hedging his claim: "Under the right circumstances [emphasis added], groups are remarkably intelligent, and are often smarter than the smartest people in them."

What are the right circumstances? A French mathematician burdened by the name Marie Jean Antoine Nicolas de Caritat, marquis de Condorcet (1743–1794) proved that a large crowd is likely to be right about a simple yes-no question only if each individual in the crowd has a greater-than-even chance of getting it right.  That is, only experts should be polled.  Condorcet offers no advice on how to qualify experts. Or how to find them.

So, is it MacKay or Surowiecki?  To be sure, two heads are better than one.  And small work groups in an organization can outperform individuals. (We’ve all been through that tiresome demonstration on our HR training days.)  But two to six people ain’t a crowd…

There are two phases in every decision: divergence, which means generating alternatives, and convergence, in which all but one of the alternatives are cut away.  Crowds are good at divergent creation of ideas – the more the merrier – but when it comes to deciding (a word that means “cutting away”), I see no signs of mobs doing better than small groups or singles.

Just look at the bankers who rushed to buy securitized mortgage instruments, or voters in the 2004 presidential election. MacKay takes the day!

p.s. I’ve treated this subject in turgid academese, in a paper called “Change in Socio-Technical Systems: Researching the Multis, the Biggers, and the More Connecteds.

No comments: