« Hot Teams With Great Chemistry They're Not | Main | The People's Champ Speaks! »

January 9, 2005

The Wisdom of Cliques

In which we explore the collective intelligence of the folks who pass through my apartment

James Surowiecki's nifty new book The Wisdom of Crowds (buy it here) opens with the story of the 1906 West of England Fat Stock and Poultry Exhibition. Among the featured attractions at the exhibition was a contest where visitors could wager on the weight of an ox; the contestant with the guess closest to the actual weight of the ox would win a prize. In attendance at the Exhibition was an eighty-five-year-old scientist named Francis Galton, who wanted to use the guesses from the contest to prove some points about the intrinsic failings of democracy; specifically, he wanted to show that the average contestant didn't know very much at all. Galton was then as shocked as anyone to discover that the average contestant's guess (that is, the arithmetic mean of all 787 guesses) was 1,197 pounds. The actual weight of the ox was 1,198 pounds. The crowd was almost exactly correct.

Pretty cool stuff. Surowiecki's book is chock full of examples of how collective decision-making (that is, harnessing the power of a group of people) routinely outperforms single decision makers, from financial markets to pro football point spreads to predicting which films will win Oscars. Emboldened by the book as well as a course I had taken in Behavioral Finance (for a good overview of this nascent field, check out this site), I wanted to see if I could recreate the uncanny effects of collective wisdom in my apartment.

The premise was fairly simple. At the end of November, I filled a glass jar with jellybeans and posted instructions next to the jar: write your name and guess on a slip of paper, closest guess to the actual number of jellybeans as of January 1, 2005 wins a dinner cooked by yours truly (a prize of almost inestimable value!). We get a fair amount of traffic in our apartment (especially around the holidays) and I figured that we would be able to generate a decent sample size. In the interest of increasing that sample size, we were lenient about multiple guesses from single contestants -- the only requirement for multiple entries being that you couldn't guess again if you could remember what you guessed the first time. In terms of strictness and scientific precision, the contest was admittedly amateur hour, but given the simplicity of the premise and the thinking behind it, I figured that (even with all the sloppiness involved) we might still create some good data.

In total, our jar of jellybeans yielded 31 total guesses from 22 different people. (I was hoping to get 50 guesses from almost that many people; I guess we're not as popular as we thought. Sigh.) A number of different methodologies were employed, with those who chose a more scientific approach inspiring more than a few "What's the formula for the volume of a cylinder?" conversations. Of the 31 guesses, the highest guess was 3,417 (complete with extensive scratch work in we assume the hope of securing partial credit!) and the lowest was 123. The average (mean) of the 31 guesses was 712.5, while the median guess was 593 -- implying that most of the guesses were below the average, but that the average was increased by outliers on the high end. The standard deviation of the 31 guesses was just short of 577; only the maximum and minimum guesses sat outside one standard deviation of the mean.

The actual number of jellybeans in the jar was 745. The winning guess, 732, was actually very nearly dead-on. The winning guess was also the only guess that outperformed the average of 712.5. Only six of the guesses were within one hundred jellybeans of the actual number. True to the promise of The Wisdom of Crowds, the average of the guesses was more accurate than all but one of the individual guesses.

Obviously, the experiment wasn't perfect. The sample size was pretty tiny, and allowing multiple guesses per entrant compromises the results a bit. In an ideal world, we'd have solicited tons of guesses and forbidden people from entering more than once. All that said, the collective wisdom effect actually ended up working out! The average guess turned out to be pretty close to the actual number, despite all the experimental fuzziness. Vindication was ours -- our clique is just as individually dumb and collectively intelligent as the folks at the 1906 West of England Fat Stock and Poultry Exhibition! (We also, on average, have more teeth.)

Posted by thatkid at January 9, 2005 8:33 PM under Biznass

Comments

Post a comment

About ThatKid

Copyright (c) 2004-2007 thatkidinthecorner