Notes from The Wisdom of Crowds

The Wisdom of Crowds by James Surowiecki (Doubleday 2004 ISBN 0-385-50386-5) explores common assumptions and studies regarding the intelligence of groups, both large and small.

The publisher's website has a short interview which summarizes of the principles in the book. According to this interview, there are four main qualities that promote intelligent group outcomes:
  1. Diverse — different people bring different information
  2. Decentralized — nobody is controlling the final decision
  3. Method of aggregating individual verdicts — such as voting
  4. Independent — people pay attention to themselves, not pressured by others' opinions

Condensed information to help you find the source in the book, arranged by chapter. The chapter names listed here are abbreviated.

Introduction

pg. XI

Francis Galton


In 1906, Francis Galton, British heredity scientist and statician, visited a fair with a "guess-the-weight" contest about an ox. Galton's opinion was that humanity depended on a very few capable individuals. Much to his suprise, the average answer of the crowd was exact. (pgs XI-XIII)
:Francis Galton, "Vox Populi," Nature 75 (March 7, 1907)

Main Point of Book


The argument of this book is that chasing experts is a mistake, and a costly one at that. We should stop hunting and ask the crowd (which, of course, includes the geniuses...) instead. (direct quote, pg XV)

Poor Past Opinion of Collective Wisdom


The author gives examples of people who felt that groups make inferior judgements than individuals: (pgs XV-XVII)

Overview


Types of Problems that Crowds Can Solve
Structure of the Book

First half is theory, dealing with:Second half is case studies, each chapter on a different ways of organizing people, and how it fails or succeeds

Wisdom/Stupidity of Crowds Creates Our World


Our present world if formed various ways that crowds make decisions, both good and bad. The effects of collective decision making are more important than most people realise, and our world is formed by group processes that are mostly poor to bad. Thus, the effects of improving these group processes to produce collective wisdom, instead of stupidity, would be huge.

Different situations/processes make a group smarter or dumber
:Steven Johnson, Emergence (New York, Scribner, 2001)
:Howard Rheingold, Smart Mobs (Boston: Perseus, 2002)
:Kenneth J. Arrow, The Limits of Organization (New York: Norton 1974)
:Thomas C. Schelling, Micromotives and Macrobehaviour (New York: Norton, 1978)

(pgs XIX-XX)

Collaboratively Finding a Submarine


1968, US submarine Scorpian was lost. Officer John Craven created a group of diverse people specialising in many areas, and used Bayes' Theorem to aggregate their opinions-- the result was 220 yards from the actual submarine. (Read Blind Man's Bluff by Sherry Sontag and Christopher Drew for the full story.) (pgs XX-XXI)
:Sherry Sontag and Christopher Drew, Blind Man's Bluff (New York: Public Affairs, 1998): 146-50

1. The Wisdom of Crowds (Part One, THEORY)

pg 3

Who Wants to Be a Millionaire?


A television game show based on multiple-choice questions. Using the "ask an expert" option produced 65 percent correct answers, while "ask the audience" (audience's answer averaged together) was correct 91 percent of the time. (pg 4)

This isn't a scientific study, of course-- no controls. But perhaps it demonstrates a phenomenon which should be studied more. (and it is studied more, see next section) (pg. 04)

Studies on Group Averaged Results

Some sample studies::Kate H. Gordon, "Group Judgments in the Field of Lifted Weights," Journal of Experimental Psychology 7 (1924): 398-400:Jack Treynor, "Market Efficiency and the Bean Jar Experiment," Financial Analysts Journal 43 (1987): 50-53

Two interesting points of these studies:
Normal L. Johnson used a computer simulation of semi-intelligent agents running a maze. He then averaged together all of the agent's routes to produce an "average route." This average route was the shortest possible route. (pg. 6)

Space Shuttle Challenger and Stock Market


A study by Michael T Maloney and J Harold Mulherin notes that when the space shuttle Challenger crashed, the stock market reacted most unfavourably to Morton Thiokol (responsible for the O-ring), even though the commission that identified the O-rings as the problem didn't come out until six months later. (pgs. 7-8) Maloney and Mulherin tried to find inside tips that would have caused the stock market's reaction, but could not find any sign of it.
:Michale T Malony and J Harold Mulherin, "The Stock Price Reaction to the Challenger Crash: Information Disclosure in an Efficient Market" (Dec 7, 1998) http://ssrn.com/abstract=141971

Mathematically, collective judgment is likely to be correct if it comes from independent, decentralized individuals that have various opinions and pieces of information. Errors average out. (pg. 10)

Have 100 people run a race, their average time is worse than the fastest. Ask them to solve a question, and the anwser is greater than (or equal to) the best. (pg. 11)

Collective Wisdom and Predicting the Future


Good decisions need to take the future into account. If collective wisdom can be useful, it needs to be able to deal with uncertainty. This is a different sort of problem than estimating jelly beans. (pg. 11)

Betting and Predicting the Future


Betting is long-time example of finding answers from the average opinion. Bookies adjust the point spread of a bet-- for instance, one team must beat the other by 5 points, vs. not beating the other team by 5 points. This point spread is adjusted so that about half the bets are on either side of the spread: this point spread is determined by public opinion.

And does this aggregation of public opinion match reality? Teams beat the spread about 50%-50% of the time in NFL (American football) games.
:Richard Borghesi, "Price Predictability: Insight from the NFL Point-Spread Market. http://www.cba.ufl.edu/fire/phdstudents/papers/borghesi%20price%20predictability.pdf

A study of horse race betting found that actual ranking of the horses almost always matches the rankings decided by the bettors.
:Raymond D. Sauer, "The Economics of Wagering Markets," Journal of Economic Literature 36 (1998): 2021-64

(pgs. 11-15)

Google and Democratic Ranking


Google uses the collective opinion of the internet to rate pages. The more sites link to a particular site, the higher the rank of that site. Google's useful results are a stunning example of the utility of aggregating collective wisdom. (pgs. 15-17)

Prediction Markets


Prediction markets apply the good mechanics of sports betting to real-world problems, such as who will win an election, or how well a product will do. Decision markets work because of diversity, independence, decentralization. They are less vulnerable to bubbles than the stock market because they are predicting a specific event, at which point the prediction market closes (the stock market is open-ended and tends to try to predict the stock market).

Iowa Electronic Markets (IEM) a prediction market, concerned mainly with elections is 75% more accurate than polls (even months before the election). Polls behave more erratically than IEM.

Hollywood Stock Exchange (HSX) is an online game that predicts box office income etc. Does very well (better than polls) at predicting Oscar nominees and winners.(pgs. 17-22)

2. The Difference Difference Makes

pg 23

Focuses on how diversity helps cognition problems.

Diversity and Evolution in Commerce


Explores how diversity is important for expanding the pool of options. (you can't choose the 'best' answer if somebody hasn't found it yet.)

The author follows Ransom E. Olds' adventures in the auto business (late 1880s-early 1900s) when there were hundreds of companies designing various automobiles, including steam-driven (Stanley Steamer went 127 miles/hour), battery (once, 1/3rd of all US cars were battery), and gasoline. Gasoline-powered car manufacturers were the first to reach the broad (middle-class) audience. The automobile, and the production of the automobile, was developed by all these companies-- Olds began using parts from other companies (outsources) and began mass-production, Cadillac used standardized parts, Ford with the assembly line.
(pgs. 23-25)

New areas of commerce and industry have the same pattern: lots of companies explore lots of possibilities, then the market settles down to the possibilities that paid off. It seems inefficient, but it's actually most efficient when the possibilities haven't been explored yet. (pg. 26)
:(comparing internet business diversity to Cambrian era) Jeff Bezos, Business Week (September 16, 1999) http://www.businessweek.com/ebiz/9909/916bezos.htm

Bees use a the same strategy of exploring lots of strategies before choosing them. Instead of having one bee wander around, looking for flowers, a hive sends out tons of scouts which them communicate their finds to the hive (via the dance). (pg. 26-27)
:Thomas Seeley, The Wisdom of the Hive (Harvard: Harvard University Press, 1996)

Our decentralized economy allows companies to explore many different options, which creates innovation. (pg. 28)
:Rajiv Kumar Sah and Joseph E. Stiglitz, "Human Fallibility and Economic Organization," American Economic Review 75 (1985): 292-97
:Sah and Stiglitz, "The Architecture of Economic Systems: Hierarchies and Polyarchies," American Economic Review 76 (1986): 716-27

Refining option set that results from diversity


A group (as opposed to an individual) creates diverse options, now what about choosing an option? The diversity of groups helps with that, because there are many viewpoints. Markets are automatically diverse because they are open to participation; organisations need to be carefully built to be diverse. (pg. 29)

Diversity in a group is critical: Study by Scott Page found that groups of mixed agents did better than groups made of only smart agents. (pg. 30)
:Page, "Return to the Toolbox" (unpublished)
:Page and Lu Hong, "Problem Solving by Heterogeneous Agents," Journal of Economic Theory 97 (2001): 123-63

James G. March suggests that diversity in organisations (bringing in new people) is more important than redundant expertise (smart old-timers who all know the same stuff). (pg. 31)
:March, "Exploration and Exploitation in Organizational Learning," Organization Science 2 (1991): 71-87

Experts are overrated


The idea that it's better to ask a large group of people instead of one or two experts is ironically counter to popular wisdom. (pg. 31)

Expertise does exist, but it is very narrow. For instance, there are chess experts, but (this should seem obvious, but it's very important to remember) their expertise is only in chess. There is no evidence that expertise carries over into broad areas: somebody might be an expert in applied areas like chess and baking and auto racing, but people aren't expert in broad areas. (pg. 32)
:James Shanteau, "Expert Judgment and Financial Decision Making," paper for Risky Business: Risk Behaviour and Risk Management edited by Bo Green (Stockholm: Stockholm University, 1995)
:Shanteau, "Domain Differences in Expertise," Kansas State University (2002) psych dept. pdf

Experts are often wrong
(pgs. 33-34)

90% mutual fund managers underperformed from 1984-1999.
:John Bogle, John Bogle on Investing (New York: McGraw Hill, 2001): 20

Study by J Scott Armstrong: "I could find no studies that showed an important advantage for exertise." "Expertise and accuracy are unrelated."
:J. Scott Armstrong, "The Seer-Sucker Theory: The Value of Experts in Forecasting," Technology Review 83 (June-July 1980): 16-24

Some studies found that nonpsychologists predicted people's behaviour better than psychologists.

James Shanteau says some studies show that experts decisions are often not self-consistent (about 50% of the time) or consistent with other experts (below 50%).
:Shanteau, "Exert Judgment and Financial Decision Making": 2
:Shanteau, "Why Do Experts Disagree," Risk Behaviour and Risk Management in Business Life, edited by B. Green et al (Dordrecht: Kluwer Academic Press, 2000): 186-96

Experts are also bad at judging how correct they are. A study by Terrance Odean (on doctors, lawyers, engineers, bankers and more) found they knew less than they thought they did.
:Terrance Odean, "Volume, Volatility, Price, and Profit When All Traders Are Above Average," Journal of Finance 53 (1998): 1887-934

A study found that foreign-exchange traders overestimated themselves 70% of the time.

Weathermen and bridge players are the only experts that don't overestimate themselves.

The Problem of Identifying Experts

In most fields, identifying experts (that are worth listening to) from non-experts is very difficult.(pg. 35)

Experts are just as overconfident as other people. (pg. 35)

Why do people think that experts will save them, but tend to think of disparagingly of the averaged results of a group estimate? (pgs. 35-36)
:Richard Larrick and Jack B. Soll, "Intuitions About Combining Opinions: Misappreciation of the Averaging Principle," INSEAD working paper 2003 pdf

Diversity vs. Groupthink


Two big reasons for diversity:
Lack of diversity in groups produce groupthink. It is easier for homogeneous groups become cohesive, for individuals to identify and depend on the group, and to protect group beliefs and opinions from being challenged.
:Irving Janis, Groupthink: Psychological Studies of Policy Decisions and Fiascoes (Boston: Houghton Mifflin, 1982)
:Irving Janis and Leon Mann, Decision Making: A Psychological Analysis of Conflict, Choice, and Commitment (New York: The Free Press, 1977)

The Bay of Pigs invasion of Cuba (1961) is a good example of groupthink. The Kennedy administration planned this invasion, skeptics were hushed. Facts weren't checked with the intelligence departments. The result:(pgs. 36-37)

Groupthink avoids doubts, and information is rationalized away. Enforces the members' belief that their opinion is correct. Enhanced peer pressure to conform.

Solomon Asch's experiment: asked a row of people to verbally judge the length of three (obviously) different lines. When the row of people chose obviously wrong answer, the last person would sometimes go along with the group. A third went with the group more than 50% of the time. 70% of them chose the peer-pressure induced wrong answer at least once.

When Asch asked these people why they chose their answers, they said they knew they were wrong answers, but they didn't want to stand out.

Asch also went to show that it's not difficult to provide opportunities to allow people to go with their judgment and escape peer pressure: just one other person in the row who chose the right answer, the subjects rarely felt the pressure to conform.
:Solomon Asch, Social Psychology (Englewood Cliffs, NJ: Prentice-Hall, Inc., 1952)
:Asch, "Effects of Group Pressure upon the Modification and Distortion of Judgments," Groups, Leadership and Men, edited by Harold Guetzkow (New York: Russell & Russell, 1963 919510): 177-90

Thus, diversity is also important because it lets individuals feel comfortable to be indpendent from group beliefs and to express what they really think. (pgs. 37-39)

3. Imitation, Information Cascades, and Independence

pg 40

Lack of indepedence creates stupidity


While imitation (and information cascades) can play useful roles in group behaviour (especially for coordination problems, such as movement) they can become harmful to cognition problems.

An example of imitation gone awry is army ants walking in a circle until they starve (circular mill).
:William Beebe, Edge of the Jungle (New York: Holt, 1921)

Ant colonies normally work very well. Each ant is rather mindless, but out of the collective behaviour of all of the ants, intelligent behaviour emerges. However, ants' lack of independence makes them vulnerable to problems such as the circular mill (above).
:Steven Johnson, Emergence (New York: Scribner, 2001). (about behaviour in human society, too)
(pg. 40)

The indepence of humans allows the society of humans to produce more intelligent decisions than the society of ants. Most examples of collective intelligence given in this book rely on people making independent contributions and thinking different things.

Independence is important because:
The best collective intelligence comes from groups with members that have diverse points of view and can stay independent of each other (instead of getting caught up in a group bias or control). Irrational individuals won't hurt the collective intelligence as long as they are independently (uniquely?) irrational.
(pg. 41)

Independence is lauded in Western society, but unfortunately this is mostly talk. Media, peers, teachers, employment, etc. often decrease our independence and shape our beliefs and desires.
:Kenneth J. Arrow, "Methodological Individualism and Social Knowledge," American Economic Review 84.2 (1994): 1-9
:Herbert J. Simon, Administrative Behaviour, 3rd ed., (New York: Free Press, 1976): xvi
:Mark Granotvetter, "Economic Action and Social Structure: The Problem of Embeddedness," American Journal of Sociology 91 (1985): 481-510
:Ronald S. Burt, Structural Holes: The Social Structure of Competition (Cambridge: Harvard University Press, 1992)
Even if the loss of independence makes us individually smarter (because we echo other's good habits) we lose diversity and collectively become less smart.
(pg.42)

How can we maintain independence despite the influence of society, and thus create collectively intelligent decisions?
(pg. 43)

Lack of Independence: Herding Behaviour


A study placed groups of people on a street corner looking up at the sky.This is an example of people thinking that if a lot of other people are doing something, there must be a good reason for it. (this isn't an example of peer pressure) This tendency can be useful, unless everybody adopts it and we lose diversity.
:Stanley Milgram, Leonard Bickman, and Lawrence Berkowitz, "Note on the Drawing Power of Crowds of Different Size," Journal of Personality and Social Psychology 13 (1969): 79-82
(pg. 43)

NFL has a common strategy about choosing field goals over first downs. It's "conventional wisdom" to take the field goal, but it's "collective wisdom" to take the possible first down, since analysis of game histories show it is more likely to produce better results.
:David Romer, "It's Fourth Down and What Does the Bellman Equation Say?" working paper, University of California, Berkeley (2003) http://papers.nber.org/papers/w9024
Instead, conventional wisdom is to take the cautious, but foolish, strategy. Why? This is probably an example of how imitation and lack of indepent behaviour produces poor decisions. Coaches don't want to go against conventional wisdom; that's regarded as risk. Herding, "Safety in numbers," "no one ever got fired for buying IBM."
:Michael Lewis, Moneyball (New York: Norton, 2003)
(pgs. 44-49)

You see herding behaviour in mutual fund managers, too. This is because they have to convince clients that they are investing well (so they can get clients to invest). By making conventional choices, clients know that the manager isn't irrational. Thus, apparently rational behaviour is more important than actually rational behaviour.
:David S. Scharfstein and Jeremy C. Stein, "Herd Behavior and Investment," American Economic Review 80 (June 1990): 465-79

Information Cascades


"Worldy wisdom teaches that it is better for reputation to fail conventionally than to succeed unconventionally."
:John Maynard Keynes, The General Theory of Employment, Interest, and Money (New York: Harbringer, 1964): 158

In the US, the story of plank-road fever is a warning of information cascades. Because of initial success, people thought that plank roads would be profitable (build a plank road, charge a toll) and people built tons of them before it was discovered that they rotted too soon to make profit. (pgs. 51-53)
:Daniel B. Klein and John Majewski, "Plank Road Fever in Antebellum America: New York State Origins," New York History (January 1994): 39-65

Information cascades can occur where people pay attention to what others have chosen before making their decision. If the first people choose a particular dish off of the menu, other diners may follow suit, because they suspect there is a good reason that previous people made that decision. This habit may cause good decisions to cascade, but also allows bad decisions to cascade. (what if the first diners had ordered fish pie on a whim?)

If everybody has decided one way on a particular issue, it is rational to assume that they have done this for a good reason. But if everybody else is making that decision because they assumed that the other people had a good reason, then it doesn't work. (pgs. 53-55)
:Sushil Bikhchandani, David Hirshleifer, and Ivo Welch, Informational Cascades and Rational Herding
:Bikhchandani, Hirshleifer, Welch, "Learning from the Behaviour of Others: Conformity, Fads, and Informational Cascades," Journal of Economic Perspectives 12 (1998): 151-70
:Abhijit V. Banerjee, "A Simple Model of Herd Behaviour," Quarterly Journal of Economics 107 (1992): 797-817
:H. Henry Cao and David Hirshleifer, "Conversation, Observational Learning, and Informational Cascades," Dice Center working paper 2001 http://papers.ssrn.com/sol3/papers.cfm?abstract_id=267770
:Suzanne Lohmann, "The Dynamics of Informational Cascades: The Monday Demonstrations in Leipzig, East Germany, 1989-91," World Politics 47 (1994):42-101
:Duncan Watts, Six Degrees (New York: Norton, 2002) [very detailed account of mechanics of information cascades according to James]

Gladwell's excellent book looks at the social mechanics of information cascades-- how the relationships and roles in society cause cascades (epidemics in Gladwell's terminology). (pg. 55)
:Malcolm Gladwell, The Tipping Point (New York: Little, Brown, 2000)

The spread of the standardized screw in the 1860s in the U.S. is an example of a beneficial cascade. Previously, machinists would lock-in customers to their custom screws. However, the machinist who caused the cascade, William Sellers, used his influence and detailed knowledge of the psychology of the screw market to cause this cascade. In other words, his product succeeded because he knew how to do the cascade, not because he had a superior product. (luckily for the U.S., he also had a superior product.) Thus, cascades aren't a recipe for intelligent behaviour. (pgs. 56-57)
:James Surowiecik, "Turn of the Century," Wired 10.01 (January 2002) http://www.wired.com/wired/archive/10.01/standards_pr.html

Good decisions result from independence and diversity. Cascades aren't that. They depend on who went first and influential people. (pg. 57)

The internet telecommunications bubble demonstrates the bad effects of cascades. The idea that internet traffic was growing became so prevalent that people stopped following reality and overbuilt.
(pgs. 57-58)

Intelligent Imitation


Cascades and imitation are a convenient way of making decisions. You can follow what other people are doing (such as taking an umbrella today) and do pretty well. Also, it's easier to take information without checking it, especially if it's not that important and the cost of getting in wrong is small.

The decision to imitate can sometimes be an example of collective wisdom. For example, researchers of the macaque monkeys of the Koshima islands observed monkeys imitating one monkey's discovery of washing sweet potatoes and wheat. Selecting what to imitate is often a collective decision. (pgs. 58-60)
:Lee Dugatkin, The Imitation Factor (New York: The Free Press, 2001): 170-72

A study (computer simulations) suggests that the whole society benefits with some imitation, but tempered with individuals willing to decide for themselves to decide which way is better. (pg. 60)
:Robert Boyd and Peter J. Richerson, "Norms and Bounded Rationality," in Bounded Rationality: The Adaptive Toolbox, edited by Gerd Gigerenzer and Reinhard Selten (Cambridge: MIT Press, 2001): 281-96 [James says Boyd and Richerson have also written papers important to trust and cooperation]

How do we tell apart smart vs. dumb imitation?:Antonio Bernardo and Ivo Welch, "On the Evolution of Overconfidence and Entrepreneurs," Cowles Foundation discussion paper no. 1307 (2001)

A 1943 study found that most Iowa farmers didn't adopt a new type of corn even though the information was good-- they waited until others had tried it first. After that, then they personally tested the corn in a small patch. This wasn't a mindless cascade. (pgs. 61-62)
:Bryce Ryan and Neal Gross, "The Diffusion of Hybrid Seed Corn in Two Iowa Communities," Rural Sociology 8 (1943): 15-24
:Everett Rogers, The Diffusion of Innovations (New York: The Free Press, 1983) [Similar topic]
:J. S. Coleman, H. Menzel, and E. Katz, "The Diffusion of an Innovation Among Physicians," Sociometry 20 (1957): 253-70

A similar study of Indian farmers found wheat farmers (whose fields all have similar growing conditions) relied heavily on their neighbors when making seed choices. Rice farms have much more variety, so neighbors' choices weren't so important: they individually experimented first (much like the Iowan corn farmers). Here is an example of an information cascade (using other's decisions) vs a non-cascade (using private information). (pg. 62)
:Kaivan Munshi, "Social Learning in a Heterogeneous Population: Technology Diffusion in the Indian Green Revolution," Brown University working paper (2003), [[http://www.econ.brown.edu/fac/Kaivan_Munshi/ag6.pdf|]]

Fashion is cascade-driven by definition: it attempts to guess what other people like. Cascades are more likely in less important areas of life. (cautious farmers vs. fashion fads is a good example) (pg. 63)

Cascades vs. Simultaneous Decisions


The lack of diversity and independence caused by cascades are caused by sequential decisions-- if everybody chose simultaneously, it would improve the group result.

For individuals, however, cascade (herding) behaviour is still better than individual behaviour (a classroom study found that it was 80% better than an individual's non-collective result. [below]) (pgs. 63-64)

A classroom study had two urns, each containing mixed black and white marbles. One urn had 2/3rd black marbles, the other predominantly white marbles. Each student would select one marble from an urn, and then guess what the predominant colour was. Obviously, if they chose based on their private information (their marble) they would be correct 2/3rds of the time. Each student couldn't tell the other students the colour of their marble, but the other students would hear their guess.

Thus, the best strategy (for maximum individual gain) was to pay more attention to what other people guessed, and less attention to your private information (the marble you saw). Students did just this, and cascades occured 78% of the time.

When the researchers said that individuals couldn't win individually, but only in groups (if the average answer of your group was correct), then students wisely adopted a group-beneficial strategy of being more individualistic: guess the colour of your marble, and ignore the previous guesses. With this strategy, individuals were more often wrong, but the group was more often right. (pgs. 64-65)
:Angela Hung and Charles Plott, "Information Cascades: Replication and an Extension to Majority Rule and Conformity-Rewarding Institutions," American Economic Review 91 (2001): 1508-20

Simultaneous decision-making (so that you don't know what others have decided) is a good recipe to encourage independence and diversity and prevent cascades and stupidity. (pg. 65)

4. Putting the Pieces Together / Decentralization

pg 66

1946 criticism of the U.S. national intelligence agencies are given as an example of lack of integration. (pgs. 66-69)

Examples of decentralisation: self-managed teams, ant colonies, beehives, social networks, the internet. (pgs. 69-70)

People's idea of decentralization can be unformed. When people praise or criticise decentralization, they may be assigning different meanings to the word. (pg. 70)

Pros of decentralisation:Con of decentralisation::The author suggests we need good information aggregation methods to counter this, perhaps with a balance of decentralisation and centralisation (choosing between them intelligently depending on the need at hand) (pgs. 71-72)

Linux


Linux is a good example of decentralisation: isn't owned: no bosses, no commanding. People work on it out of interest. It's turned out remarkably well, making Linux the largest competitor to Microsoft.

This system makes excellent use of:
The decentralised system of Linux works because there is an aggreation process: various people or groups (depending on the project) choose from the diversity of improvements discovered by the community and integrate them back into their particular project. (pgs. 72-75)

Decentralisation isn't the solution for all problems (traffic jams, uncoordinated intelligence agencies, etc.). Also, the particular kind of decentralisation is important. It's important that all the decentralised parts can use the information and wisdom of the other parts (link computer databases, etc.) (pgs. 75-83)

5. Coordination in a Complex World

pg 84

How do people coordinate individual actions to meet their individual needs while the groups functions smoothly? This usually works well in practice-- examples include pedestrians coordinating themselves on a busy sidewalk (pgs. 84-85), or commerce.

Coordination problems are fundamentally different from cognition problems, because individuals need to be aware of what other people are doing. (pg. 86)

Two researchers suggested that individuals could reach beneficial solutions by act independently. For example, in a bar where people only enjoy their time if the bar is about 60% full, individuals form strategies from their past experiences. (pgs. 86-90)

Studies by Thomas Schelling found that people could often coordinate without communicating with each other.:Thomas C. Schelling, The Strategy of Conflict (Cambridge: Harvard University Press, 1960): 54-67

Conventions and coordination


Solutions to coordination problems can quickly become conventions. Professor Schelling taught in a building with two stairways, both of them jammed with students going both directions. He asked his students to take one stairway down and the other up to see what happened. In three days, all the other classes had adopted the same behaviour. Conventions can allow efficient, convenient solutions to coordination problems (students no longer have to figure out how to get past the student in front of them-- convention avoids that problem). (pg. 92-93)
:Jonathan Rauch, "Seeing Around Corners," The Atlantic Monthly 289 (April 2002): 35-48

The most efficient method for deciding who gets what seat and gets served in a queue (first come, first serve) is efficient because it requires the least thinking. (pgs. 93-94)

Stanley Milgram in 1980s had his class experiment with asking subway riders for their seats. About 50% of the time they people they asked would give them his or her seat. His students reported that the most difficult aspect of the process was overcoming their own inhibition to brazenly ask for the seat. (pgs. 94-95)
: The Individual in a Social World, edited by Stanley Milgram (New York: McGraw-Hill, 1992): xix-xxxiii

Milgram's experiment on cutting into queues found that:
A single queue is much better than multiple queues (such as supermarkets) since a single queue is first-come-first-served and doesn't take much cognitive effort (convention based, not cognitive). (pgs. 95-96)

Inefficient Conventions


Convention is often inefficient. Laying off employees vs. cutting wages, sharecropping, pricing goods, movie theatre pricing are all examples of conventions which don't coordinate well with reality. (pgs. 97-101)
:Liran Einav and Barak Orbach, "Uniform Prices for Differentiated Goods: The Case of the Movie-Theater Industry," Harvard Olin discussion paper no. 337 (2001)

Coordination in flocks and markets


A flock of birds coordinates evasion from a hawk with this strategy:
The flow of goods in a market is a similar phenomenon. When you look for an item at the supermarket, and it's almost always there. The components of the supply chain (farmers, packagers, etc.) act with local rules (like the flock of birds) and the result is coordinated and meets people's demands. The free market is a mechanism to solve coordination, much like flocking. The successful behaviour of the market emerges from independent parts, not from central planners controlling everything. (pgs. 101-103)

Experiments by Vernon L. Smith show that buyers and sellers with very little information would quickly produce a market that reached mutually agreeable prices. This is notable because, although theories predict this result (beginning from Adam Smith's Wealth of Nations), these theories presumed agents will complete information. Smith's experiments proved that free markets still coordinates well even with diverse individuals working from independent information. (pgs. 103-107)

:Vernon L. Smith, Bargaining and Market Behavior (Cambridge: Cambridge University Press, 2000)
:Kenneth J. Arrow and Gerard Debreu, "Existence of an Equilibrium for a Competitive Economy," Econometrica 22 (1954): 265-90
:Vernon L. Smith, "Constructivist and Ecological Rationality in Economics," American Economic Review 93 (2003): 465-508


The work of Smith shows that individuals can act selfishly and coordinate for mutual benefit.

6. Taxes, Tipping, Television, and Trust

pg 108

This chapter is about cooperation: people working together to achieve a common benefit.

Coordination vs Cooperation

The solving of coordination and cooperation problems both require that participants have awareness of each other's actions. However, coordination problems (such as markets and walking on a busy sidewalk) are easily solved by people working for their own interests, while cooperation problems need people to work together for the group's greater good-- sometimes difficult since an group members which aren't helping still receive the benefits. (pg. 110)

Ultimatum Game

This game has two players-- one player decides how to split $10, and the other player either:Low offers are usually refused: people would rather give up a few bucks and see a greedy person punished. 50% splits were most common. (Not only humans did this; a study on monkeys observed similar behaviour.) This behaviour is notable in that it shows people are willing to give up personal benefits in order to improve society as a whole. (pgs 112-113)

:Vernon L. Smith, "Constructivist and Ecological Rationality in Economics," American Economic Review 93 (2003): 465-508
:Alvin E. Roth, Vesna Prasnikar, Masahiro Okuno-Fujiwara, and Shmuel Zamir, "Bargaining and Market Behaviour in Jerusalem, Ljubljana, Pittsburgh, and Tokyo: An Experimental Study," American Economic Review 81 (1991): 1068-95 [examines ultimatum game in different countries]
:Sarah F. Brosnan and Frans B. M. de Waal, "Monkeys Reject Unequal Pay," Nature 425 (2003): 297-99


The ultimatum game of how people go out of their way to encourage fairness. It's different in different countries; for example, American CEOs earn more than Japanese CEOs, and high incomes don't bother Americans as much as Europeans.
:Alberto Alesina, Rafael di Tella, and Robert MacCulloch, "Inequality and Happiness: Are Europeans and Americans Different?," National Bureau of Economic Research working paper no. 8198 (2011)

Tackling the cooperation problem can be expressed in two different extremese::rather than acting at personal cost that would benefit everybody:acting at personal cost that benefits everybody: example is the ultimatum game, where participants' refusal of money doesn't help them (they don't get the money) but it makes sure that the social norm is to try to play fairly

Prosocial Behaviour and Cooperation


Cooperation depends on prosocial behaviour-- an awareness of the future and the big picture, and a tendency to discourage or punish noncooperative behaviour. Making non-selfish decisions in the short-term produce long-term benefits. Trust and cooperation in economics, and behaviour like tipping, are examples of this. (pgs. 116-118)

:Samuel Bowles and Herbert Gintis, "Prosocial Emotions," Santa Fe Institute working paper no. 02-07-028 (2003)
:Robert J. Shiller, Maxim Boycko, and Vladimir Korobov, "Popular Attitudes Toward Free Markets: The Soviet Union and the United States Compared," American Economic Review 81 (1991): 385-400
:Rick L. Riolo, Michael D. Cohen, and Robert Axelrod, "Evolution of Cooperation Without Reciprocity," Nature 414 (2001): 441-43 http://www-personal.umich.edu/~axe/research_papers.html (see also "The Role of Social Structure in the Maintenance of Cooperative Regimes" (2001) and "On Six Advances in Cooperation Theory.") (2000)


Why do we cooperate with strangers? Robert Wright suggests that it is because we have internalized the reality that trade is not a zero-sum game: both parties can win.
:Robert Wright, Nonzero: The Logic of Human Destiny (New York: Pantheon, 2000)

Trust is Vital to Commerce


Apparent paradox of cooperation problems::because acting would cost ourselves and only benefit society:because punishing antisocial behaviour (as seen in the Ultimatum Game) comes at personal cost, but benefits the group

Commerce is very difficult without trust-- that your goods will not be stolen, etc. Developed countries that rely on commerce tend to be more respectful of property. Commerce during Europe's medieval era was difficult-- it took extreme measures, like embargo, to prevent ships from being confiscated by the locals. People tended to trust people only in the same clan or locality. (pg 122)
:Avner Greif, "Contract Enforceability and Economic Institutions in Early Trade: The Maghribi Traders' Coalition," American Economic Review 83 (1993): 525-48
:Avner Greif, "Self-Enforcing Political Systems and Economic Growth: Late-Medieval Genoa," Analytic Narratives, edited by Robert H. Bates et al. (Princeton: Princeton University Press, 1998): 23-63

Quakers were trustworthy, thus they ran a large portion of the British economy in the 1700-1800s-- over half of the ironworks, lots of banking (Barclays and Lloyds) etc. This is an example of how trust and honesty pay of well in the long-term. (pg. 119)
:James Walvin, The Quakers, Money and Morals (London: Trafalgar Square, 1998)

In the 1800s, the idea that trust = profit began to spread through the public awareness. It's good business to treat your customers fairly instead of trying to maximize each transaction-- they come back. (pg. 122)
:John Mueller, Capitalism, Democracy, and Ralph's Pretty Good Grocery (Princeton: Princeton University Press, 1998)

Establishing a system of honor and trust allows the economy to flourish. Experiences in undeveloped countries sometimes show how the type of cooperation and honesty (social capital) that enables modern business may not be well developed. (pgs. 121-122)

:Thomas Schelling, Choice and Consequence (Cambridge: Harvard University Press, 1985): 210
:Douglass C. North, "Economic Performance Through Time," American Economic Review 84 (1994): 359-68
:Kenneth J. Arrow, "Observations on Social Capital," Social Capital: A multifaceted Perspective, edited by Partha Dasgupta and Ismail Seregeldin (Washington, DC: The World Bank, 1999)
:Arrow, "The Economics of Moral Hazard-- Further Comment," American Economic Review 58 (1968): 537-38


By adopting impersonal trust, trade in enabled outside of tight, homogenous groups. Relationships in commerce are now less personal, but they allow greater diversity and focus on mutual profit. (pg. 123)
:Clifford Geertz, "The Bazaar Economy: Information and Search in Peasant Marketing," American Economic Review 68.2 (1978): 28-32

"Individuals in higher-trust societies spend less to protect themselves from being exploited in economic transactions. Written contracts are less likely to be needed, and they do not have to specify every possible contingency." Stephen Knack and Philip Keefer, "Does Social Capital Have an Economic Payoff? A Cross-Country Investigation," Quarterly Journal of Economics 112 (1997): 1251-88

People may think that markets are about greed, not trust. But studies with tribal societies show that groups with greater interaction with markets demonstrated greater trust and prosocial behaviour. (pgs. 125-126)

:Joseph Henrich et al., "'Economic Man' in Cross-Cultural Perspective: Behavioral Experiments in 15 Small-Scale Societies," Santa Fe Institute working paper (2001) http://www.santafe.edu/research/publications/wpabstract/200111063
:Henrich et al., "In Search of Homo Economicus: Experiments in 15 Small-Scale Societies," American Economic Review 91 (2001): 73-78


Trust but Verify


The stock market bubble in the U.S. in the late 1990s shows that trust without verification can have bad results. The institutions that were supposed to be verifying corporation's honesty were no longer verifying, and people weren't checking on those institutions. Many CEOs realised that the long-term picture was bad, but they could make short-term profits by cheating. (pgs. 126-128)

Lack of cooperation in TV advertisers


Advertisers in TV industry use Nielsen Media Research's statistics to determine who is whatching what and when. The current system has many problems, but nobody will make the move to make the expensive overhaul-- such a move would cost the company who did this, but benefit the entire industry. This is an example of failed cooperation: individually rational, collectively dumb. (pgs. 128-135)

Taxes


Taxes are a coordination problem: they cost the individual, but benefit everybody (roads etc.). People who don't pay still benefit: there are not individual rewards for paying. Yet Americans usually comply and pay (more often than Europeans). Margaret Levi attributes this to the attitude that people don't mind chipping in if they think that others are also helping. (pgs. 137-138)

:Margaret Levi, "A State of Trust," in Trust and Governance, edited by Valerie Braithwaite and Margaret Levi (New York: Russell Sage Foundation, 1999)
:Joel Slemrod, "On Voluntary Compliance, Voluntary Taxes, and Social Capital," National Tax Journal 51 (1998): 485-91


A study found that people are happier about paying taxes if they trust the government to spend their tax money well. (pg. 140)
:John T. Scholz and Mark Lubell, "Adaptive Political Attitudes: Duty, Trust, and Fear as Monitors of Tax Policy," American Journal of Political Science 42 (1998): 903-20
:Scholz and Lubell, "Trust and Taxpaying: Testing the Heuristic Approach to Collective Action," American Journal of Political Science 42 (1998): 398-417

Public Goods Game


Each player begins with tokens. Each turn, if they invest a token, every player earns 0.4 tokens. Thus, if everybody invests tokens, the whole group is richer. Players tend to be prosocial unless they see other people acting selfishly.

A study suggests three types of people:
The researchers modified the game so that everybody was aware of who was contributing or not, and added the ability to punish the free-riders (use part of a token to remove one of their tokens).:Ernst Fehr and Simon Gächter, "Cooperation and Punishment in Public Goods Experiments," American Economic Review 90 (2000): 980-94

7. Traffic / Failure to Coordinate (Part Two, EXAMPLES)

pg 145

Bad London traffic was solved by charging people fees to drive into London during busy times ("congestion pricing"), and put the money raised towards building public transportation. It worked. Attempts in other places for congestion pricing (such as San Francisco) have met strong resistance.(pgs. 145-146)
:C. Robin Lindsey and Erik T. Verhoef, "Traffic Congestion and Congestion Pricing," Tinbergen Institute discussion paper TI2000-101/3 (2000)

Congestion pricing has been adopted in non-auto areas: early-bird specials, happy-hour specials, cheaper telephone at night, etc.

Singapore has had congestion pricing on its roads since 1975. It automatically deducts money from your account when you drive, depending on the zone and the time. This alleviates problems (and Signapore has the highest cars per mile) and leaves the choices in the hands of the driver. Mexico tried a different strategy that tried stricter control (banning certain cars on certain days) but it didn't work. Signapore's strategy sets the price and allows individual decisions to produce a collectively good result. Dan Baum, "The Ultimate Jam Session," Wired 9.11http://www.wired.com/wired/archive/9.11/singapore_pr.html

Congestion pricing isn't fair since poorer people have less decisions. This can be countered-- for example, Signapore uses the fees it collects to build superb public transit. Carlos Daganzo suggests a system that allows people free days (one seventh can drive on each day for free) so that poorer people have a chance, while still utilising the benefits of congestion pricing. The author suggests an ideal system that charges people depending on the current congestion of a road. (pgs. 147-149)

Traffic fails to solve the coordination problem and results in frequent jams. This is built into the way the behaviour of individual drivers adds up; there doesn't seem much to do about it.
(pgs. 150-153)
:Philip Ball, "Further On Down the Road," Nature (November 26, 1998)

Diversity is good for cognition problems, but not coordination problems. An simulation clearly demonstrated this: regularly released cars and regular speeds produced smooth flow, but variations created jams. (pg. 153)
:Mitch Resnick, Turtles, Termites, and Traffic Jams

California's PATH program experimented with computer controlled cars in 1997. The cars were automatically synchronized. Such a system would remove/reduce traffic jams. (pgs. 154-155)

8. Science: Collaboration, Competition, and Reputation

pg 158

Collaboration and Success in Science

It's normal for scientists to collaborate, research and publishing in large groups. For instance, credit for the discovery of the "top quark" goes to 450 physicists. (pg 161)

The most successfull scientists are very collaborative. (pg 162)

:Barry Bozeman and Sooho Lee, "The Impact of Research Collaboration on Scientific Productivity," prepared for the annual meeting of the American Association for the Advancement of Science (February 2003): 24-25
:Paula Stephan, "The Economics of Science," Journal of Economic Literature 45 (1996): 1220-21
:Etienne Wenger, Richard McDermott, and William M. Snyder, Cultivating Communities of Practice (Boston: Harvard Business Scholl Press, 2002): 10
:D. J. de Solla Price and Donald B. Beaver, "Collaboration in an Invisible College," American Psychologist 21 (1966): 1101-17


Collaboratively discovering SARS


Laboratories collaborated world-wide to pinpoint the virus that caused SARS. They discussed their work and shared data daily. Together they discovered the virus in weeks-- individually it would have taken months or years. (pgs 158-160)
:WHO, "A Multicentre Collaboration to Investigate the Cause of Severe Acute Respiratory Syndrome," The Lancet 361 (2003): 1730-33

Motivating Factors and Collaboration


Reputation is the main "coin" of the scientific community, not money. Science is both collective and cumulative. Scientists work on their particular interests, then share their information, gaining reputation and helping others. This independent approach is normal. Nowadays, however, scientists working under the commands of a corporation (where money, rather than interest, are presumably more important) are becoming more common.:Robert K. Merton, "The Matthew Effect," Science 159 (1968): 56-63 pdf

Isaac Newton's famous statement If I have seen further, it is by standing on the shoulders of giants isn't as pro-collaborative as it may seem at first: it doesn't admit to gaining knowledge from contemporaries, and it only credits the "giants," or those with already large reputations.

Henry Oldenburg

Henry Oldenburg pioneered the idea of open science-- that sharing what you discover, rather than hiding it. Knowledge, unlike solid possessions, does not lessen when you give it away. When scientists share their discoveries, not only do they get recognition for their work, but other people can refine, disprove or build off of their discoveries. (pg 166)
:Lisa Jardine, Ingenious Pursuits (New York: Doubleday, 1999)
:Joel Mokyr, The Gifts of Athena: Historical Origins of the Knowledge Economy (Princeton: Princeton University Press, 2002): 36, 54

Modern science depends on::To save time, experiments are usually accepted on how likely they seem, rather than by repeating the experiment yourself. Occasionally, this results in a poorly done study becoming accepted and entrenched, and experiments that disagree with an accepted study often doubted.

Reputation is more important than the idea

Unfortunately, the reputation of the scientist doing the work usually have more effect than the value of the discoveries themselves. People will pay much more attention to the work of famous scientists. This has a poor effect on science. (pgs 170-172)
:Robert K. Merton, "The Matthew Effect II: Cumulative Advantage and the Symbolism of Intellectual Property," Isis 79 (1988): 606-23

9. How Small Groups Can Be Made to Work

pg 173

The space shuttle Columbia disaster is an example of bad small group dynamics. (pgs. 173-182):James Glanz and John Schwartz, "Engineers: NASA Leaders Ignored Safety Pleas," The New York Times (September 26, 2003)

One of the biggest problems in small group dynamics for the Columbia was the hierarchical structure of the group. (pg. 182)
:N. R. F. Maier and A. R. Solem, "The Contribution of a Discussion Leader to the Quality of Group Thinking: The Effective Use of Minority Opinions," human Relations 5 (1952): 277-88

A common small group problem is to begin with the conclusion, then fit new evidence into the conclusion. A study found that people tend to only talk about information that they have in common (instead of sharing information they don't know) and, alarmingly, this leads decisions coming from casual discussions to ignore the bulk of the private information. (pg. 183)
:Garold Stasser, "The Uncertain Role of Unshared Information in Collective Choice," Shared Knowledge in Organizations, edited by Leigh L. Thompson, John Levine, and David Messick (New York: Lawrence Erlbaum, 1999).

Identity issues

A small group has an identity-- people feel their membership of small groups much more vividly than their membership in big groups. The more you identify with the group, the harder it is to disagree.
(pg. 176)

Counter this with:*Diversity of opinions are very good-- lots of viewpoints.

Group polarization


Diversity is vital in small groups. A metastudy by Chandra Nemeth found that even one minority opinion (even a poor one) can improve the group process in mock juries. Unfortunately, without this diversity, group discussion becomes extreme: opinions are strengthened, positions become more extreme. This unfortunate process is called group polarization. This may happen because people compare their position to others' positions; perhaps without a spread of opinion to compare themselves to, people's bearings communally shift off to one direction.

:Cass Sunstein, Why Societies Need Dissent (Cambridge: Harvard University Press, 2003)
:D. G. Myers and M. F. Kaplan, "Group-Induced Polarization in Simulated Juries," Personality and Social Psychology Bulletin 2 (1976): 63-66
:A. Vinokur and E. Burnstein, "Novel Argumentation and Attitude Change: The Case of Polarization Following Group Discussion," European Journal of Social Psychology 8 (1978): 335-48

Problems with people having too much influence

The strength of the proponents of an idea may not indicate the validity of the idea. These kinds of people tend to talk more (and be listened to more)::E. P. Torrence, "Some Consequences of Power Differences on Decisions in B-26 Crews," United States Air Force Personnel and Training Research Center research bulletin 54-128 (1954)
Extremists tend to be more overconfident and stubborn.
:Brock Blomberg and Joseph Harrington, "A Theory of Flexible Moderates and Rigid Extremists with an Application to the U.S. Congress," American Economic Review 90 (2000): 605-20

Successful small groups


A study arranged people into groups of six, that were made of two sub-groups (3+3) that had opposing viewpoints. This helped polarization and increased scores on factual questions.
:E. P. Torrence, "Some Consequences of Power Differences in Permanent and Temporary Three-Man Groups," Small Groups, edited by A. P. Hare et al. (New York: Knopf, 1955)

Small groups with good dynamics usually make better decisions than their smartest member.

A study tried to see if small groups could make good decisions as quick as individuals. Two experiments:
Identifying shifts in pattern
Setting interest rates in a simulation of central banking:
:Alan S. Blinder and John Morgan, "Are Two Heads Better Than One? Monetary Policy by Committee," National Bureau of Economic Review working paper 7909 (2000)
A Bank of England study confirmed the above conclusions. (pgs. 189-190)
:Clare Lombardelli, James Proudman, and James Talbot, "Committees versus Individuals: An Experimental Analysis of Monetary Policy Decision-Making," Bank of England working paper 165 (2002) http://www.bankofengland.co.uk/wp/

Groups have a reputation for making decisions slowly; this is incorrect, since slowness and incorrectness etc. derives from poor dynamics (lack of diversity, no facilitation structure for speed, etc.) Also, groups don't matter whose decisions don't matter (because they have no read control, as in the hierarchical Columbia control).

10. The Company / Same as the Old Boss?

pg 192

Zara


Zara is a Spanish clothing company that sells worldwide.:David Bovet and Joseph Martha, Value Nets (New York: John Wiley & Sons, 2000)

Zara can do these amazing things because it coordinates so well with its customers and suppliers. It takes advantage of the collaborative functioning of markets. Then why does it need to control its employees? Why not outsource everything? (pgs 195-196)

Ronald Coase


In The Nature of the Firm (1937) Ronald Coase said that outsourcing takes more energy than doing it in-house. Transaction costs-- finding people, determining a price, agreeing, billing, etc.-- the costs of the market-- are averted by building a non-market enclosure within a company. (pg 196) (read more at wikipedia)

Inhouse vs. Outsource


Inhouse advantages:
Outsource advantages:(pgs 196-197)

11. Markets

pg 224

12. Democracy: Dreams of the Common Good

pg 259