Just finished James Surowiecki’s “The Wisdom of Crowds.” This is a great book about the “behavioral economics” involved in groups. The basic premise of the book is that with certain conditions met, the decisions and actions taken by a group of people produces consistently better results than those that would have been taken by any single person in the group, including the smartest and or best informed individuals.
Surowiecki explores numerous examples that encompass problems of cognition, cooperation and coordination. Through these examples, he shows how very often performance of these groups is strikingly better than the performance of any single person. A group of people with a diverse set of information and opinions on a topic, can better predict the right answer if they have a way of systematically aggregating all of their input. Each person has a different set of incomplete information for a particular problem, some of it correct, some of it incorrect, but all of it incomplete. The aggregated results of these different proposed answers cancel out the incorrect information, and produce a better (not necessarily “best”) result.
One of the things I found most interesting in the book was the discussion of predictive markets. A predictive decision market is a form of game theory whereby different actors are allowed to trade on the given likelihood of possible events. The idea is that a group of people betting on potential outcomes can predict certain events as more likely than others, giving government or private enterprise more actionable information.
You might remember a few years back a DARPA project called the Policy Analysis Market (PAM) (which was lumped under a program called Total Information Awareness) that set up a market that predicted the possibility of world events including terrorist attacks, assassinations, and other world events. Some eager beaver congressional-types leapt on the program as morbid and “crazy.” It’s unfortunate, as it could be one way to better aggregate and surface the U.S.’s maze of intelligence agencies. Today the United States’ intelligence information is spread out across myriad agencies (FBI, CIA, NSA, DIA, ATF, etc.) that don’t have incentive to always work together. PAM would have had actors across these organizations make “bets” on potential world events. No, not with real money – though predictive decision markets get better when the actors have real skin in the game of one sort or another. The U.S.’s solution to the cross agency information problem? Make them all report to a single person, the National Intelligence Director. If Surowiecki’s theory is to be believed (and it’s a convincing argument) this is the exact wrong action – in effect, placing all of the responsibility on a single person (see the CEO-worship section near the end of the book). We can only hope that the predictive decision market has been setup anyway, but that it’s actually classified.
A real predictive decision market that you can go to and play with (and bet real money) is called Intrade Prediction markets. Intrade allows you to bet on real events, including current events, financial index milestones, entertainment, and even the weather. For example, you can bet that on both the Republican and Democratic nominations for the 2008 election, as well as the eventual winner. For example as of today the market predicts a 51% likelihood that Hillary Clinton will win the Democratic ticket (Obama is tracking next at 23.4, and Edwards at around 11, plus other candidates which are more noise.) Just to mark it down, if you were to take the leaders in both Presidential and VP nominations, the Democratic ticket will be Clinton/Obama. Republican ticket will be either Giuliani or Mccain (they’re tracking 31 and 30), with Mitt Romney as V.P. We’ll see. The idea of the market is that as we get closer and closer to an event the market reads information better or sees a given outcome as more likely. Actors in the market can buy and sell contracts at any time.
Predictive decision markets are a really interesting area. Definitely not the last you’ve heard of this topic.
The end of the book describes an instance of crowds not acting in concert for a better benefit, stock bubbles. Surowiecki details multiple examples of stock bubbles, and potential explanations of how and why they are caused (no one knows exactly). One reason (or at least sign that one is happening) is that people are basing their actions on other’s actions, rather than their own information. That is, when people are buying an asset because prices have gone up (opposite usual trend when a price goes up) and they think prices are going to continue to go up – sign of a bubble.
This is a really interesting book – great companion to Freakanomics, The Tipping Point, and The Undercover Economist. Highly recommend it.
Surowiecki explores numerous examples that encompass problems of cognition, cooperation and coordination. Through these examples, he shows how very often performance of these groups is strikingly better than the performance of any single person. A group of people with a diverse set of information and opinions on a topic, can better predict the right answer if they have a way of systematically aggregating all of their input. Each person has a different set of incomplete information for a particular problem, some of it correct, some of it incorrect, but all of it incomplete. The aggregated results of these different proposed answers cancel out the incorrect information, and produce a better (not necessarily “best”) result.
One of the things I found most interesting in the book was the discussion of predictive markets. A predictive decision market is a form of game theory whereby different actors are allowed to trade on the given likelihood of possible events. The idea is that a group of people betting on potential outcomes can predict certain events as more likely than others, giving government or private enterprise more actionable information.
You might remember a few years back a DARPA project called the Policy Analysis Market (PAM) (which was lumped under a program called Total Information Awareness) that set up a market that predicted the possibility of world events including terrorist attacks, assassinations, and other world events. Some eager beaver congressional-types leapt on the program as morbid and “crazy.” It’s unfortunate, as it could be one way to better aggregate and surface the U.S.’s maze of intelligence agencies. Today the United States’ intelligence information is spread out across myriad agencies (FBI, CIA, NSA, DIA, ATF, etc.) that don’t have incentive to always work together. PAM would have had actors across these organizations make “bets” on potential world events. No, not with real money – though predictive decision markets get better when the actors have real skin in the game of one sort or another. The U.S.’s solution to the cross agency information problem? Make them all report to a single person, the National Intelligence Director. If Surowiecki’s theory is to be believed (and it’s a convincing argument) this is the exact wrong action – in effect, placing all of the responsibility on a single person (see the CEO-worship section near the end of the book). We can only hope that the predictive decision market has been setup anyway, but that it’s actually classified.
A real predictive decision market that you can go to and play with (and bet real money) is called Intrade Prediction markets. Intrade allows you to bet on real events, including current events, financial index milestones, entertainment, and even the weather. For example, you can bet that on both the Republican and Democratic nominations for the 2008 election, as well as the eventual winner. For example as of today the market predicts a 51% likelihood that Hillary Clinton will win the Democratic ticket (Obama is tracking next at 23.4, and Edwards at around 11, plus other candidates which are more noise.) Just to mark it down, if you were to take the leaders in both Presidential and VP nominations, the Democratic ticket will be Clinton/Obama. Republican ticket will be either Giuliani or Mccain (they’re tracking 31 and 30), with Mitt Romney as V.P. We’ll see. The idea of the market is that as we get closer and closer to an event the market reads information better or sees a given outcome as more likely. Actors in the market can buy and sell contracts at any time.
Predictive decision markets are a really interesting area. Definitely not the last you’ve heard of this topic.
The end of the book describes an instance of crowds not acting in concert for a better benefit, stock bubbles. Surowiecki details multiple examples of stock bubbles, and potential explanations of how and why they are caused (no one knows exactly). One reason (or at least sign that one is happening) is that people are basing their actions on other’s actions, rather than their own information. That is, when people are buying an asset because prices have gone up (opposite usual trend when a price goes up) and they think prices are going to continue to go up – sign of a bubble.
This is a really interesting book – great companion to Freakanomics, The Tipping Point, and The Undercover Economist. Highly recommend it.
