SFI BulletinSFI BulletinS A N T A F E I N S T I T U T E • F A L L 1 9 9 9 • V O L U M E 1 4 • N U M B E R 2

H ET ERA RCHYI N T H E A L L E Y

F A L L 1 9 9 9 • V O L U M E 1 4 • N U M B E R 2

SFI Bulletin

The Bulletin of the Santa Fe Institute is published bySFI to keep its friends and supporters informed aboutits work. The Bulletin is free of charge and may beobtained by writing to the managing editor at theaddress below.

The Santa Fe Institute is a private, independent,multidisciplinary research and education centerfounded in 1984. Since its founding, SFI has devoteditself to creating a new kind of scientific researchcommunity, pursuing emerging synthesis in science.Operating as a visiting institution, SFI seeks tocatalyze new collaborative, multidisciplinary research;to break down the barriers between the traditionaldisciplines; to spread its ideas and methodologies toother institutions; and to encourage the practicalapplication of its results.

Published by the Santa Fe Institute 1399 Hyde Park RoadSanta Fe, New Mexico 87501, USA Phone (505) 984-8800 fax (505) 982-0565 home page: http://www.santafe.edu

Note: The SFI Bulletin may be read at the website:www.santafe.edu/sfi/publications/Bulletin/.If you would prefer to read the Bulletin on your com-puter rather than receive a printed version, contact Patrisia Brunello at 984-8800, Ext. 269 orpdb@santafe.edu

EDITORIAL STAFF:Ginger RichardsonLesley S. KingAndi Sutherland

CONTRIBUTORS:Cosma Rohilla ShaliziJanet StitesHollis Walker

DESIGN & PRODUCTION:Patrick McFarlin

In this issue

6

8

COVER ILLUSTRATION: PATRICK MCKELVEYBASED ON PHOTO BY JULIE GRABER P

HO

TO:

JULI

E G

RA

BE

RP

HO

TO:

HA

WLE

Y J

OH

NS

ON

2

S A N T A F E I N S T I T U T E B U L L E T I N • F A L L 1 9 9 9 1

FEATURES

The Organization of Diversity—Travels with Sociologist David Stark . . . . . .2

Discovering Patterns—The Interface Between Art and Science . . . . . . . . . .9

Homo Reciprocans: Political Economy and Cultural Evolution . . . . . . . . . .16

Pursuing Complexity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .21

WORK IN PROGRESS

Kevin Bacon, the Small-World, and Why It All Matters . . . . . .center section

NEWS

SFI Artificial Stock Market Soon Online . . . . . . . . . . . . . . . . . . . . . . . . . .6

SFI Trustee John Powers, 1916-1999 . . . . . . . . . . . . . . . . . . . . . . . . . . .7

New Books From Oxford . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .8

Oprea Earns Ph.D. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .8

New Members Named to SFI Boards, External Faculty . . . . . . . . . . . . . . .24

Motorola’s Galvin Takes SFI Post . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .27

INSIDE SFI

Language Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .28

9

21

PH

OTO

: JU

LIE

G

RA

BE

R

16

center section

PH

OTO

: A

MY

SN

YD

ER

, C

OU

RTE

SY

OF

EXP

LOR

ATO

RIU

M,

SA

N F

RA

NC

ISC

O

AR

T:

HO

LLY

RO

BE

RTS

, C

OU

RTE

SY

LE

WA

LLE

N C

ON

TEM

PO

RA

RY,

SA

NTA

FE

DIG

ITA

L C

OLL

AG

E:

PATR

ICK

MC

KE

LVE

Y

Position

scan

strip in

seps

2 S A N T A F E I N S T I T U T E B U L L E T I N F A L L 1 9 9 9

profile

PH

OTO

: JU

LIE

GR

AB

ER

HETERARCHY

The Organization of Diversity—

Travels with Sociologist David Stark

through the Uncertain World of Heterarchy

by Janet Stites

ECONOMIC SOCIOLOGIST DAVID STARK has spent the greater part ofhis academic career traveling to Eastern Europe, particu-larly Hungary, to gather data for his research on how post-socialist enterprises reorganize and recombine resourcesto shape their own distinctive path to capitalism. But it’sin his own backyard where the Columbia University pro-fessor and Santa Fe Institute external faculty member isnow studying the emergence of new organizational formsin the raw lofts and hidden cubbies of downtownManhattan, where since the advent of the graphical userinterface for the World Wide Web, an industry dubbedSilicon Alley has emerged to become one of the mostinfluential centers of the digital economy.

Stark, Arnold A. Saltzman Professor of Sociology andInternational Affairs and chair of the Department ofSociology at Columbia University, has found many sim-ilarities between the organizations of Silicon Alley andthose in Hungary. “I ask the question,” says the soft-spoken Stark, “are there some forms of organization thatare more likely to be able to rede-fine, redeploy, recombine assets?The answer: Yes, organizationswith a capacity for reflexivity. Whatis the basis of that? Active rivalry ofcoexisting principles—the organi-zation of diversity.”

Central to his work is a phe-nomenon he calls “heterarchy.” Hedefines organizations operatingunder the principles of heterarchyas those which operate with mini-mal hierarchy and which have orga-nizational heterogeneity. There isuncertainty and self-organization.To prosper in such a situation,Stark believes managementbecomes the art of facilitating orga-nizations that can perpetually reor-ganize themselves. “The solution is to minimize hierar-chy,” he says. “Authority is no longer delegated verti-cally, but emerges laterally.”

To be sure, in Silicon Alley, firms are working atlightning speed, and traditional management modelswith academic flow charts, regular quarterly reviews andthe like, are often disregarded if ever implemented atall. In some cases, it’s not even clear if successful man-agers from the traditional corporate world can operate inthe start-up world of the digital economy.

“I’m interested in finding out to what extent hierarchyversus heterarchy is there in the new kind of firm which isemerging,” Stark says. The benefit of heterarchy, accord-ing to Stark, is that firms become capable of learning.

In the case of Hungary and Silicon Alley, both

emerging markets find themselves mired in dissonance.Players with varied backgrounds, motives, and even lan-guages, have to learn to work together. There is noestablished corporate ladder per se, often no companyhandbook, metaphoric or otherwise. In the case ofSilicon Alley, the CEO might want to go public and cashout as soon as possible only to start over with a new idea,the director of business development wants to spendmoney to acquire competitors, the marketing directorwants to produce award-winning design, the program-mer wants to write immortal code.

“Dissonance can be positive,” Stark says (which isprobably easier to swallow in theory than in reality).“These organizations can benefit from the active rivalryof competing belief systems. Rivalry fosters cross-fertil-ization.” The example he uses is the infantry officerwho instructs drummers to disrupt the cadence ofmarching soldiers while they are crossing bridges, “lestthe resonance of uniformly marching feet bring calami-

ty,” and cause the bridge to col-lapse. The goal, Stark contends, isto coordinate diverse identitieswithout suppressing differences.

Stark sees both post-socialistHungary and Silicon Alley as sociallaboratories that researchers canuse to test competing theories,because in both, people are active-ly experimenting with new organi-zational forms. Also common toboth is the idea that their experi-mentation is like bricolage—mak-ing do with what is available, rede-ploying assets for new uses, recom-bining resources within and acrossorganizational boundaries.

In Hungary, that might meanworking suddenly without govern-

ment support, with workers who need retraining, withold equipment, or not enough supplies. In Silicon Alley,it means working without the resources of a corporatestructure, in offices where workspace and hierarchy arenot well defined, or having to use speculative stockoptions and new media panache to compete for an oth-erwise pampered workforce.

But there are differences between post-socialistHungary and Silicon Alley. “The Hungarian firms knewwhat they were working toward; there was a model,”Stark says. “They knew what capitalism was and how itworked. In Silicon Alley no one knows what the marketwill be. They are still working in the dark. It’s likebuilding a ship while out at sea.”

Indeed, Silicon Alley hosts an eclectic group of peo-

S A N T A F E I N S T I T U T E B U L L E T I N • F A L L 1 9 9 9 3

ple scrambling to build businesses, some of which willdevelop technologies which will be obsolete before theyare launched (remember “Push?”), others only to bebeaten to market by a competitor or monolith likeMicrosoft. Some will be acquired and then quietlyabsorbed or disbanded altogether; others will go publicin a flurry of publicity, some successfully, others unsuc-cessfully to the point of embarrassment.

As an external faculty member of SFI, Stark hasfound himself most recently discussing the economiesof Hungary and Silicon Alley under the sun of Santa Fe.Stark was first invited to SFI in 1997 by SFI externalfaculty member John Padgett. Having read Stark’s work,Padgett, a political scientist at the University ofChicago, recognized Stark’s cross-disciplinary approachto studying social transformations. Out of a cast of 20 orso social scientists invited to the Institute in the fall of1997, Stark now meets and corresponds with a workinggroup of four, including Padgett, Walter “Woody”Powell from the University of Arizona, and GernotGrabher, an economic geographer from the Universityof Bonn. (In 1997, Stark and Grabher publishedRestructuring Networks in Post-Socialism: Legacies, Linkages,and Localities, Oxford UniversityPress.)

Each brings to the group similardata sets from vastly dissimilareconomies. Padgett’s data is gatheredfrom 15th-century Florence (see SFIBulletin, Summer 1998); Powell’sfrom the bio-tech industry; andGrabher from the advertising industryin London’s Soho.

In each, the scientists studied thenetwork of firms and the people oper-ating the firms. They looked to find where the networkscrossed, where people crossed, and if the firms werelasting. What is common and of great value amongStark’s small group of peers, is that they all have theorysupported by a solid data set, and each studied an econ-omy where there was a fast-paced change. “All theeconomies had a one- to two-year period of vastupheaval, and sometimes external shock,” Stark said.“There was a high degree of uncertainty in the organi-zation environment.”

Stark says he was thinking in SFI terms long beforehe got there. “I was looking at changes in economies,not as planned or designed, but as evolving and self-organizing,” he says. While browsing the library of theWissenschaft Zentrum in Berlin, Stark came across anarticle in Daedalus on advances in computer program-ming written by SFI external faculty member JohnHolland. “It was a paper on using ‘cross fertilization’ for

computer programs that would be capable of adapting tonew problems in the environment. I read it around 1990at the height of the craze of foreign advisors makingrecipes, blueprints, and formulas in Eastern Europe—and it meshed with my criticism of ‘designer capital-ism.’” Since, Stark has been influenced by other SFIscientists, most importantly, economist David Lane forhis work on “complex strategy horizons,” and theoreti-cal chemist Walter Fontana. Lane, Fontana, Padgett,Powell, and Grabher will be among the participants inan ongoing faculty seminar on “Heterarchy” that Starkis organizing at Columbia in 1999-2000, supported by amajor grant from the Andrew W. Mellon Foundation incooperation with SFI.

In regard to his work on-site in Santa Fe, Starkbelieves the benefits of being exposed to creative ideasfrom scientists from different fields will be long-term.“In a certain way, my involvement at SFI mirrors myown view of how organizational innovation comesabout,” Stark says.

While hesitant to shift the focus of his work inEastern Europe, Stark believes that what is going on inNew York City’s Silicon Alley is worth trading in his

passport for a Metrocard. So signifi-cant is the new industry, he and sever-al colleagues, including his wifeanthropologist Monique Girard, are inthe process of setting up a center tochart the emergence of collaborativeorganizational forms in an era of inter-active media.

Already the group is collecting anarchive of the websites of 200 of SiliconAlley’s most prominent firms to exploreconnections among the firms through

hyperlinks. They are collecting data on the activities offirms in the industry from trade journals and business pub-lications, and developing a database of transactions acrossfirms by using a search engine to look for phrases like“merged,” “invested in,” “inked a deal.” They have spenttime closely observing the operations of a number of Alleycompanies and interviewed numerous Alley entrepreneurs.

But to focus on Silicon Alley is to begin to let go ofEastern Europe, where he had been doing researchsince 1977. “Staying in familiar territory would havebeen the easier route,” Stark says. “Beginning a com-pletely new project is like launching a start-up company.It’s difficult, but exhilarating. The learning curve issteep.”

Stark is currently leading a year-long SawyerSeminar entitled “Distributed Intelligence and theOrganization of Diversity” sponsored by theDepartment of Sociology at Columbia, with support

4 S A N T A F E I N S T I T U T E B U L L E T I N • F A L L 1 9 9 9

GIO

RG

IO B

AS

AR

IM (

15

11

-73

) “V

IEW

OF

FLO

RE

NC

E”

FRE

SC

O

A B O U T S I L I C O N A L L E Y

New York's Silicon Alley started with a whimper, not a bang, onthe eve of the Internet revolution, when a few individuals—longon foresight—saw infinite possibilities available in consumer andbusiness applications via the World Wide Web. This, thanks tothe introduction of a graphical user interface that would becomeknown as a browser.

Initially, the core of the industry was numerous small Webdesign shops which, because of the agility small businessesafforded them, were able to scoop corporate design shops and tra-ditional advertising agencies to lock in Fortune 1000 design jobs.

It became evident very quickly that there was much to do onthe Web beyond design. New York, being the media capital of theworld, suddenly found itself rife with companies hoping to devel-op "content" for the Web. They included iVillage.com, now a Website for women, initially for parenting; theglobe.com, started bytwo young Cornell University graduates, who worked tirelessly tocorner the market on Generation X; and FEED Magazine, the

Internet's answer to Harper's or The Atlantic Monthly.Beyond design and content, companies like Internet

Advertising Network (now DoubleClick) helped to definethe way advertising on the Internet would work. Othersspecialized in extending the effectiveness of the bannerad or measuring "clicks." With interest in the Internethigh, more traditional software companies, previouslyworking in near anonymity, began to surface. Suddenly, ifyou weren't working in new media, you weren't anybody atall. Otherwise well-paid people traded in their $50-per-square-foot office space, regular salaries and great bene-fits, to work in dingy makeshift lofts, or in second-ratebuildings in the financial district which had stood emptysince the late ‘80s bust of the market. From the FlatironDistrict to Wall Street, a new type of company emerged onthe landscape—the “start-up.” A journalist dubbed thearea “Silicon Alley” and the moniker, as well as the indus-try, stuck.

Now the term Silicon Alley is more metaphor thanactual place, representing the growth of technology or

Internet-based companies throughout the New York City metro-politan area-and beyond. Companies like DoubleClick, iVillage,and theglobe.com have made international headlines with theirsuccessful public offerings. The venture capital community andinvestment banks, once skeptical, are swarming the market. Bigcity that it is, one can hear people discussing their businessplans in parks, on the subways, in coffee shops. Once almost aforeign concept to the New York area (particularly the city), smallbusiness and entrepreneurism have surfaced from the shadow ofthe alleys and skyscrapers to become the darlings of economicdevelopment and Wall Street—the core of the big apple.

URLs:www.iVillage.comwww.theglobe.comwww.feedmag.comwww.doubleclick.net Janet Stites

from the Andrew W. Mellon Foundation. The aim ofthis faculty seminar is to explore the emergence anddynamics of new organizational forms in response tothe extraordinary uncertainties of rapid technological,economic, and social change.

On the subject of learning and adapting, Starklikes to tell the story about a souvenir he keeps on hisdesk: “I have a tin can that I bought in Budapest inthe autumn of 1989. It’s considerably smaller thanyour standard tuna can and extremely light in weight.If you tap your fingernail on it, it gives a hollow ring.But the label, complete with a universal bar code,announces in bold letters that, in fact, it’s not empty:Kommunizmus Utolso Lehelete—The Last Breath ofCommunism.”

The trinket came not from a clever entrepreneurin post-socialist Eastern Europe, but from a state-owned work team which took advantage of legislationthat allowed employees of socialist firms to form“intra-enterprise partnerships.” Such practices werethe beginning of organiza-tional hedging, accordingto Stark—one toe in thewater.

“The challenge of themodern firm, whether itbe a post-socialist firmcoping with the uncer-tainties of system changeor a digital technologiesfirm coping with unpre-dictable strategy hori-zons,” Stark writes in hispaper, “Heterarchy: AssetAmbiguity and theOrganization of Diversityin Post-socialist Firms”(July 1996), “is the chal-lenge of building organizations that are capable oflearning. Flexibility requires an ability to redefine andrecombine assets: in short, a pragmatic reflexivity.”

Stark believes these attributes are imperative in avolatile, fast-moving market. “It’s the difference,” hesays, “between a firm that is adaptable and one that isjust ‘adapted.’”

Janet Stites is a free lance writer and publisher of AlleyCatNews (www.alleycatnews.com), a New York-based monthlymagazine which covers the business of Silicon Alley. She haswritten for OMNI Magazine, Newsweek, and TheNew York Times.

S A N T A F E I N S T I T U T E B U L L E T I N • F A L L 1 9 9 9 5

PH

OTO

S:

HAW

LEY

JO

HN

SO

N

news6 S A N T A F E I N S T I T U T E B U L L E T I N • F A L L 1 9 9 9

SFI ARTIFICIALSTOCK MARKETSOON ONLINERecent work on the Santa Fe InstituteArtificial Stock Market (ASM) model by SFIgraduate student Shareen Joshi, in collabo-ration with Reed College advisors JeffreyParker and Mark Bedau, suggests that thismodel may be used to understand someinteresting features of financial markets,such as the destabilizing effects of techni-cal trading rules, and the profitability ofsuch rules. In short, the research suggeststhat the frequent revision of forecasts,accompanied by high technical trading,leads to higher volatility and less overallwealth.

With public distributionof the ASM code comingsoon to the SFI web site,others will be able to probethese results for them-selves. The Artificial StockMarket, one of the firstagent-based models of afinancial market, was devel-oped in the early 1990s byW. Brian Arthur, JohnHolland, Blake LeBaron,Richard Palmer, Paul Tayler,and Brandon Weber at theSanta Fe Institute. It con-tains a population ofmyopic, imperfectly ratio-nal, heterogenous agentswho make investment deci-sions by forecasting thefuture states of the market,and who also learn fromtheir experience over time.The model illustrates howsimple interactions amongsuch agents may lead tothe appearance of realistic market behavior.

Results suggest that many features ofreal-world markets (such as bubbles andcrashes in prices) that have in the pastbeen attributed to external influences mayactually be internally generated within themarket structure itself.

THE MODELThe simulated market contains two

assets: a risky stock, in finite supply, and arisk-free bond, available in infinite supply.

Agents, initially endowed with a certain sumof money, must decide in each time periodof the simulation how to allocate their capi-tal between the two assets. They do this byforecasting the price of the stock, andassessing its riskiness (measured by thevariance of the prices). Forecasting rulesare IF-THEN statements: IF (a certain mar-ket state occurs) THEN (a certain forecastis made).

Agents may recognize two differentkinds of market states (possibly simultane-ously): technical and fundamental. A marketstate detected by an agent is “technical” ifit identifies a pattern in the past price his-tory, and is fundamental if it identifies animmediate over- or under-valuation of thestock. An example of a technical statewould be “the price is greater than the 50period moving average,” and an example of

a fundamental state would be “the price isover-valued by 10 percent.”

If the market state in a given periodmatches the descriptor of a forecasting rule,the rule is said to be activated. A number ofan agent’s forecasting rules may be activat-ed at a given time, thus giving the agentmany possible forecasts from which tochoose. An agent decides which of theactive forecasts to use by choosing at ran-dom among the active forecasts with a prob-ability proportional to its accuracy, a mea-

sure that indicates how well the rule hasperformed in the past. Once the agent haschosen a specific rule to use, the rule isemployed to make an investment decision.

Agents determine how much stock tobuy, sell or hold, using a standard risk-aver-sion calculation. They submit their decisionsto the market specialist, an extra agent inthe market who functions as a market-maker. The specialist may use a number ofdifferent techniques to declare a market-clearing price (available as parameters).

A genetic algorithm (GA) provides forthe evolution of the population of forecast-ing rules over time. Whenever the GA isinvoked, it substitutes new forecastingrules for a fraction of the least-fit forecast-ing rules in each agent’s pool of rules. Arule’s success or “fitness” is determinedby its accuracy and by how complex it is

(the GA has a bias against complex rules).New rules are created by first applying thegenetic operators of mutation andcrossover to the bit strings of the more suc-cessful rules in the agent’s rule pool. TheGA may be compared to a real-world con-sultant. It replaces current poorly perform-ing rules with rules that are likely to per-form better, much the same way as a con-sultant urges her client to replace poorlyperforming trading strategies with thosethat are likely to be more profitable.

WE

B S

ITE

IN

PR

OG

RE

SS

—B

RA

ND

ON

WE

BE

R

newsS A N T A F E I N S T I T U T E B U L L E T I N • F A L L 1 9 9 9 7

It is important to note that agents inthis model learn in two ways: First, as eachrule’s accuracy varies from time period totime period, each agent preferentially usesthe more accurate of the rules available toher; and, second, on an evolutionary timescale, the pool of rules as a whole improvesthrough the action of the genetic algorithm.

RESULTSThe most significant early finding was

that this market exhibits two quite differentkinds of behavior, corresponding to differ-ent rates at which market-forecasting rulesare being revised by the genetic algorithm.When the GA-invocation interval is large(between 1,000 and 10,000) resulting inforecasting rules evolving relatively slowly,prices are more stable; evolved forecastingrules are simple; levels of technical tradingare low; trading volumes are low; and thereis little evidence of nonlinearity. Since thiskind of behavior resembles the predictionsof the theory of efficient markets, thisregime has been termed the “RationalExpectations Regime.”

On the other hand, when the GA-invoca-tion interval is small (between 10 and 100)it results in forecasting rules evolving rela-tively quickly, and the variance of the pricetime series is relatively high; the evolvedrules are complex; levels of technical trad-ing are high; trading volumes are higher;and there is strong evidence of nonlinearity.This regime is called the “ComplexRegime.”

Since 1998 Joshi, Parker, and Bedauhave been further studying the dynamics ofthe ASM. Using a simple game theoreticalmodel together with the Santa Fe Stock

Market, theyattempted to find theoptimal rate at whichtraders shouldrevise their reper-toire of market-fore-casting rules usingthe GA. They showedthat the market hasonly one symmetricNash equilibrium,and that this equilib-rium lay in the“Complex Regime.”Most important ofall, they concludedthat this symmetricNash equilibrium

was “sub-optimal” because in this regimethe wealth accumulated by agents waslower than in the Rational ExpectationsRegime, the asset was riskier (prices wereless stable), and the market was noisier.

These recent results suggest that finan-cial markets can end up in situations anal-ogous to a multi-person Prisoner’s Dilemmagame in which frequent revision of fore-casting rules can lead to increased pricevariability and thus reduced overall earn-ings. When traders do not know a prioriwhat other traders are doing, their optimalstrategy is to revise forecasting rules fre-quently. But when this dominant strategyprevails and market beliefs co-evolve rapid-ly, the market falls into a symmetric Nashequilibrium with relatively low average earn-ings for traders. In other words, this workindicates that if every trader used technicalanalysis the result would be a general lossof profit.

Much research remains to be done inestablishing the robustness of these resultsto variations both in the model’s parametersand in the structural design of the modelitself. As part of this effort SFI plans toestablish a Web site for public distribution ofthe ASM code. The Web site will be designedto encourage general testing of the code, val-idation against empirical data, and collectionof detailed simulation studies.

SFI TRUSTEE JOHN POWERS 1916-1999

SFI Trustee John Powers, a long-time sup-porter of the Institute’s activities and aninfluential figure in the modern art commu-nity, died in September at the age of 83.

Powers and his wife Kimiko moved fromNew York to live full-time in Aspen in themid-1970s when Powers retired as presi-dent of Prentice Hall Publishing Company.The couple lived first in one of the AspenInstitute trustee houses, and then latermoved to Carbondale, Colorado.

Powers donated to the Institute manyworks of contemporary art—including printsby Cristo, Roy Lichtenstein, and JasperJohns-—which today grace the campus.SFI Founding President George Cowannotes, “John’s support of SFI activitiesnever wavered through the years. His ownintellectual curiosity, coupled with his cre-ative and artistic sensibilities, found ahome here at SFI and at the Aspen Institutewhere he was also active for many years.We will all miss him.”

past experience

goals

feedback actions

population of models[fittest in bold]

machineryto decideaction

Environment

Agent

A Typical Agent-Based Model

PH

OTO

: M

AR

Y E

SH

BA

UG

H H

AYE

S

news8 S A N T A F E I N S T I T U T E B U L L E T I N • F A L L 1 9 9 9

NEW BOOKS FROM OXFORDSFI is pleased to announce two new booksfrom Oxford University Press, publisher ofthe Santa Fe Institute Studies in theSciences of Complexity (SISOC) bookseries. Orders can be placed by calling 1-800-451-7556.

Swarm Intelligence: From Natural to Artificial Systems

By Eric Bonabeau, Marco Dorigo, and Guy Theraulaz

Social insects—ants, bees, termites,and wasps—provide us with a powerfulmetaphor to create decentralized problem-solving systems composed of simple inter-acting, and often mobile, agents. The emer-gent collective intelligence of social insects,swarm intelligence, lies not in complex indi-vidual capabilities but rather in networks ofinteractions that exist among individuals andbetween individuals and their environment.

Swarm intelligence offers another wayof designing “intelligent” systems, whereautonomy, emergence, and distributed func-tioning replace control, preprogramming,and centralization. This book surveys sev-eral examples of swarm intelligence insocial insects and describes how to designdistributed algorithms, multi-agent sys-tems, and groups of robots according to thesocial insect metaphor.

Dynamics of Human and PrimateSocieties: Agent-Based Modelingof Social and Spatial Processes

Edited by Timothy A. Kohler and George J. Gumerman

This volume presents a series of stud-ies from archaeologists, enthnographers,primatologists, computer scientists, sociol-ogists, and philosophers who use agent-based models to examine social and spatialdynamics. The contributors, convened at aninternational conference at the Santa FeInstitute in December 1997, consider avariety of societies, from ones as apparent-ly simple as those of primates, to ones ascomplex as those of contemporary industri-al nations. These papers ask and find pro-visional answers to fundamental questionssuch as: How do levels of selection andspatial configuration of resources interactin the evolution of cooperation? How can weexplain the evolution of inference? Andwhat is the role of warfare in the emer-gence of state-level societies? Other papersare concerned with understanding how set-tlement patterns are generated amongMesolithic foragers in the SouthernHebrides or small-scale Neolithic societiesin the North American Southwest.

OPREA EARNS PH.D.In April 1999, Mihaela Oprea was awardeda doctorate in computer science from theUniversity of New Mexico. The main thrustof Oprea’s work focuses on the evolutionaryaspects of the immune system: how anti-body gene libraries encode informationabout the pathogen environment of thespecies; how the immune system improvesits efficiency of generating high affinity anti-bodies during on-going immune responses;how mutation rates are estimated; andwhat mechanism is responsible for somatichypermutation of antibody genes.

Before coming to the United States,Oprea earned an M.D. from the Universityof Medicine and Pharmacy at Timisoara,Romania.

Her thesis is entitled “AntibodyRepertoires and Pathogen Recognition:The Role of Germline Diversity and SomaticMutation.” Much of her research has beendone at the Santa Fe Institute in collabora-tion with external faculty members TomKepler and Alan Perelson under the aus-pices of the SFI Theoretical Immunologyprogram supported in part by the Joseph P.and Jeanne M. Sullivan Foundation.Currently Oprea is continuing her researchwith Perelson as a postdoctoral fellow atLos Alamos National Laboratory.

PH

OTO

OF

MIH

AE

LA O

PR

EA

: JU

LIE

GR

AB

ER

S A N T A F E I N S T I T U T E B U L L E T I N • F A L L 1 9 9 9 9

PH

OTO

: PA

TRIC

K M

CFA

RLI

NP

HO

TO:

DA

N B

AR

SO

TTI

DESIGNING THE

DISCOVERING PATTERNS—THE INTERFACE BETWEEN ART AND SCIENCE

BY HOLLIS WALKER

10 S A N T A F E I N S T I T U T E B U L L E T I N • F A L L 1 9 9 9

These days, Jim Crutchfield is thinking aboutvocabularies—about how humans continuously extendour vocabularies to describe new realms of experience.His concerns run counter to those espoused by philoso-pher Ludwig Wittgenstein early in this century, whosaid that if you don’t know about something, you can'ttalk about it. Crutchfield is more worried about how ourperception of the world is limited by our vocabulariesand how we transcend those limitations. “How do youextend your vocabulary in a dynamic way? How do youteach yourself to see new patterns youhaven’t seen before?” he asks.

This fascination with what he calls“pattern discovery” is one of the rea-sons Crutchfield spends more timehanging out with artists than other sci-entists. Like mathematics, art can bean iconic substitute for language—and at the same time a vocabularyunto itself. “In a way, art is a theoryabout the way the world looks tohuman beings,” MitchellFeigenbaum told James Gleick,author of Chaos: Making A New Science.“What artists have accomplished isrealizing that there’s only a smallamount of stuff that’s important, andthen seeing what it was. So they cando some of my research for me.”

Crutchfield says his long-standingfriendships with manyartists, including NedKahn, Sara Roberts,and Gail Wight, haveinspired some of hismore rigorous scientificinquiries. Kahn, withwhom Crutchfieldworked on the 1996-97San FranciscoExploratorium exhibit,Turbulent Landscapes:The Natural ForcesThat Shape Our World(http://www.explorato-rium.edu/complexity),creates art installationsinspired by atmospher-ic physics, geology,astronomy, and fluidmotion: flapping flags,dust devils, swirlingstreams.

During the years Crutchfield was at UC Berkeley,Kahn would call him for help on the scientific end of hisconstructions: “Hey, Jim, I've been squirting water intoa satellite dish, and the vortex detaches from the drain.What do you think is happening?” Often Kahn’s querieswould send Crutchfield into his Berkeley physics lab,his own curiosity piqued. The artist, meanwhile, waslooking for a way to put a frame around an active, natur-al system, then create a way for observers to alter it—ineffect, manufacturing curiosity to engage ordinary peo-

ple with nature, science, and art. While Kahn formalizes opportuni-

ties to perceive and effect naturalphenomena, Sara Roberts borrowsthe mathematics used to describethose phenomena and employs it inthe design of her anthropomorphiccomputer programs. Roberts teachesat the California Institute of the Artsin Valencia, where she also foundedand now directs the Integrated MediaProgram.

“Dynamical systems theory haslots of useful material for me,”Roberts explains. “It’s not that I’minterested in looking at the worldthrough a dynamical systems filter.It’s useful to me as technique.” Muchof Roberts’ work, including her 1994Elective Affinities (based on the

Goethe novella), useddynamical systemsmodels to drive the“emotional engines” ofmultimedia installa-tions. In ElectiveAffinities, the installa-tion became ametaphor for thedynamics of complexhuman relationships.

A married coupleand two close friendsare riding in a cartogether and, as a resultof sexual innuendoesbandied about at a pic-nic from which they’rereturning, they arethinking about betray-ing each other. Thecharacters are projectedvideo images in front

top: “Circling Wave” by Ned Kahn, Exploratorium, San Francisco.bottom: “Encircled Stream” by Ned Kahn, Seattle, Washington

PH

OTO

S C

OU

RTE

SY

OF

THE

EXP

LOR

ATO

RIU

M,

SA

N F

RA

NC

ISC

O

S A N T A F E I N S T I T U T E B U L L E T I N • F A L L 1 9 9 9 11

and back seats repre-sented by video bustson a pedestal; thescenery outside the carrushes away from themon a screen mountedon the wall behind.Each character runs onits own computer andowns its own emotionalprogram and databaseof thoughts; all four aren e t w o r k e d .Occasionally, one char-acter glances at anoth-er. That glance altersthe state of the emo-tional engine in theother character according to a set ofrules. “The spectator looking at themdoesn't see the system in action, butwhen you walk close to each charac-ter’s pedestal, you can hear theirthoughts,” Roberts says.

Another artist who has workedwith Crutchfield is Gail Wight of SanFrancisco, who is developing a newelectronic arts program at MillsCollege. Wight met Crutchfield whilecreating a piece for TurbulentLandscapes on biological self-organi-zation in dictostelium slime mold.Biologists are fascinated with thisspecies of slime mold because when adictostelium cell is in danger of dyingof starvation or thirst, it sends outchemical signals to surroundingcells—and they aggregate by movingin synchronized waves into a slug-likecreature that forms a budding stalk,which explodes, sending spores flyingto distant, and perhaps more hos-pitable, environments.

Wight was supposed to grow dic-tostelium as part of herExploratorium installation, but thespores she was given were a differentspecies (physarum) that only grows in tree-like struc-tures and does not shift from individual cells to a multi-cellular organism—something she did not discover untilafter months of waiting for her slime cells to organizeinto traveling waves. The experience caused her to lookat science differently, to question her implicit trust, and

led to the way she nowuses science in herwork—by questioningit. “I started being verysuspicious of my owninfatuation with sci-ence,” she said. “Ibegan to wonder, howdid we get to this place,this thing we now callscience?”

Her question isapparent in a laterinstallation in whichshe put 50 mice in indi-vidual cages; their envi-ronments illustratedmoments in the history

of genetics. One was a portrayal ofMendel’s pea garden. Inside the cagewas a miniature pea garden, which,eventually, the mouse ate. Anothercage illustrated the studies fromwhich scientists concluded twins hadlittle in common genetically—only tolater realize that their own biases pre-vented them from recognizing thetwins' shared attributes. One mousein the twin pair had a tiny baby grandpiano in his cage; he decided to sleepinside it. The other had a shabbyupright piano; he ate the instrument.

Recently, Wight designed a similarinstallation; this one includes five tinytableaux from history representing“how we came to conceive of our-selves as electrochemical entities.” Ofthis exhibit, Wight said, “These tinytableaux are sitting inside a squarePlexiglas maze, and there’s a rat thatlives inside of it. The rat will hopeful-ly eat away the tableaux. The rat issort of the artist.”

In the winter of 1998-99,Crutchfield organized a public lectureseries in Santa Fe titled “Arts of theArtificial.” Motivated by an interest in

how art and science will determine the structure of “vir-tual spaces” created by networked computers, the seriesincluded talks by Gail Wight, Roberts, art critic DaveHickey, and Rodney Brooks, director of MIT's ArtificialIntelligence Lab.

The idea of the series was that public exposure to the

PH

OTO

S C

OU

RTE

SY

OF

GA

IL W

IGH

T/IN

STA

LLAT

ION

PH

OTO

: B

EN

BLA

CK

WE

LL.

12 S A N T A F E I N S T I T U T E B U L L E T I N • F A L L 1 9 9 9

thinking of those onthe forefront of pat-tern discovery in artand science mightinspire others towardsimilar inquiries.Putting artists and sci-entists together oftencauses each to recog-nize old things in newways—that mysteriousprocess of pattern dis-covery at whichhumans are so good.

Kahn was recentlyworking on a geologi-cal installation thatrepresents a slicethrough a volcaniclandscape. Air ispumped up throughtwo sheets of glass,fluidizing a powderymixture, erupting tothe surface and creat-ing a caldera. “When Igot this working Icalled this geophysi-cist, Raymond Jean-Luz, at (UC)Berkeley,” Kahnrecalled. “He was sointo this thing hespent three hours juststaring at it. He cameback the next day witha graduate student andthey spent all day star-ing at it. They werelooking at somethingreal. It reminded themof why they got inter-ested in geology in thefirst place.”

Kahn says lookingat real phenomenaprompts a differentkind of thinking. “Your mind is working on a lot of lev-els. You’re processing this visual information, and you’rerecognizing patterns, some so subtle you probably can'tdescribe what you're seeing, but on some level aesthet-ically . . . there's an indication that there is an order inthere.”

That’s similar to what Crutchfield experienced inthe laboratory during his experiments on video feed-back. These were not focused so much on the rich pat-terns generated by that system, but on the process of hisown perception of those patterns. Eventually, this ledhim to develop a mathematical framework to describe

“Techne and Eros: Human Space and the Machine” drew participants from around the world.Above; student participant. Below David Dunn.

PH

OTO

S:

DA

N B

AR

SO

TTI

S A N T A F E I N S T I T U T E B U L L E T I N • F A L L 1 9 9 9 13

the process of pattern discovery.He would be looking at the videomonitor, thinking, “This patternlooks familiar...similar to some-thing else I saw a few days ago,”he recalls. “The empirical factsthat I concentrated on were notimages that appeared on thescreen but the first intuitiveimpressions of regularity thatoccurred as I began to see newpatterns.”

One of the projectsCrutchfield began recentlyinvolves magnetoencephalogra-phy (MEG), a new imaging tech-nique that measures neural activ-ity via magnetic signals generatedby the functioning brain—apotentially more sensitivemethod than the more familiarelectroencephalography (EEG).A clinician typically analyzes suchdata by visual inspection, that is,studying temporal informationrecorded on strips of paper or on ascreen, and recognizing certainpatterns within the data.

The problem is that the moresophisticated MEG machines,such as the 122-channel one atthe Veterans AdministrationHospital in Albuquerque, pro-duce gigabytes of data in just afew minutes of recording from asubject, more data than a clini-cian could ever analyze by eye.What's needed is the ability toanalyze such quantities of datafor hundreds of people, overtime, to identify norms andanomalies associated with illness.“Can we teach a machine to auto-matically discover patterns insuch huge quantities of data?”Crutchfield asks.

In the past, human beingshave been constrained by thelimits of our physical world andour evolutionary heritage.Although it may not be possiblefor the human mind to perceivepatterns in more than four or five

SP

IRA

L D

ES

IGN

GE

NE

RAT

ED

BY

JIM

CR

UTC

HFI

ELD

. P

HO

TOS

: PA

TRIC

K M

CFA

RLI

N,

GA

IL W

IGH

T

M U R R A Y G E L L - M A N N

A N D T H E C R E A T I V E

P R O C E S S

SFI Distinguished Professor Murray Gell-Mann uses neckties asan easy way to talk about simplicity and complexity. “If you’relooking at a pattern of a necktie and it’s just regimental stripes,it’s simple,” he said. “But you’ve seen neckties with much morecomplex patterns.” His point is that those complex patternshave regularities that it would take a long time to describe.

Gell-Mann’s interest in questions of simplicity and complex-ity and their intersections with art led last fall to a forum co-sponsored with SITE Santa Fe called “Simplicity and Complexityin the Arts and the Creative Process.” The forum brought togeth-er a number of scientists and artists in discussions at the SantaFe Institute that culminated in a public presentation at SITESanta Fe. Gell-Mann and his wife poet Marcia Southwick (whoselatest book is A Saturday Night at The Flying Dog and OtherPoems, Oberlin College Press, 1999) together organized theforum. Among the scientists attending were Chuck Stevens andJim Crutchfield of SFI. Arts panelists included novelist CormacMcCarthy, architect Moshe Safdie, poet David St. John, andvisual artist Joseph Kosuth.

Gell-Mann and Southwick also have attended recent meet-ings (along with others on the faculty at SFI) on connectionsbetween complexity and the arts in Abisko, Sweden and CatalinaIsland, California. Some of the topics explored at the meetingshave included: regularities in the visual and musical arts thathave counterparts in human brain function; the universal appealof poetry; and measures of effective complexity in art.

Currently, Gell-Mann, with his assistant Marla Karmesin, istrying to compile a Digital Video Disk of material from the SantaFe forum, including videotaped lectures, photographs of the art-objects shown, recordings and supplementary materials. TheDVD will provide a jumping-off point for future discussions.

Hollis WalkerHollis Walker is arts and entertainment editor at The SantaFe New Mexican. She was a Pew Charitable Trusts’National Arts Journalism Fellow in 1996-97.

dimensions, once trained, machines may be able to do itfor us, Crutchfield says. Such pattern-discoverymachines could become our proxies in worlds we cannotvisualize. A more intelligent MEG machine would notonly produce massive quantities of data, but also be ableto recognize patterns in that data—and then point themout to us. Can Crutchfield and his colleagues teachmachines to see patterns and regularities in high-dimen-sional spaces, for example, to analyze those mountains ofMEG data? That’s their goal.

Some of these ideas will no doubt be on the agendaof a new research facility just formed in Santa Fe. What'stentatively being called The Art and Science Laboratorywill involve Crutchfield and pioneer electronic com-posers and artists including David Dunn and Steina andWoody Vasulka (founders in the 1960s of The Kitchen,an electronic art performance space in New York City).This core group—plus composer/electronic artist MortonSubotnik and composer/vocalist Joan La Barbara—pre-sented a six-week series of workshops, “Techne andEros: Human Space and the Machine,” at the Santa FeArt Institute this summer that drew students fromaround the world. A permanent exploratory science-artsfacility in Santa Fe will offer them and others working inthese loosely defined arenas a way to easily interact withthe researchers at SFI and other institutions.

New machines that can think better than we can,communicate in languages we do not speak, in realms ofwhich we cannot perceive—it sounds like science fiction.In fact, these are the characteristics of cyberspace, onlythe first of the novel non-physical/non-biological realmshumans are creating. The inventors of these new, verysocial spaces should include artists as well as scientistsand technologists, Crutchfield says. Why? Consider theinnovation in magnetic materials that led to the small,powerful motors which drive Sony Walkmans. Now peo-ple the world over are running, riding the subway, racingthrough their ordinary lives while wearing the ubiquitousheadphones attached to miniature music boxes. Scienceaffects technology which drives culture, and culture indi-rectly determines the directions in which society choosesto invest scientifically.

“Shift that feedback loop into the new virtualspaces,” Crutchfield suggests. Imagine that artists, aswell as scientists, have primary input into the structure ofsuch new realms. He adds, “It will be an entirely differ-ent world, one in which physical and biological con-straints are markedly less dominant, and aesthetic choiceand design are primary.”

14 S A N T A F E I N S T I T U T E B U L L E T I N • F A L L 1 9 9 9

PH

OTO

:UU

RR

AE

HAY

NE

S

SFI Bulletin

Small-World Networks

DIE-CUT

art, do not print

this white circle

THE KEVIN BACON GAME is a curious thing to be sure. Forthose who don’t know him, Kevin Bacon is an actor bestknown for not being the star of many films. But a fewyears ago, Brett Tjaden—a computer scientist at theUniversity of Virginia—catapulted Bacon to true inter-national recognition with the claim that he wassomehow at the center of the movie uni-verse. This is how the game goes:

• Think of an actor or actress.• If they have ever been in a

film with Kevin Bacon, then theyhave a “Bacon Number” of one.

• If they have never been in afilm with Kevin Bacon but havebeen in a film with somebody elsewho has, then they have a BaconNumber of two, and so on.

The claim is that no one who hasbeen in an American film, ever has a BaconNumber of greater than four. Elvis Presley, forexample, has a Bacon Number of two. For real enthusi-asts, Tjaden created a web site that provides the BaconNumber and shortest path to the great man for the mostobscure of choices. In fact, Tjaden later fireproofed hisclaim by conducting an exhaustive survey of theInternet Movie Database, and determined that thehighest finite Bacon Number (for any nationality) iseight. This may seem nothing more than a quirky factabout an already bizarre industry, but in fact it is a par-ticularly clear example of a phenomenon that increas-ingly pervades our day-to-day existence: somethingknown as the “small-world phenomenon.”

The small-world phenomenon formalizes the anec-dotal notion that “you are only ever six ‘degrees of sep-aration’ away from anybody else on the planet.” Almosteveryone is familiar with the sensation of running into acomplete stranger at a party or in some public arena and,after a short conversation, discovering that they know

somebody unexpected in common. “Well, it’s a small-world!” they exclaim. The small-world phenomenon isa generalized version of this experience, the claim beingthat even when two people do not have a friend in com-mon, only a short chain of intermediaries separates

them. Stanley Milgram made the first experimentalassault on the problem (confined to the United

States) by sending a series of traceable let-ters from originating points in Kansas and

Nebraska to one of two destinations inBoston. The letters could be sent onlyto someone whom the current holderknew by first name and who was pre-sumably more likely than the holder toknow the person to whom the letter was

ultimately addressed. By requiring eachintermediary to report their receipt of the

letter, Milgram kept track of the letters andthe demographic characteristics of their han-

dlers. His results indicated a median chain length ofabout six, thus supporting the notion of “six degrees ofseparation,” after which both a play and its movie adap-tation have since been named.

This result was both striking and surprising and con-tinues to be so today, because the conscious construc-tion of such chains of intermediaries is very difficult todo. Ordinarily, our perception of the social world is con-fined to our group of immediate acquaintances, andwithin this group there is a great deal of redundancy;that is, within any one circle of acquaintances, most ofthem know each other. Furthermore, our average num-ber of acquaintances is very much less than the size ofthe global population (at most thousands, comparedwith billions). So the claim that some very short chain ofacquaintances exists that links us to any other person,anywhere in the world, does seem unlikely.

From Small-Worlds by Duncan Watts, PrincetonUniversity Press, 1999

Kevin Bacon, the Small-World,and Why It All Matters

S A N T A F E I N S T I T U T E B U L L E T I N • F A L L 1 9 9 9

Under what conditions can any network, social or otherwise, be deemed“small”? And if small-world networks can be shown to exist, what does thisimply about the behavior of systems that are connected up that way? Thesequestions, among others, have been the research focus of SFI PostdoctoralFellow Duncan Watts and his collaborators.

REGULAR SMALL WORLD RANDOM

i n c r e a s i n g r a n d o m n e s sp=0

This illustrates the random rewiring procedure for interpolating between a regular ring lattice and a random network, without altering thevertices in the graph. Three realizations of this process are shown, for different values of p. For p=0 the original ring is unchanged; as pincreases the graph becomes increasingly disordered until for p=1, all edges are rewired randomly. For intermediate values of p, the graphis a small-world network—highly clustered like a regular graph, yet with small characteristic path length, like a random graph.

p=1

Do we live in a small world?

S A N T A F E I N S T I T U T E B U L L E T I N • F A L L 1 9 9 9

CLOSE CONNECTIONSNetworks are ubiquitous. The brain is a network of

neurons. Organizations are networks of people. Theglobal economy is a network of national economies,which are networks of markets, which in turn are net-works of producers and consumers. Diseases and rumorsboth transmit themselves through social networks, andcomputer viruses propagate via the Internet.

Any kind of network can be represented abstractlyby a graph, composed of nodes (or vertices) and a set oflines, edges, joining the nodes. The nodes represent,say, members of a population, and the edges, their inter-personal ties, business ties, friendships, etc. . . .Traditionally, however, networks have been modeled aseither completely ordered or completely random. In anordered network, like a crystal lattice, each node has thesame number of edges that join a small number ofneighboring nodes in a tightly clustered pattern. In arandom network, each node is arbitrarily connected tonodes that can lie anywhere. Although ordered and ran-dom networks are in one sense extreme opposites, theyshare the common feature of uniformity; that is, locallyeach network “looks” the same everywhere, and thissimplifies their analysis.

However, most real-world networks appear to fallsomewhere in between the ordered and randomextremes. Friendship networks are a good example ofthis in-between state. Since people meet most newfriends through existing friends, the networks are locallyordered. (Here order means that if A knows B and Bknows C, then A is more likely to know C than someother random element.) The outcome of local orderingin such a network is that one individual’s friends aremore likely than not to know one another: a characteris-tic that is called “clustering.” Many real-world networks,including friendship networks, tend to be highly clus-tered, but they are not entirely so. If a person joins a cluband meets new people or moves to a different city totake a job, new connections can form that are notordered by the existing network.

In order to simulate this kind of intermediate systemone might take an ordered network and deliberatelyintroduce increasing amounts of randomness into it.Watts and Steven Strogatz, Watts’ thesis advisor atCornell University, took this approach, called “randomrewiring” in order to explore more deeply what wouldhappen to the properties of initially-ordered networks.For example, beginning with the graph of a 1D lattice(simply a ring of nodes, each connected only to its near-est neighbors within a specified radius) they beganreplacing near-neighbor edges with edges to randomlyselected nodes chosen uniformly throughout the net-work.

To understand the resulting networks, Watts andStrogatz computed two statistics. The first—the charac-teristic path length—is defined as the length of the short-est path (i.e., smallest number of edges) required to con-nect one node to another, averaged over all pairs of nodes.The second parameter—the clustering coefficient—mea-sures the average probability that two nodes with a mutu-al “friend” will be connected, or in other words, the aver-age cliquishness of local neighborhoods. According tothese measures, lattice-like networks have long charac-teristic path lengths and large clustering coefficients,whereas random networks are “small” and exhibit verylittle clustering at all.

Studying intermediate kinds of networks led to thediscovery that when just a few long-range, random con-nections replace the local edges of a lattice-like network,the characteristic path length decreases dramatically; a“shortcut” occurs. And while the first random rewiring hasa great impact on path length, the clustering changes verylittle. Even when the separation of elements in a networkis very small the clustering can remain almost as high aspossible. This result is what Watts and Strogatz call a“small-world network.” The name derives from the factthat it exhibits the short global separations that are typi-fied (anecdotally) by social interactions while maintainingthe high degree of clustering exhibited in most social net-works.

REAL-WORLD NETWORKS, SMALL-WORLD NETWORKS

Idealized models like the one just described suggestthat the small-world phenomenon might be common insparse networks with many vertices, as even a tiny frac-tion of long-range shortcuts would suffice to make theworld “small.” But does it arise in the real world? Wattsand Strogatz set out to check this, selecting three differ-ent real-world networks, for which all the data necessaryto compute characteristic path length and clusteringcoefficient was available. The first system was a data-base of feature-film actors ordered by their appearancein different films; the second was the electric power gridof the Western United States; and the third was theneural network of the nematode worm C. elegans. AsWatts and Strogatz wrote in a letter in Nature, “All threegraphs are of scientific interest. The graph of film actorsis a surrogate for a social network, with the advantage ofbeing much more easily specified. It is also akin to thegraph of mathematical collaborations centered, tradi-tionally, on P. Erdos. The graph of the power grid is rel-evant to the efficiency and robustness of power net-works, and C. elegans is the sole example of a complete-ly mapped neural network.”

Each of the three graphs turned out to be a small-

S A N T A F E I N S T I T U T E B U L L E T I N • F A L L 1 9 9 9

S A N T A F E I N S T I T U T E B U L L E T I N • F A L L 1 9 9 9

Empirical examples of small-world networks

L actual L random Cactual Crandom

Film actors 3.65 2.99 0.79 0.00027

Power grid 18.7 12.4 0.080 0.005

C. elegans 2.65 2.25 0.28 0.05

Characteristic path length L and clustering coefficient C for three realnetworks, compared to random graphs with the same number of vertices(n) and average number of edges per vertex (k). All three networks showthe small-world phenomenon L > L random but C >>Crandom.

6º

DIG

ITA

L C

OLL

AG

E:

PATR

ICK

MC

KE

LVE

Y

S A N T A F E I N S T I T U T E B U L L E T I N • F A L L 1 9 9 9

world network. The scientists were careful to note thatthese examples were not handpicked. They were cho-sen for their inherent scientific interest and complete-ness of available data. What this research suggests is thatthe small-world phenomenon may be common for manylarge networks found in nature; it is not merely an arti-fact of an idealized model.

NETWORK STRUCTURE AND THEBEHAVIOR OF DYNAMICAL SYSTEMS

An obvious question that arises from these conclu-sions is: What impact might this phenomenon have onthe dynamical behavior of a distributed system? Saywe’re looking at social structure: What is its role in gen-erating globally observable, dynamical features in asocial system? In general, if a set of relatively smallchanges to the edge set of a graph can have a dramaticimpact on its global structural properties, might thesame changes affect the behavior of dynamical systemsthat are coupled according to such a graph?

The question is far from straight forward. Even inidealized systems whose behavior offers a relativelyclear interpretation, the relationship between structureand dynamics builds on more than one factor. One mustconsider network structure, as well as a whole literatureof distributed dynamical systems, which again tradition-ally assumes that the relevant coupling topology iseither completely ordered or completely random.However, Watts and Strogatz’s work suggests that real-world networks may combine significant elements oforder and randomness with resultant properties (likesmall-world connectivity), that cannot be captured byeither traditional approach. Thus a realization aboutnetwork structure suggests a new question for distrib-uted dynamical systems: Do significant new dynamicalphenomena emerge when the corresponding network isa small world?

SPREAD OF INFECTIOUS DISEASEA very simple kind of distributed dynamical system

is that of a disease spreading from a small seed of initia-tors into a much larger population, whose structure isdescribed by some underlying graph. Typically, work onthe spread of diseases focuses on populations in whichcomplete mixing is assumed between elements. Withthis assumption, subsequent analysis of populationstructure can be ignored and only the relevant sizes ofhealthy, infected, and immune populations along withthe rate of infectiousness need to be known.

Watts and Strogatz took a different approach, simu-lating the spread of an infectious disease on a simplesmall-world network model. At time t=0 a single infec-tive is introduced into an otherwise healthy population.

After one unit of time, the infective is “removed”(either because it dies or becomes immune) but in thatinterval it can infect (with some probability) each of itshealthy neighbors. The process is then repeated until itreaches a steady state.

Their findings show three distinct regimes of behav-ior. In the first (for diseases with low infectiousness), thedisease infects little of the population before dying out.In the second, a highly infectious disease infects theentire population regardless of its connective topology,but the time taken to reach this steady state varies dra-matically as a function of characteristic path length of thenetwork. (Shorter path length implies faster spreading ofthe disease.) For intermediate levels of infectiousness,there is some complicated relationship between struc-ture and dynamics, which has not yet been completelycharacterized. Nevertheless, there is a clear correlationbetween critical infectiousness—the point at which thedisease infects a macroscopic fraction of the popula-tion—and the amount of randomness in the network.Beyond those conclusions, not much more can be said.However, it is clear that for this dynamical system theattractor for the global dynamics does depend on thecoupling topology.

In epidemiological terms, small-world networksimply that the level of infectiousness required for a dis-ease to grow to epidemic proportions can be highly sen-sitive to the connective topology of the population. Thismay change our way of looking at social diseases, whichare often perceived as confined to isolated subgroups ofa population. The highly clustered nature of small-world graphs can lead one to believe that a given diseaseis “far away” when in fact it is very close. In other words,when looked at on a local level, the change in structurethat causes the disease to spread much further and fastermay not be observable by an individual who has accessonly to local information.

GAMES ON GRAPHSWhile the spread of disease presents a relatively sim-

ple dynamical system, the question of cooperativebehavior emerging among competitive agents playing amany-player game is significantly more complicated.Since disease is involuntary and mechanical, it is only onthe edge of truly social behavior. However, humanbehavior as measured through games can bring up morecomplex behavior.

One such game is the Prisoner’s Dilemma, whichmodels a situation of two partners in crime who are cap-tured and locked in separate cells between which theyare unable to communicate. The dilemma concernswhether each prisoner ought to “cooperate” by remain-ing silent or “defect” by selling out their partner in

S A N T A F E I N S T I T U T E B U L L E T I N • F A L L 1 9 9 9

order to reap a reward. This game actsas a model for many kinds of interac-tions (from common pool resourceproblems to international arms negotia-tions). The main focus is that essential-ly competitive agents need to over-come the temptation to exploit eachother if they are to reap the rewards ofreciprocal cooperation.

Some early and influential researchon the subject was done by RobertAxelrod at the University of Michigan,who devised a computer tournament inwhich players (computer programs)using different strategies competedagainst each other. Axelrod found thatthe strategies that performed best didso not by exploiting the weaknesses ofothers, but by eliciting cooperation, thus allowing bothsides to do well. A strategy known as “Tit-for-Tat”devised by Anatol Rapoport (on the basis of his obser-vations of human behavior two decades earlier) emergedas the most effective. Its rules are simple: “Cooperate atfirst and then, on succeeding turns, do whatever theother player did on the previous turn.” The trademarksof TFT—being “nice,” “forgiving,” “retaliatory” and“transparent”—emerged as generic markers for anystrategy to be successful in a sufficiently heterogeneousenvironment.

In much of the early work on the Prisoner’sDilemma, the population is treated as essentially struc-tureless, a reasonable starting assumption, but not veryrealistic for large social systems. Lately, interest hasemerged into the evolution of cooperation in the pres-ence of population structure, but generally this work hasfocused on either one- or two-dimensional space. Wattsand Strogatz, interested in this question, adopted a gen-eralized version of Tit-for-Tat that could be used byplayers located on an arbitrary graph. In their versionthe formulation was entirely local; players could onlyreact to the conditions within their immediate social andtemporal neighborhoods.

Similar to the case with the spread of disease, Wattsand Strogatz found three regimes of behavior in theirsimulation. For large h, where now the variable para-meter is the average hardness (h) of players—the “hard-er” a player is, the more reluctant it is to cooperate.Regardless of population structure, cooperation alwaysdies out; for sufficiently low h, cooperation always dom-inates, but the timescale depends quite sensitively onthe fraction of shortcuts; and for intermediate h, howmuch cooperation succeeds depends on the fraction ofshortcuts. What became clear throughout is that as the

number of shortcuts increases, coopera-tion is increasingly difficult to sustain.This is due to the fact that cooperationin the Prisoner’s Dilemma contextrequires that potential cooperatorsinteract with each other preferentially.In this case, in a highly clustered graph,cooperators located in the same clustercan survive, even thrive, even within anoncooperative majority. However, theintroduction of even a few shortcutscan weaken the cooperators’ position.Those who believe in altruism will behappy to know that if the greater popu-lation is sufficiently predisposed tocooperate in the first place, then coop-eration will spread much faster in asmall world than in a large one.

For those seeking to optimize both the spread andsustenance of cooperation, the primary task is striking abalance between high clustering and speedy transmis-sion. Since small-world topology exhibits both thesefeatures it may be truly useful to such an endeavor.However, the matter is not straightforward, even for thismodel, because for any given constant population struc-ture, any change in the disposition toward cooperation(hardness) can tip the balance from rapid growth torapid decline. Nevertheless, in the design of organiza-tional structures, the idea of optimizing between clus-tering and length may be a useful concept.

NEXT STEPSThe work by Watts and Strogatz suggests that dis-

tributed systems can exhibit dramatically differentbehavior within the structural context of small-worldnetworks. In fact, the research has set off a smallavalanche among researchers in both the natural andsocial sciences to explore the implications of the small-world phenomenon. At the Institute, Watts and SFIResearch Professor Mark Newman have been examin-ing the scaling properties, phase transitions, and sitepercolation properties of small-world graphs. Anotheraspect of the phenomenon will be studied in a workinggroup organized by Charles Sabel on using the intu-itions from small-world dynamics to increase the effec-tiveness of organizations solving complex problems inrapidly changing environments. As Watts notes in one ofhis early papers on the subject, the notion of small-world connectivity “may have implications in fields asdiverse as public health, organizational behavior, anddesign.” The work has just begun.

C DC (3,3) (0,5)D (5,0) (1,1)

PRISONER’S DILEMMAThe conundrum of the Prisoner’sDilemma is captured in the chartabove. Defection (D) yields either 5 or1, and cooperation yields either 3 or0. The rational decision is to defect. Ifboth prisoners are rational, since theyboth have the same information, bothdefect, yielding a payoff of 1 each. Ifboth prisoners cooperate—if bothremain silent-—both will earn a 3, aconsiderably better outcome than thatgenerated by the rational action.

PH

OTO

: B

EN

N M

ITC

HE

LL/I

MA

GE

BA

NK

16 S A N T A F E I N S T I T U T E B U L L E T I N • F A L L 1 9 9 9

Homo Reciprocans:Political Economy

and Cultural EvolutionBy Cosma Rohilla Shalizi

AR

T:

HO

LLY

RO

BE

RTS

, “O

LD D

OG

” O

IL O

N S

ILV

ER

PR

INT

ON

PA

NE

L, C

OU

RTE

SY

LE

WA

LLE

N C

ON

TEM

PO

RA

RY,

SA

NTA

FE

Position scan

strip in seps

S A N T A F E I N S T I T U T E B U L L E T I N • F A L L 1 9 9 9 17

p=1

English-speaking social science, espe-cially economics, is dominated by atradition going back to Adam Smithand the other late 18th- and early19th-century British political econo-mists and historians of civil society.It focuses on individuals, and seestheir acts and choices as primary.Larger entities—markets, states,institutions, cultures, and classes—are shorthand ways of speakingabout patterns in the acts of manyindividuals.

Partly because they lend them-selves to precise, mathematicalexpression, individualist theorieshave proven theoretically insightful,practically useful, and surprisinglypowerful. They are also basicallyunrealistic. The standard individualeconomic agent, Homo economicus,has been called a “hedonisticsociopath.” He also has no culture atall, and is far too smart. Nobody, noteven exponents of “rational choice”theories, is much like Homo econom-icus, which is good for humanity, butbad for those theories.

There is another social sciencetradition, going back to Herder,Hegel, and other German contempo-raries of Smith, which evades theseproblems by focusing on collectiveentities like cultures and classes:these entities, proponents say, arereal, and they do things to people;indeed, they shape the people whobelong to them in fundamental ways.This isn’t much of an alternative,however, because it amounts to say-ing that cultural effects are producedby—culture. This is like saying thatopium puts people to sleep becauseit possesses a “dormative virtue.”This is a dilemma for social scien-tists: do they invoke incredible crea-tures like Homo economicus, or vac-uous entities that don’t really explainanything?

Though most find Homo econom-icus an implausible caricature ofhuman behavior, for want of anyreplacement, he has had to do.However, one of the most exciting

developments in the social sciencesin recent years is the emergence ofsomeone to take his place—calledHomo reciprocans by SamuelBowles, a professor of economics atthe University of Massachusetts atAmherst and a member of SFI’s coor-dinating committee for its KeckFoundation evolutionary dynamicsprogram. Bowles described “recipro-cans” during his talk “SocialOrganization and the Evolution ofNorms” at SFI’s May 1999 ScienceBoard Symposium which focused on“Humans and Other Social Animals.”

Perhaps the most striking way tointroduce this character is with someresults from experimental economics.Take public goods games: the experi-menter gives his subjects somemoney and explains that they canchoose, separately, how much tokeep and how much to contribute to acommon pool, which will be, say, dou-bled, to pay for a benefit in which allwill share equally. The payoffs aresuch that contributing nothing maxi-mizes one’s individual gains. In such“collective action” situations, Homoeconomicus contributes nothing, andhopes to exploit everyone else. Inreal-life experiments, however, few,often less than a half of the subjects,start out doing this. When given theopportunity, experimental subjectscan be surprisingly determined topunish those who cheated them,even at considerable cost to them-selves. More surprisingly, this is trueeven on the last round of the game,when they couldn’t hope that punish-ing the cheaters now would changetheir behavior in the future. Homoeconomicus, by contrast, realizesthat punishing cheaters under theseconditions is, like contributing to thecommon pool, a pure waste ofmoney, and so refrains from doing so.

H. L. Mencken once defined con-science as “the inner voice thatwarns us somebody is looking,” andsocial scientists and biologists haveoften interpreted apparently gener-ous acts as self-interest in disguise.

But consider the results of the “ulti-matum game,” with two players, anexperiment that has been carried outin over one hundred studies in twentycountries with highly consistentresults. The experimenter picks aplayer at random, hands him a wad ofcash, and tells him to divide itbetween himself and the other player.The second player can either acceptthe offer, in which case they split thepot as agreed, or reject, in whichcase both get nothing. But Homoeconomicus, playing against anothereconomicus, offers only one cent,which is accepted. Most people offerbetween forty and fifty percent, androutinely reject offers of less than athird, even in one-shot games (wherethere’s no chance for retaliation),even when the pot amounts to sever-al months’ earnings. That peoplemake large offers is striking enough,but what really rules out Homo eco-nomicus is that people reject quitesubstantial offers in order to punishothers for not cooperating, even whenit costs a lot to do so.

This suggests a very differentview of what economic agents areactually like, and thus emergesHomo reciprocans. As Bowles puts itin an essay with his long-time collab-orator and fellow U-Mass economistHerbert Gintis: “Homo reciprocanscomes to new social situations witha propensity to cooperate and share,responds to cooperative behavior bymaintaining or increasing his level ofcooperation, and responds to self-ish, free-riding behavior on the partof others by retaliating against theoffenders, even at a cost to himself,and even when he could not reason-ably expect future personal gainsfrom such retaliation.” This is cer-tainly in line with empirical observa-tions: people do produce publicgoods, they do observe normativerestraints on the pursuit of self-inter-est (even when there is nobodywatching), and they will put them-selves to a lot of trouble to hurt rule-breakers.

PH

OTO

: B

EN

N M

ITC

HE

LL/I

MA

GE

BA

NK

S A N T A F E I N S T I T U T E B U L L E T I N • F A L L 1 9 9 9 19

Besides the fact that typicallythere have been no other alterna-tives, another objection to abandon-ing Homo economicus is that he is,in his own way, a reliable standard.There is just one way of being ahyper-intelligent hedonisticsociopath, but there are at least asmany ways of being someone mostlyinclined to follow norms as there arenorms to follow. A single, definite,unambiguous prediction is, for some,superior to an endless series of“maybe” and “it could be thisnorm...on the other hand it could bethat one” predictions. There are twoways out of this, and Bowles, char-acteristically, takes both.

First, which norms a given groupof people follow is a factual ques-tion, which can be investigated. Forexample, to see whether the resultsof the ultimatum game are uniformacross very different types of soci-eties, Bowles, along with anthropolo-gist Robert Boyd, experimentalistErnst Fehr, and Gintis, have orga-nized field experiments of the ultima-tum and public goods games in adozen simple societies around theworld, including some, like theMachiguenga in Amazonian Peru,with very limited exposure to mar-kets and other modern institutions.(No noble savages have turned up sofar: the Machiguenga, in fact, are theclosest approximations to Homo eco-nomicus yet discovered.)

Second, norms do not vary inarbitrary and indefinite ways; thereare certain patterns which appear tobe common across societies. Inexperimental games, subjectsexplain their acts by saying that self-seeking behavior would not be “fair.”Fairness need not mean equality, butinequality does have to be justifiedsomehow. They must have reasonsfor it; a person may be rewarded forskill or effort; for virtue; or (a sur-prisingly common move) becausethey are more than human, at thevery least a different and much bet-ter kind of human. (This is borne out

by the experimental results: if, forinstance, you get to be the proposerin the ultimatum game by passingsome test, even a trivial one, youoffer less, and the other playeraccepts less.) Even when normsallow for inequality, they still enjoinsome reciprocity; the players acceptmutual, if not equal, obligations.

Readers familiar with evolution-ary psychology will remember theelegant experiments of LedaCosmides and John Tooby on “cogni-tive adaptations for socialexchange.” They showed that mostpeople can solve certain kinds oflogic puzzles when the problem isphrased as one of detecting peoplebreaking rules, even if those samepeople cannot solve formally identi-cal problems which are presentedabstractly or with different subjectmatter. This suggests that the capa-bility for reciprocans-type behavior issomething very deeply wired into ourbrains.

This only makes more pressingthe question, which will have alreadyoccurred to readers familiar withsociobiology, of how (if at all) Homoreciprocans can evolve and sustainitself in a population which containssome exploiters. In the presence ofsuch exploiters, natural selection willtend to eliminate organisms whichengage in unprofitable behaviors,such as helping others or engaging incostly punishments. One of the stan-dard theories of the evolution ofcooperation between relativesevades this by postulating “assorta-tive” interactions between kin—iforganisms tend to interact with theirclose relatives, which carry many ofthe same genes, then altruisticbehaviors can establish themselves,even in populations which containmany exploiters. Applying such rea-soning to non-kin, one of Bowles andGintis’ most recent papers showsthat Homo reciprocans could haveevolved during the Paleolithic era in asimilar way: reciprocators could haveproliferated in a population, even if

their fitness was reduced by uphold-ing norms, so long as reciprocatorswere more likely to find themselveswith other reciprocators in well-ordered groups capable of survivingbouts of material scarcity andattacks that norm-flouting groupscould not endure.

Since norms differ from place toplace and time to time (sometimeseven within a single society at a sin-gle time), we would like a theory thatexplains what determines dif fer-ences in norms. Such a theory wouldalso help close a significant gap incurrent individualist models. Theyregard agents’ preferences as“exogenous,” as fixed, given—andinexplicable. The question is how tomake them “endogenous,” to bringthem within our models?

The acquisition of norms andpreferences (and other bits of cul-ture) is an abiding concern forBowles; his first book with Gintis, onthe role of public education in mod-ern America, was, precisely, a studyof how people are acculturated, anda study of the ways in which what islearned is affected by economic cir-cumstances. Acculturation is an indi-vidual-level process: we get it notfrom our culture, but from parents,siblings, other relatives, neighbors,playmates, colleagues, and, ofcourse, teachers. Sometimes, as informal schooling in developed coun-tries, this is a very deliberateprocess, and people are taught cer-tain skills, beliefs, norms and prefer-ences, because these are economi-cally useful. In other cases, there isa mutual influence between socialand economic organization and cul-ture: people are apt to imitate thosewho achieve social success.

This kind of “replicator dynamic”is actually easier to model than isdeliberate instruction, using exten-sions of models from population biol-ogy originally developed by MarcusFeldman with Luca Cavalli-Sforza, aswell as the related evolutionary mod-els of Rober t Boyd and Peter

20 S A N T A F E I N S T I T U T E B U L L E T I N • F A L L 1 9 9 9

Richerson. Individual rewards in suchmodels are affected both by materialcircumstances and by economic andsocial organization—“who meetswhom, to undertake which joint activ-ities, with what payoffs and opportu-nities to acquire new traits,” asBowles puts it. This structure ofsocial interactions, Bowles shows,controls the rate and direction of thedifferential replication of tastes,habits and norms, whether the repli-cation is genetic, cultural, or somecombination of both.

What happens when material cir-cumstances and social structurechange? In particular, what happenswhen those who follow previouslyaccepted norms are liable to fail,whereas those who break them, whoadhere to some other set of norms,may be seen to prosper? Such a sit-uation is unsustainable; ultimately,either the norms have to change, orthe material situation does. Becausemost people are merely Homo recip-rocans, and not “rebels in defense oftradition,” it’s safe to bet on thenorms changing, but this can be alengthy process. One very commonreaction, for instance, to introducingmarkets, trade, and a money-econo-my into societies which previouslydidn’t have such institutions is toproduce fringe groups of people whoare “rootless”—cut off from theolder social groups, much moregiven to the pursuit of individual self-interest. (Bowles’ interest in culturalevolution was awakened by witness-ing precisely such a sequence ofevents as a teacher in a remote partof Nigeria in the early 1960s: “Whathappened in two hundred years ofEuropean history unfolded before myeyes in the course of a couple ofyears.”)

In highly marketized societies,careful, intelligent, competitive maxi-mizers of personal gain, unfetteredby sentiment or scruple, can do verywell for themselves. Certain restric-tions on competition are enforced(CEOs of other software companies

can’t have Bill Gates assassinated),but these have very little culture-spe-cific content, and continual effortsare made to remove any lingeringspecificity. It’s not so much that mar-kets make people into Homo eco-nomicus, but that they present situa-tions which evoke behavior thatresembles his, and reward it. (Inexperimental Prisoner’s Dilemmagames, subjects tend to cooperate ifthe game is called “Community” anddefect if it’s “Wall Street.”)

As Adam Smith knew, the institu-tions of the market can only work ifmany people (e.g.,police, judges, par-ents, soldiers) donot, in the line ofduty, act like Homoeconomicus at all,but instead actmore like Homo rec-iprocans. Balancingtwo dif ferent,incompatible setsof norms—one forthe marketplace,the other for thehome, and for rela-tions with friendsand workmates—is not an easy task,and there is a natural tendency forthe balance to tilt in one direction orthe other, for the domain governedby one set of conventions to grow atthe expense of the other’s. Nobodyreally knows whether this will happenin our case, or whether we’ll contin-ue our uneasy impersonation ofHomo economicus, or even whetherwe’ll hit upon new rules. Bowlesquite openly hopes for new rulesmore conducive to humans flourish-ing throughout the planet.

Bowles will coordinate a work-shop at the Institute next January.“Coevolution of Institutions andPreferences” will bring togethereconomists and other social scien-tists to discuss the dynamics of insti-tutional and individual behavioralevolution. The goal is to understandhow social interactions—defined by

markets, intergroup bargaining,firms, and other economic organiza-tions—shape the evolution of individ-ual preferences, and in turn howthese preferences shape the evolu-tion, and in particular, the emer-gence of new economic organization.As part of the Institute’s continuingprogram on evolutionary dynamics,Bowles is planning subsequent work-shops including one on the role ofgroup formation, the distinctionbetween insiders and outsiders, andgroup extinctions in evolutionaryprocesses.

Bowles’ papers are available athttp://www-unix.oit.umass.edu/~bowles

Position

scan

strip in

seps

S A N T A F E I N S T I T U T E B U L L E T I N • F A L L 1 9 9 9 21

PURSUINGCOMPLEXI-

PH

OTO

S:

JULI

E G

RA

BE

R

22 S A N T A F E I N S T I T U T E B U L L E T I N • F A L L 1 9 9 9

The undergraduate interns toiling in SFI’s offices thissummer face an interesting challenge. Most of the insti-tutions from which they come offer only limited coursework in the area of complexity studies, and yet the stu-dents are hungry for knowledge of this approach. Theseundergraduates have chosen to explore this new scienceout of their own initiative and curiosity, rather than aspart of an educational track laid before them.

So, what will they gain?“We’ll be prophets,” said Russ Tedrake, jokingly,

when the undergraduates convened for a roundtablediscussion on an SFI patio overlooking Santa Fe. Russ isone of seven interns who is spending 10 weeks partici-pating in this National Science Foundation supportedannual Research Experience for Undergraduates (REU)program. Each undergraduate is matched with one ormore mentors with whom they work on aresearch project; often these mentorshipsresult in continuing collaborations.

Russ tempered his “prophet” joke byadding, “There are in fact quite a few atMichigan who have a broader perspec-tive. I’ll be one of them.” Russ’s experi-ence in his undergraduate work has beendifferent from many SFI undergradinterns. At the University of Michigan hebenefits from the presence of such schol-ars as John Holland and Rick Riolo. “Butall the complexity courses are graduatecourses,” he said, which means that evenat an institution where complexity is acommon word, courses are not immedi-ately available to undergraduates. Forthis reason his experience at SFI hasbeen rich. “It’s opening my eyes,” hesaid.

While at the Institute, Russ is work-ing with SFI Postdoctoral Fellow Tim Hely attemptingto further the results from a recent paper by PetrMarsalek, Christof Koch, and John Maunsell on therelationship between synaptic input and spike outputjitter in individual neurons.

Looking enthusiastically around at the group, JamesBrink, a sophomore majoring in cognitive science, com-puter science, and mathematics at Indiana University,said his experience at his home institution has involveda considerable amount of complexity science. “A lot ofissues we deal with here at SFI are touched on by peo-ple at IU. But there’s not a complex systems departmentthere. A few of the undergrad cognitive science pro-grams discuss issues being discussed here, so I guessyou could say it’s not a unified program but is scatteredaround different departments.”

Brink is working with Resident Research ProfessorsJames Crutchfield and Cris Moore, using classical infer-ence techniques to attempt to give/reproduce the struc-ture of data produced by a quantum source.

Meanwhile, Jessica Kleiss expressed a similar experi-ence at her home institution. “At MIT I would say thereare a lot of studies pertaining to complexity, but they’rein small, hidden pockets,” said this junior majoring inmathematics and atmospheric sciences with a concen-tration in physics. “You can’t minor in complexity stud-ies, but both the Media Lab and the ArtificialIntelligence Lab deal a lot with complexity issues, andthe math department does a good deal of nonlinearstudies.”

While at SFI, Jessica’s mentor is SFI ExternalFaculty Member Alfred Hubler from the University of

Illinois. Their project involves studyingthe motions of a spring-mass systembeing fed through a viscous medium,like oil. They have found different statesof motion, like waves, loops and kinks inthe springs, and they’re using the resultsto gain insight about fluid dynamics.

Next, Amy Nelson told the group howexposure to complexity science has uni-fied her interests. She comes fromStanford University, where she’s a juniormajoring in philosophy with a concentra-tion in metaphysics and epistemology.Since she encounters little exposure tocomplexity science at Stanford, she’seager to explore the approach during thissummer experience. “I’m identifyingpatterns in areas I’m interested in andfinding relationships that I might nothave seen otherwise,” she said.

Amy is also working with Tim Helyinvestigating the role of the synchronicity of neuronalfiring rates in the perceptual task of feature binding.The project investigates the strategy of temporal bind-ing in the selection and organization of information fromhighly parallel and distributed functional areas in thecortex.

The group sat silent for a moment as if ponderingAmy’s notion of unifying interests. Then Matt Bell, asophomore at Stanford who is majoring in computer sci-ence and psychology, offered an interesting perspectiveon why complexity science isn’t popular in computerscience labs at his home institution. “For better or forworse, the computer science department at Stanford isclosely tied to Silicon Valley. So most researchers arevery concerned about applications. Consequently,there’s little interest in research that does not have

PURSUINGCOMPLEXI-

TY

S A N T A F E I N S T I T U T E B U L L E T I N • F A L L 1 9 9 9 23

financial value. However, there is one course in geneticalgorithms taught by John Koza.”

While at SFI, Matt is working with SFI ResearchProfessor Walter Fontana examining the evolvability ofRNA structures, both in terms of understanding theevolution of RNA itself and as a framework for lookingat evolutionary theory in general.

Next, Hunter Fraser told the group he’d heard aboutSFI from his father and then followed up with his ownresearch. “When I told people in the BiologyDepartment at MIT that I was coming to SFI, theythought I was weird. They thought I was going to studymath for the summer,” said this sophomore, whoseinterests include biological networks, the origin of life,and immunology.

His mentor is SFI Visitor Charles Sidman from theUniversity of Cincinnati. Theyare exploring epigenetic interac-tions and how they may cause thegenome to act as a complex non-linear system in determining anorganism’s phenotype, instead ofthe traditional “one gene-onetrait” idea of genetics.

Like Hunter, Roger Turner, asophomore majoring in the histo-ry of science at Brown University,found SFI on his own, rather thanthrough his home institution, andhe’s impressed with what he’sexperiencing. “It’s exciting to seethe nature of the place—all thesepeople who majored in one field,got their masters in another, andtheir Ph.D. in another,” he said.“Things have become so splin-tered in academia. People haveone small thing they know about.But here we learn to see patterns.We can study them and applythem to different systems.”

His mentor is SFI Visitor Douglas White from theUniversity of California at Irvine. Roger is assisting withthe Institute’s secondary school computer modelingworkshop. This workshop introduced high school stu-dents and teachers to the concepts of computer model-ing using the Starlogo programming environment. He’salso been studying how philosophical and ideologicalcritiques of science, particularly feminist critiques, cansuggest improvements in science education.

Roger’s comment led the students to talk moreabout their experience as interns. In some ways they seeSFI as a world without boundaries, both in terms of the

concrete and the abstract. Many commented on theschedule at the Institute, where work goes on day andnight. “You might come in at one a.m. and find all thesepeople here talking,” said Amy.

They were somewhat surprised that everyone eatslunch together. This allows the interns to converse with“major scientists.” They are encouraged by this kind offreedom at the Institute. “You can even show up atsomeone’s door and ask what they’re working on,” saidAmy. “I’m struck by how people communicate acrossdisciplines here. I’ve never seen people working andtalking together to that extent.”

This comment led to discussion about where the SFIexperience might lead them. James said his time at theInstitute will help him explore new areas when he returnsto IU. “From this experience, I’ll be inclined to talk to

people from different projects.” Meanwhile, Russ introduced

the question of getting too broadtoo fast. “I have a professor whowould never have gotten broaduntil he got a grounding in onefield. So while I’m here I try tolisten to what people say andtake away what relates to me, towhat I’m interested in.”

Jessica countered, “Some sayin grad school you’ll learn almosteverything about almost nothingor almost nothing about almosteverything.”

Most of the group chuckledat this, while some nodded theirheads.

“But the undergrad yearsare the best years to explore,”said Matt. “It’s then that youhave a chance to get a generalfoundation.”

Amy, the philosophy major,threw in a joke, “What I’ve been exposed to here hasheightened my prospects beyond bartending when Igraduate.”

Everyone laughed at that one. All of the undergraduates agreed their experience at

SFI would expand their ways of thinking and help themdesign their own tracks into the future. Said Roger,“There’s a wealth of interesting ideas flowing aroundhere.”

Lesley S. King

clockwise beginning in the upper left: Hunter Fraser,James Brink, Amy Nelson, Jessica Kleiss, RogerTurner, Matt Bell, and Russ Tedrake.

24 S A N T A F E I N S T I T U T E B U L L E T I N • F A L L 1 9 9 9

T R U S T E E S

John Koza is a consulting professor in the MedicalInformatics Section and the Electrical EngineeringDepartment at Stanford University. He is president of ThirdMillennium Venture Capital Limited and of GeneticProgramming, Inc. Koza is the author of numerous papersand serves on the editorial boards of Genetic Programmingand Evolvable Machines, IEEE Transactions on EvolutionaryComputation, Evolutionary Computation, and Artificial Life.He is the author of three books on genetic programmingincluding Genetic Programming III: Darwinian Invention andProblem Solving, published this year.

Ford Rowan is a former NBC news correspondent andhost of International Edition, a weekly program on publictelevision. He is the principal author of Crisis Prevention,Management and Communications (1991). Rowan has writ-ten widely in the media and communications field, includingarticles on topics such as news ethics and information tech-nology. Rowan has taught at Northwestern University’sMedill School of Journalism and at the University ofSouthern California’s Washington Public Affairs Center.

S C I E N C E B O A R D

Frances Arnold is professor of chemical engineering andbiochemistry at the California Institute of Technology. Arnoldreceived her Ph.D. in chemical engineering from theUniversity of California at Berkeley in 1985. At Caltech,Arnold’s group is applying state-of-the-art methods toaddress central issues in protein design and the evolutionof enzymes and biosynthetic pathways. This researchrequires contributions from a variety of disciplines, includingbiochemistry, molecular biology, chemical engineering,chemistry, and applied physics.

Marjorie Blumenthal is executive director of theComputer Science and Telecommunications Board of theNational Research Council, which is part of the NationalAcademy of Sciences. One of the purposes of the Board isto foster interaction among computer science, computingand telecommunications technologies, and other pure andapplied science and technology. Last year Blumenthal was avisiting scientist at the MIT Laboratory for Computer Sciencewhere she designed and taught a graduate seminar on com-puting and public policy.

New Members Namedto SFI Boards, External Faculty

PH

OTO

: D

AN

BA

RS

OTT

I

Manfred Eigen, 1967 Nobel Laureate in chemistry, pio-neered and remains a leader in the field of molecular evolu-tion. His interests in biochemistry range from hydrogenbridges of nucleic acids, through the dynamics of codetransfer, to enzymes and lipid membranes. Biological controland regulation processes, and the problem of the storage ofinformation in the central nervous system also occupy hisattention. In 1957 Max Planck Institute in Göttingen appoint-ed Eigen a Scientific Member, where in 1964 he became thehead of the Institute. In 1967, he was elected managingdirector of the Institute for a period of three years. At thesame time he was appointed to the Scientific Council of theGerman Federal Republic. Eigen is recipient of numerousprizes in addition to the Nobel; these include ForeignHonorary Member of the American Academy of Arts andSciences and the Linus Pauling Medal of the AmericanChemical Society.

Hans Frauenfelder is director of the Center for NonlinearStudies at Los Alamos National Laboratory. During his morethan 50 years of research in physics, Frauenfelder hasmoved through a number of different fields. He notes, “Istarted by studying nuclear energy levels, explored the sur-face effects with radioactivity, discovered perturbed angularcorrelation, helped elucidate parity violation in the weakinteractions, used the Mossbauer effect, and finally beganto investigate the physics of proteins. I find biomolecularphysics as fascinating (or more challenging) than any otherbranch of physics and continue to do work in this field. Oneof my goals at the Center for Nonlinear Studies is to fostermore interaction between theory and experiment, andincrease work in biological physics.”

Mimi Koehl is a professor in the Integrative BiologyDepartment at the University of California, Berkeley. Herresearch involves the application of fluid dynamics and solidmechanics to the study of biological structure. The aim is tobetter understand basic physical rules that apply acrosstaxa and to provide tools for understanding and predictinghow organisms interact physically with each other and withtheir abiotic environments. Koehl’s work places a strongemphasis on field work as well as on laboratory experimen-tation, and investigates structure and function on variouslevels of organization including tissue, organismal, and envi-ronmental.

Laura Landweber received her Ph.D. from HarvardUniversity in 1993. Her area of study was biology in theDepartment of Cellular and Developmental Biology. She hasbeen an assistant professor of biology in the Department ofEcology & Evolutionary Biology at Princeton since 1994. Arecipient of a Burroughs Wellcome Fund New InvestigatorAward in molecular parasitology, her main interest is theevolution of biological information processing, or complexmolecular systems, both in test-tube experiments in the lab-

oratory and in organisms as far ranging as ciliates or try-panosomes. Her work on “gene unscrambling” and RNAediting in these organisms offers a fresh way of thinkingabout the construction of functional genes from encryptedpieces of the genome, as biological computation.

Harold Morowitz is Clarence J. Robinson Professor ofBiology and Natural Philosophy at George Mason University.Morowitz became a Robinson Professor after teaching atYale University as professor of molecular biophysics and bio-chemistry and serving for five years as master of PiersonCollege. The author of several books, Morowitz has writtenextensively on the thermodynamics of living systems, aswell as on popular topics in science. In his current research,Morowitz is investigating the interface of biology and infor-mation sciences and continues his exploration of the originsof life. His books include The Origin of Cellular Life:Metabolism Recapitulates Biogenesis and The Facts of Life(co-authored with James Trefil). He is director of theKrasnow Institute for Advanced Study and editor-in-chief ofthe journal Complexity.

Mitchel Resnick is currently associate professor at theMedia Laboratory at Massachusetts Institute of Technology,where he holds the LEGO Papert Chair. Resnick’s researchinterests include the role of technology in learning and edu-cation, design of computational systems for nonexperts andchildren, decentralized systems and decentralized thinking,informal learning environments, and learning in virtual com-munities. Resnick co-developed LEGO/Logo, a computer-con-trolled construction set and PlayWrite, a reading/writing com-puter program. He also developed Starlogo, a programmablemodeling environment designed to help students exploredecentralized systems and self-organizing phenomena.

Shripad Tuljapurkar is president and chief scientist atMountain View Research (MVR) a population-scienceresearch firm in California. MVR creates innovative tools forpopulation forecasting. Tuljapurkar has a 1976 Ph.D., andhas held faculty positions at Portland State University, theUniversity of California at Berkeley, and Stanford University.His research focuses on uncertainty in human, ecological,and evolutionary processes. Tuljapurkar is author and co-author of numerous scientific papers and two books. In1990, he was elected Fellow of the American Association ofArts and Sciences, and in 1996 received the Mindel Shepsaward from the Population Association of America.

S A N T A F E I N S T I T U T E B U L L E T I N • F A L L 1 9 9 9 25

26 S A N T A F E I N S T I T U T E B U L L E T I N • F A L L 1 9 9 9

E X T E R N A L F A C U L T Y

Two former members of the Institute’s residentialresearch community have become members of the SFIExternal Faculty:

W. Brian Arthur is Citibank Professor at the Santa FeInstitute and PricewaterhouseCoopers Fellow. From 1983 to1996 he was Dean and Virginia Morrison Professor ofEconomics and Population Studies at Stanford University.He holds a Ph.D. from Berkeley in operations research, andhas other degrees in economics, engineering and mathe-matics. Arthur was the first director of the EconomicsProgram at the Santa Fe Institute in New Mexico; and he cur-rently serves on the Institute’s Board of Trustees. Arthurpioneered the study of positive feedbacks or increasingreturns in the economy—in particular their role in magnify-ing small, random events in the economy. His current inter-ests are the economics of high technology, the “new econo-my” and how business evolves in an era of high technology,cognition in the economy, and financial markets.

Melanie Mitchell received a Ph.D. in computer sciencefrom the University of Michigan in 1990. From 1992 to1999 she was research professor at the Santa Fe Institute,and directed the Institute’s program in adaptive computa-tion. She is currently a technical staff member in theBiophysics Group at Los Alamos National Laboratory.Mitchell’s research interests include intelligent systems andmachine learning; evolutionary computation and artificiallife; decentralized parallel computation in spatially extendedsystems such as cellular automata; and cognitive science,particularly computer modeling of perception and analogy-making, emergent computation and representation, andphilosophical foundations of cognitive science.

OTHER EXTERNAL FACULTY MEMBERS BEGINNING TERMS ARE:

Elizabeth Bradley received her Ph.D. in electrical engi-neering and computer science from MIT in 1992. She isassociate professor at the University of Colorado at Boulder,holding a joint appointment in the Computer Science andElectrical and Computer Engineering Departments. Herresearch interests focus on artificial intelligence, or com-puter tools that autonomously analyze and/or design things,and on techniques for characterizing and exploiting theunique properties of chaos. Bradley, currently a PackardFellow, also holds the 1999 John & Mercedes PeeblesInnovation in Teaching Award.

Stephen Lansing is Professor of Anthropology, NaturalResources, and Environment at the University of Michigan.Lansing’s research has involved the study of the relation-ship between Balinese religious systems, farming, andhuman ecology. He has demonstrated that ideology hasstrong control over land use and other environmental factorson this island nation. In 1995, he was the recipient of theJ.I. Staley Prize from the School of American Research forhis book Priests and Programmers: Technologies of Power inthe Engineered Landscape of Bali. He currently has a bookin preparation titled Ecology, Complexity, and Social Theoryfor Princeton University Press.

University of Iowa economist Scott Page’s work involvesthe application of computational and complex systemsmethods to economics and political economy. Some of hiswork has focused on how models with interactive agents canilluminate political phenomena such as party platform char-acteristics and the advantages of incumbency. Page hasalso done research into diversity and problem-solving and isdeveloping a theory of how different skills produce syner-gies that result in powerful aggregate problem-solvingcapacity. With John Miller, Page has been co-director of SFI’sGraduate Workshop in Computational Economics since1995, and he serves on the steering committee of theInstitute’s Fellows-at-Large initiative.

John Reinitz is associate professor at the BrooksdaleCenter for Molecular Biology, Mount Sinai Medical School.Reinitz performs both theoretical and experimental studieson mechanisms of segmentation genes expression duringearly Drosophila embryogenesis. The qualitative character-istics of segmentation genes expression patterns are usedas variables in a dynamic model for genes expression pat-tern formation. The validity and utility of the model has beendemonstrated by its successful application to several impor-tant problems of embryo development.

David Stark is the Arnold A. Saltzman Professor ofSociology and International Affairs and chair of theDepartment of Sociology at Columbia University. Stark is

S A N T A F E I N S T I T U T E B U L L E T I N • F A L L 1 9 9 9 27

MOTOROLA’S GALVIN TAKES SFI POST

Robert W. Galvin was named chairman of the Santa Fe Institute Board ofTrustees at its May 1999 meeting. Galvin started his career at Motorola in1940. He held the senior officership position at the company from 1959 until1990, when he became chairman of the Executive Committee. Currently, heserves as a full-time officer of the company.

Galvin attended the University of Notre Dame and the University of Chicago,and is now a member and was the recent chairman of the Board of Trusteesof the Illinois Institute of Technology.

He has been awarded honorary degrees and other recognitions, includingelection to the National Business Hall of Fame and receipt of the NationalMedal of Technology in 1991.

Continuing as vice-chair of the Board is Robert J. Denison, founder andchairman of First Security Management, Inc.

Galvin and Denison replace David Liddle and Robert Maxfield who haveserved as chair and vice-chair since 1994. Both Liddle and Maxfield remain onthe Board of Trustees.

“I shall always be grateful to David Liddle and Bob Maxfield for their helpin moving SFI forward,” said SFI President Ellen Goldberg at the May meeting.“I look forward to their continued involvement as active members of our Board.

“I am delighted that Bob Denison has agreed to continue on as vice-chair-man of the Board. His knowledge of fiduciary issues provides all of us withsound advice, and his breadth of knowledge in a number of fields provides theInstitute with a wonderful resource for counsel.

“As to having Bob Galvin agree to become chairman of the Board, I am trulythrilled. He is a visionary who will help move SFI to new heights. I look forwardto continuing to work with the entire Board of Trustees, and I thank all of themembers for their commitment to SFI. Their guidance has been and continuesto be very much appreciated.”

currently doing research on neworganizational forms among firms inManhattan’s Silicon Alley. In postso-cialist Eastern Europe, he studiedhow interfirm networks facilitatedand impeded economic restructur-ing. His recent publications includeHeterarchy: Distributed Intelligenceand the Organization of Diversity,for thcoming from PrincetonUniversity Press. PostsocialistPathways: Transforming Politics andProperty in Eastern Europe, withLaszlo Bruszt, Cambridge UniversityPress, is a comparative study of theopportunities and dilemmas posedby the simultaneous extension ofproper ty rights and citizenshiprights.

Andreas Wagner is assistantprofessor in biology at the Universityof New Mexico. Wagner received hisPh.D. from Yale University in 1994.Following fellowships at the Institutefor Advanced Study in Berlin andthen with Leo Buss at Yale, Wagnerwas a postdoctoral fellow at theSanta Fe Institute from 1996 to1998. While at SFI Wagner pursuedresearch projects on the evolution ofgenetic redundancy and developedmathematical and computationaltechniques for the analysis of wholegenomes. His SFI redundancy workrests on a population geneticsmodel based on the fact that redun-dancy provides protection againstdeleterious mutations. Within thisframework he was able to show thatselection cannot only maintain butincrease genetic redundancy in largepopulations.

28 S A N T A F E I N S T I T U T E B U L L E T I N • F A L L 1 9 9 9

Language Issues

BY KEN BAAKE

Issues of language surface fre-quently at the Santa Fe Institute,where discussions among membersoften lead to debates about wordmeaning. One of the biggest chal-lenges in this arena is finding wordsand metaphors that can stimulateproductive scientific thinking with-out distorting that thinking.

Dynamic new concepts are con-stantly emerging at the Institute.Just ask librarian MargaretAlexander about her list of scientific“buzz words” that she uses whenshopping for books to add to thelibrary. The list is always changing,she says.

The word “complexity,” one ofthe central concepts at the Institute,presents possibly the biggest defini-tion challenge. A number of scien-tists acknowledge that they do notreally know what it means. Theapparent vagueness of the term,however, may be what makes it sovaluable as a catalyst for thought. Aword like complexity is new andunresolved; it is not an inert tool ofscientific description, but rather anidea whose meaning evolves as itinteracts with researchers.

SFI President Ellen Goldbergsays that complexity involves“interacting parts with very simplerules,” but she is quick to add thatthe term does not reduce to a con-stant definition across disciplines.For that reason Goldberg likes theterm; it is flexible and permits mul-tiple definitions. Former SFI presi-dent George Cowan also accepts thepolysemantic nature of the word.“Its chief value is that it embraces anumber of possible systems,”Cowan says.

The definition challengesincrease when you bring metaphorsinto the rhetorical equation.

Language scholars define metaphorin various ways, but most say that itis a rhetorical device for transportingknowledge by using a word thatbrings connotations from one fieldinto play in another field. The wordhas at its roots the Greek word“phora,” which means locomotion.SFI theoretical chemist WalterFontana explains the power of suchlocomotion in SFI science.“Metaphor makes people realizethat certain old questions can becast as new ones. It triggers newthoughts and speculations.”

It can also trigger a lot of discus-sion. Take for instance the publiclecture by Indiana University politi-cal scientist Elinor Ostrom at lastsummer’s Integrated ThemesWorkshop. Ostrom used a game the-oretical approach to look at publicpolicy questions. She examined theways in which a community decidespublic policy, such as the allocationof irrigation rights among farmers.

Much of the question andanswer portion of her talk con-cerned debate over the word“rules,” a term Ostrom used todescribe how members of the com-munity would interact with eachother and what policy decisions theywould make. But one researcherquestioned the use of the word“rules” because it implies legal reg-ulations, while much human inter-

action has nothing to do with laws,he said. Clearly, the semanticaldebate centered around the specificuse of language in different scientif-ic disciplines. As Ostrom explainedlater, “rules” for biologists might beobserved regular behavior or strate-gies, while “rules” for political sci-entists might be enforceable lawsagreed upon by members of agroup.

But terminology can be prob-lematic even when confined to asingle scientific discipline.Theoretical biologist MichaelLachmann calls attention to theterm “signaling” in biology as anexample of a metaphor that is rich,but also defies precise definition.What does it mean when we say ayellow bee is signaling that it is dan-gerous to eat? “Does that mean thata brown bee is not dangerous toeat?” Lachmann asks. Signalingimplies an intentionality on the partof some living agent, but the bee’scolor is purely the product of evolu-tion. Can we use a metaphor likesignaling, with all its connotations ofintentionality, to refer to a DNA-coded process? “Who intends thebee to be yellow?” Lachmann asks.

The good news is that suchquestions lead to further explo-ration. SFI scientists are constantlystretching terms to help explainnew insights. For example, theoret-

inside SFI

ILLU

STR

ATIO

N:

G &

V C

HA

PM

AN

/IM

AG

E B

AN

K

S A N T A F E I N S T I T U T E B U L L E T I N • F A L L 1 9 9 9

ical chemist Walter Fontana hastaken the common word “neutrali-ty” and used it to argue that thegenetic code underlying a com-pound such as RNA (ribonucleicacid) can change frequently withoutaffecting the performance of thecompound—until at some point thecompound seems to “suddenly”evolve to a higher level of perfor-mance. Such “neutral” changes ofthe underlying genetic sequencemake evolution possible, Fontanasays.

Still, Fontana acknowledgesthat the neutrality may only be tem-porary; eventually the compoundevolves. So, it seems appropriate toask if a change can be called “neu-tral” when it has no apparent imme-diate effect on a compound or anorganism, yet over time, in combi-nation with other changes, it leadsto a profound impact. Again, it isobvious that when scientists strug-gle with word meaning they oftenare led to new insights—and newquestions.

Fontana’s research has piquedthe interest of social scientists affili-ated with SFI who are consideringwhether seemingly unimportant“neutral” events in human historyand human economies could havemajor consequences over time. Theneutrality concept also may apply topresent-day business. Suppose youare a chief executive officer whowants to reorganize your firm,Fontana says. You know that a suc-cessful reorganization would requirea lot of small changes affecting theway employees go about their jobs,but you have no way of knowing inadvance the best combination ofchanges to make. Neutrality sug-gests that you can make smallchanges gradually—even in anuncertain environment—withoutrisk that you will sink your firm.

Clearly the way scientists stretchmetaphors and other terms helpselucidate ideas. But metaphors can

lead to distortions as well by spawn-ing unintended connotations thatinfluence society far beyond theintention of those scientists whodeveloped the metaphors as aids tocognition. Einstein’s concept of rel-ativity, for example, has rippledbeyond the confines of physics andcosmology into the world at large,where some might argue that it hasreinforced a moral relativism inWestern society. SFI physicistCris Moore presents anotherexample—the metaphor ofgenetic mapping, which impliesa certain determinism that couldpossibly lead to social eugenics.

Metaphors are so powerfulthat they can paralyze aresearcher who becomes soenamored of a term that she or heloses the flexibility to see the worldoutside of the frame of themetaphor. As Fontana says, “Youfall in line with ideas and then youbecome brittle.” For example,Moore notes that metaphors import-ed from biology into economics maydistort impressions of the economyby making it seem more naturalthan it really is. “The economy isn’ta jungle,” Moore says. Economicsystems have governments, whichare not present in the rain forest.

Other times the metaphors areso catchy that they imply quick-fix,real-world applications for a particu-lar theory that cannot be delivered,at least not in the short term.Suzanne Dulle, SFI director of busi-ness relations and external affairs,says that the language of complexityhas flooded the popular business-book press, leading to an expecta-tion among some business peoplethat they can spend a few days atSFI and come away knowing how tosave their company money. Somebusiness managers hearing aboutself-organization have asked half-seriously if they should step asideand let the company run on its own,Dulle says.

Metaphor is a major rhetoricalfocus at SFI, but it is not the onlyone. Other problems involveresearchers having to modify theirspeaking and writing according tothe needs of a lay audience. One ofthe most difficult rhetorical chal-lenges scientists face when explain-ing their research to the lay publiccomes in answering the question,

“And so what?” Dulle notes.People often wonder what theresearch means to their lives, tothe world at large.

SFI scientists also face aunique challenge when writingto an audience of specialists inanother field. At times they’refrustrated when trying to publishtheir results in scholarly journals

that may not be receptive to theunique and almost poetic languageof complexity science. For example,economist graduate researcherShareen Joshi says she often usesconcepts like “the edge of chaos”when envisioning the economy inher mind, but she would never usesuch terms in a journal article.

Joshi’s compromise with elo-quent language reveals the trueessence of metaphor as a midwife ofknowledge. Metaphor can open theresearcher’s mind to fresh scientificinsights that might remain occludedwithout the powerful cognitive“locomotion” that metaphor entails.Yet, because metaphor is so powerfuland perhaps intrinsic to human cog-nition, it also carries the dangersinherent in any powerful tool;metaphoric language can distort thetruth and imply a greater sense of sci-entific certainty than the results ofrigorous scientific exploration wouldsupport. Most at SFI seem willing toaccept this paradox; they know thatthe benefits outweigh the risks.

Ken Baake is a doctoral candidate in Rhetoric

and Professional Communication at New Mexico

State University in Las Cruces. For his disserta-

tion he is examining issues of language at SFI.

SANTA FE INSTITUTE1399 HYDE PARK ROAD SANTA FE, NEW MEXICO 87501U.S.A.

NON-PROFITORGANIZATIONU.S. POSTAGE

PAIDPERMIT NO. 49LUBBOCK, TEXAS