You Know What I Want to Know - The Wisdom of Crowds

I guess you could read many people's views about James Surowiecki's The Wisdom of Crowds, and that's probably the best and most fitting way to find out if you will like it. Surowiecki's idea is that, contrary to the message of negative terms such as mob psychology or groupthink, a crowd can be smarter than an expert.

Something about this concept is so appealling, but after reading it I am even more convinced that crowds are stupid and that only individuals thinking independently (while comparing notes) are smart.

If you look closely at his many examples, they don't show that a random collection of people is smart; they show that taking the average of the views of a collection of people thinking individually leads you to an accurate result:

  • estimating a bull's weight (Francis Galton calculated the average of 787 "votes" which was accurate to within 10 grams)
  • identifying the most likely position for a lost submarine (John Craven used Bayes' Theorem to approximate the location, only 220 yards off)
  • the "surprising" accuracy of the audience in the game show Who Wants To Be A Millionaire; his quoted rate of 91% is not correlated with the actual value or difficulty of the question; since the initial questions are intended to be 'common knowledge', this result is to be expected
Several examples allow for communication between the individuals, or even a cyclical accumulation of information:
  • the collective result of people navigating an unknown maze multiple times; how surprising is it that the average path (a journey comprised of the average decision at each branch in the maze) turned out to be optimal? don't we expect people to learn, and then mostly remember what they learned just minutes earlier?
  • the actions of the stock market; I would argue that this is similar to a transit of a maze; the decisions of the individual stock traders are not independent, they are often sequential and influenced by earlier decisions
  • the specific example of the stock market identifying Morton Thiokol (manufacturer of the solid-fuel booster rocket) as the cause of the space shuttle Challenger explosion seems promising, but given that the solid booster rocket is the part that is active during initial lift-off (Challenger blew up after 74 seconds), it's intuitively the safest bet; clearly not everyone is going to make this connection, but it's possible that enough smart individuals start some stock movement, and the rest can be explained by herd action or momentum
  • Google accuracy - ok, this is truly amazing, but it also demonstrates an accumulated result; since users posting information often use Google to find information, they have information that enables them to create a positive feedback loop that increases the accuracy and relevance of the result; in essence, I believe this simply shows that there are often an infinite number of wrong answers, but only a few right answers which many individuals will find independently

In my view, the suggested wisdom of "crowds" is the result of either positive feedback loops or an algorithm that averages a collection of individual less accurate results. But what does this really add to the use of Bayes' Theorem to use conditional probabilties to process subjective accounts of evidence? An individual who has a chance to learn from another is more likely to achieve a better result.

Don't get me wrong - this is a lovely book, well worth reading; it is often inspirational to see examples of collections of individuals achieving a good or even perfect result. But these collections of people are not groups let alone crowds in the normal sense.

For me the standout best example of "the wisdom of crowds" (according to Surowiecki's meaning of "crowd") is open-source software development. When the individual hackers are talented and the quality standards of users are high, the results of a rapid feedback loop of continuous development can be excellent. But this is a "group" only in the sense of a collection of individuals who modify their individual behaviour, not in the sense of people who are coordinating their group activity.

In a nutshell, when and why do groups fail? Here's my attempt to state this; they fail when they look inward as a group, developing group norms that override individual intelligence; the group then is simply an exaggeration of the impulses of herd leaders. And the group succeeds when it truly seeks the best result from a collection of individuals, each individually seeking the best answer and modifying it based on information and opinions from the others.

This here is a blog entry. How did you find it? Beats me, but the most successful blogs are by definition the result of many iterations of people making a decision based on an accumulated or average decision made by others. Fame or recognition operates in the same way, in part based on individual decision, in part based on herd momentum. The wisdom of crowds? I can't call it that. To me it is positive feedback loops, new sources of information and each individual making the most intelligent decision that they can at any moment in time that will lead to the best result. Not so romantic, but essentially the truth.

Should you buy The Wisdom of Crowds? Well, don't buy it because it's in some top 10 list; buy it because you value and agree with the reviews of others who liked it.

As Monty Python has it, we are all individuals. So long as that remains true, then collectively we're pretty intelligent.


Have you ever done the team building exercise where you rank what you will take with you for a walk across the desert after a plane crash. You rank the items first by yourself and then as a team. The team almost always does better than the individual. You are right though it is not the crowd that does it. It is the diverse combined wisdom of the individuals in the team. I think we sometimes think we know best than a group. We forget that groups are made up of individuals I've just started reading the book. Thanks for an interesting blog.

Posted by Marion Vermazen on December 03, 2004 at 07:54 AM PST #

Hi Marion, I haven't tried the desert exercise, but I've done something similar where a team have to drop a raw egg two stories to the ground - and use some insubstantial materials to keep it from cracking! It was a lot of fun - plenty of opportunity for individual creativity, some tough team decisions to prioritise which techniques to try and how best to use the resources - and then the moment where you do the egg drop. Appreciate the comments, and hope to read your reaction to the book when you're finished.

Posted by Colm Smyth on December 03, 2004 at 09:44 AM PST #

Post a Comment:
Comments are closed for this entry.



« July 2016