Bulletin Spring 2008, Vol

SFI Bulletin spring 2008, vol. 23 #1

conflict: what can it teach us— about our brains, the Internet, ecosystems, and financial markets? The Bulletin of the Santa Fe Institute is Balancing Convention published by SFI to keep its friends and and all of our researchers who contributed to this effort, be considered outliers supporters informed about its work.The and the staff for seeing it to fruition. It’s gratifying on the conventional Santa Fe Institute is a private, indepen- and Risk to Create to know that even though we support research that scientific landscape. dent, multidisciplinary research and sometimes doesn’t quite fit, we can still be successful when For example, David education center founded in 1984. Since its founding, SFI has devoted itself to Transformation we play on the conventional academic playing field. Krakauer, together creating a new kind of scientific research Although this award represents less than 10 percent with Yale History community, pursuing emerging synthe- of our funding (indeed, by design, only about one third Professor John

By Geoffrey West, President fe institute a

sis in science. Operating as a visiting is derived from government sources), it’s extremely Gaddes, is organizing nt a institution, SFI seeks to catalyze new important and we value it tremendously. We need a workshop asking to collaborative, multidisciplinary research;

The advent of a New Year is a good time to that credibility and, by extension, the approval from what extent history mpbell / s to break down the barriers between the a traditional disciplines; to spread its ideas take a retrospective look at some of the developments and the academic community, when it comes to soliciting can be viewed as a and methodologies to other institutions; achievements of last year and to speculate on where the funding from private foundations, universities, and other science. This may © Andrew C and to encourage the practical application Institute might be headed in the coming year. This issue of government agencies. not be as crazy as it of its results. the Bulletin does some of that. It explores many subjects, Regarding research trends in 2007, we’ve had some sounds; after all, both astronomy and geology are accepted

Published by the Santa Fe Institute but a common thread is that of conflict and threat. Topics tremendous successes, as evidenced by the wealth of SFI as “historical sciences.” In addition, we are planning to 1399 Hyde Park Road ranging from the threat of “malware” to the loss of diversity publications and reporting about SFI in books, articles look at law as a complex system. Can we learn something Santa Fe, New Mexico 87501, USA in ecosystems and financial markets are examined from the in professional journals such as Nature and Science, and about jurisprudence, about the conceptual framework of phone (505) 984-8800 broad, big-picture viewpoint characteristic of SFI. in popular publications such as Wired, New Scientist, The law that might impact new emerging legal questions of the fax (505) 982-0565 One of the more gratifying events of 2007 was the Atlantic Monthly, The New York Times, and many more. 21st century? home page: http://www.santafe.edu approval of our core umbrella proposal to the National For example, Samuel Bowles’ work on altruism was Another incipient area of exploration is what we are Note: The SFI Bulletin may be read at the website: Science Foundation. Our award is unique because, almost featured in The Atlantic and on National Public Radio. calling international negotiations and complexity. This http://www.santafe.edu/bulletin.html without exception, the NSF typically entertains proposals Michael Mauboussin’s financial wisdom was also featured arose because people involved in international negotiations If you would prefer to read the Bulletin on that are for specific, relatively narrow research projects, at on NPR, as was my own work on scaling and lifespan, as approached us stating that the traditional system of your computer rather than receive a printed least compared to the scale of SFI activities! Our proposal well as on cities and innovation, done with my wonderful diplomacy is broken. They suggested that this system version, contact encompasses a mix of questions ranging from anthropology group of collaborators. Many accolades came to SFI evolved in a “simpler” world of the 19th century where Patrisia Brunello at 505/984-8800, Ext. 2700 and archaeology, through economics, biomedical sciences researchers. Among those recognized were Nina Fedoroff, the negotiating paradigm was “linear and Newtonian”; or [email protected]. and biological sciences, across mathematics, computer named Science and Technology Advisor to the U.S. international relations have changed dramatically and can Editorial Staff: science and physics, all the way to the interpretation of Secretary of State, who also received the National Medal now be characterized as “complex”! Such a line of thinking Ginger Richardson quantum mechanics and the wave-function of the universe! of Science; David Sherrington won the 2007 Dirac Medal naturally leads to the Santa Fe Institute. Like all of these Lesley S. King Andi Sutherland Despite its good intentions and professed commitment and Prize from the Institute of Physics; Cormac McCarthy new areas of potential investigation we don’t yet know to cross-disciplinary, transformative research, the NSF received the Pulitzer Prize for his book The Road; and both if progress will be made by bringing together unlikely Contributors: doesn’t yet have in place a formal structure for handling Duncan Watts and I were featured in Harvard Business combinations of creative people. It is the very nature of Jenna Beck such unusually broad proposals. Needless to say, this Review’s list of Breakthrough Ideas for 2007 (which meant SFI to explore such avenues, to see if there might be new Devon Jackson Melanie Mitchell put ours at some risk. The “normal” mechanism would that two of the 20 such ideas were associated with SFI). insights provided by a “complexity” lens. Don Monroe break ours into 10 to 15 separate proposals, each In looking at the future, we have a number of exciting A continuous challenge for the Institute is identifying Julie J. Rehmeyer focused on a specific topic within some sub-discipline, research areas we plan to address. This year we began talented people who might be attracted to the spirit of Daniel Rockmore which, of course, is antithetical to a basic philosophy of to recognize that much of the work at the Institute SFI, those exciting minds who really have eclectic, wide James Trefil SFI: transdisciplinary research transcending canonical falls under the umbrella of sustainability: e.g., our tastes, and a passion for some of the big questions that John Whitfield disciplinary boundaries. programs on innovation, robustness, financial markets, transcend traditional boundaries, and who, at the same Jon Wilkins Establishing this and its implications for problems in networks, and those areas that interface with energy, the time, have the powerful technical tools, background and Design & Production: complexity as a major and exciting component of the environment, and organization. While we are not going discipline to execute their ideas. How do we identify those Paula Eastwood research landscape has arguably been SFI’s greatest impact. to solve or provide detailed models of global warming or people and get them on board here as resident faculty? www.eastwooddesignsf.com I am delighted to report that the award was a 30 percent sustainable urban infrastructures, we look to provide a Indeed, I believe this is our major challenge, but one we cover: simulation of a malware attack; Illustration by catalogtree.net for wired magazine increase over the previous one. It is to the credit of the holistic, integrated conceptual scientific framework for can meet. Our past track record and recent kudos, along

printed on recycled paper NSF that they were willing to entertain our proposal understanding some of these critical world concerns. with the many awards and publications, offer strong despite its unique character. I would like to thank them We are also looking at a series of questions that might encouragement for our future direction. t

Santa Fe Institute Bulletin SPRING 2008 The Bulletin of the Santa Fe Institute is Balancing Convention published by SFI to keep its friends and and all of our researchers who contributed to this effort, be considered outliers supporters informed about its work.The and the staff for seeing it to fruition. It’s gratifying on the conventional Santa Fe Institute is a private, indepen- and Risk to Create to know that even though we support research that scientific landscape. dent, multidisciplinary research and sometimes doesn’t quite fit, we can still be successful when For example, David education center founded in 1984. Since its founding, SFI has devoted itself to Transformation we play on the conventional academic playing field. Krakauer, together creating a new kind of scientific research Although this award represents less than 10 percent with Yale History community, pursuing emerging synthe- of our funding (indeed, by design, only about one third Professor John

By Geoffrey West, President fe institute a

Santa Fe Institute Bulletin SPRING 2008

SFI Bulletin Spring 2008 Vol. 23 #1 55 contents

When Diversity Vanishes 38

Economics and Markets as Complex Systems 44

26 Visual Understanding 50

The Evolution of Individuality and Sex 55

The Conflicted Brain 60

Patterns of Terror 18

Pablo Marquet: Guerilla Ecologist 22 66 32

Malware Wars 26

Risk in Financial Markets— Learning from Nature 32 12 60 Cherry Picking the Periodic Table: A New View of Life 66 Balancing Convention and Risk to Create Transformation SFI Community Lectures 70 Inside Front Cover Transitions 73 Cities: Large is Smart 4

The Revolutionary Lizard 9

Sleep to Repair 12 9 PHOTO CREDITS— PHOTO CREDITS— TOP RIGHT: COURTESY OF RICHARD MICHOD, University of Arizona TOP LEFT: VEER.COM; TOP RIGHT: CELIA JOHNSON / RAPPART.COM; center: paula eastwood; center right: iSTOCKPHOTO.COM RIGHT: Hannah Gal / Photo Researchers, Inc. RIGHT: indiana state university photography services

Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008

SFI Bulletin Spring 2008 Vol. 23 #1 55 contents

When Diversity Vanishes 38

Economics and Markets as Complex Systems 44

26 Visual Understanding 50

The Evolution of Individuality and Sex 55

The Conflicted Brain 60

Patterns of Terror 18

Pablo Marquet: Guerilla Ecologist 22 66 32

Malware Wars 26

The Revolutionary Lizard 9

Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 research report Cities:Large is Smart

By Jenna Beck

n an attempt to discover whether cities consume energy in the same way as biological organisms, SFI External Professor Luis Bettencourt, SFI President Geoffrey West, and their colleagues from Arizona State University and the Dresden University of Technology found that the size of a metropolis not only affects energy consumption, but also the pace at which people work, earn, walk, spend, and steal. Their study, published in the April 24th issue of the Proceedings of the National Academy of Sciences USA, lays out scaling laws that appear to connect cities of different sizes and in different nations. The most surprising of these laws suggests that a city may have a social metabolism that, as the city grows, increases faster than the population itself. I“A city’s not just a place, just a culture, what people eat, what language they speak,” says Luis Bettencourt. “It’s a space of interaction. When you put more and more people together in a city, some things increase in a very predictable and interesting way.” Using nc data from international census and statistics bureaus, the researchers compared infrastruc- I ture, energy consumption, and a handful of social phenomena for cities large, small, and in between. They found that as a city’s size increases, the infrastructure networks realize esearchers, economies of scale, while the social phenomena grow faster than the population. In other R words, the surface areas of streets and the lengths of electrical cables lag behind a city’s td / Photo L

This satellite image of North America at night illustrates the density of populations in the coastal regions as well as major inland cities, with large stretches of plains, deserts, and mountains remaining dark. Dawn is breaking over the

eastern coast of the U.S.A. PlanetaryVisions

Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 research report Cities:Large is Smart

By Jenna Beck

eastern coast of the U.S.A. PlanetaryVisions

Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 not extend to all urbanites. The report projects To the researchers’ surprise, the biological par- that “poor people will make up a large part of allel failed for the third category of data: social future urban growth,” since much of the growth is currency. Unlike an animal’s metabolism, which projected to occur in the developing world, where slows with increasing size, as city size increased, cities do not generate enough jobs to accommo- the social metabolism sped up. “Superlinear” date their populations. In light of the global ur- is the term the researchers use to describe this ban upswell, questions of how a city grows, how it metabolic growth rate, which, when graphed, is sustained and affects the environment, and how, maps a steep upward curve rather than a straight as an environment, it affects its citizens become line. “It was quite a surprise when we discovered increasingly important. The scaling laws presented this whole superlinear scaling. Like many things, in the researchers’ paper lay down a quantitative once you see it and think about it, it seems obvi- base for exploring such questions. ous,” says Geoffrey West. “But it wasn’t obvious After analyzing datasets from American, Chi- at first.” The social currency category included nese, and European government bureaus, the a range of seemingly unrelated measurements. researchers found three different scaling laws for Some of these—number of new patents, inven- three categories of data: individual human needs, tors, employment in research and development, material infrastructure, and data related to social currency. Individual human needs included housing, employment, household electrical consumption, and household water consumption. These scaled linearly for cities of different sizes, which suggests that a person’s basic requirements for work, a home, and utilities remain the same whether he lives in New York or Nantucket. Infrastructure was measured by road surface area, length of electrical cables, gasoline sales, and number of gas stations. The infrastructure scaled sub-linearly as city size increased, which Big cities will continue to grow, researchers believe, and understanding such places may help protect them from economic crisis. means it didn’t grow as quickly as the population. Violent crime per capita increases with city size (USA, 2005) as a scaling law. Here Shanghai’s notable sector, The Bund, along the Huangpu River, buzzes with activity. When more people use fewer roads, gas stations, etc., a city realizes an economy of scale. Large population growth, whereas income, crime rates, the number of small cities with populations of cities, up to a point, are a more economic way and creative employment—were loose measures disease transmission, and certain measures of fewer than 500,000 people. of distributing resources than small cities and of innovation. Others were measures of con- innovation outpace it. From an economic perspective, a growing city is rural communities. This mirrors a relationship sumption: total wages, total bank deposits, gross Understanding these phenomena is a timely usually seen as a healthy one. After all, the world’s seen in biological organisms, where large animals domestic product, and total electrical consump- endeavor. For the first time in history there are largest urban centers currently drive the world’s economize on blood vessels by pumping more tion. The researchers also took account of such roughly the same number of people living in largest economies. But while growth can benefit blood through proportionally fewer veins and things as the number of new AIDS cases and cities as in rural areas, and the United Nations a city’s economy, a city’s future is by no means capillaries. Fewer vessels support more mass, but serious crimes. projects that cities will absorb the bulk of the ensured. According to the UNFPA State of World the large animal’s metabolism slackens because As diverse as the social data were, they all population growth in the next 30 years. Big cit- Population 2007 Report, there are actually more the blood is delivered more slowly. When a city scaled superlinearly with roughly the same expo- e: © Wei yan / masterfile.comWeiyan e: © ies will grow bigger, as will the number of small people moving out of Buenos Aires, Calcutta, v grows, the highways deliver more people to and nents. This led the researchers to conclude that cities. Between now and 2015, half of the urban Mexico City, São Paulo, and Seoul than are mov- from their destinations, but traffic slows as the an underlying social metabolism was driving hoto abo growth is expected to come from an increase in ing in. Furthermore, the economic advantages do p roads become congested. their growth. “We went into the study trying to

Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 odically reset its growth rate. Such “resetting” can come from innovations that revitalize the economy, or from outside factors, such as shifts in immigration. The pattern that an ever-growing city falls into is one of successive growth cycles—each one shorter than the last as the size of the city increases. “You’re on this treadmill and you’ve got to go on making these changes, these innovative changes, faster and faster because if you don’t you’ll stagnate and collapse,” West says. “It’s probably hard to have control over these things because a city is embedded in something Total wages per metropolitan area indicate increasing per capita returns versus bigger. My interpretation is that where that cycle population size. ends up might depend upon whether you have a test the economies of scale theory, first for en- city with a good administration in place, but the ergy consumption,” Bettencourt says. “It’s only general trend and the coarse-grained scale of it is because we kept finding superlinear scaling for probably determined for you.” energy consumption—and then in everything If a growing city does not reset, the growth economic—that it became so obvious that every- will collapse when the city exhausts its resources, thing social showed increasing returns to popula- or when its infrastructure and economy cannot tion. All these quantities are rates, meaning that support the population. Bettencourt cautions their increase per capita with the size of the city that a city can also collapse when the dark side is an expression of acceleration of social life.” of social metabolism accelerates more quickly As a city grows, its social metabolism speeds than the rest: “My intuition, which is something up. Individual productivity rises (15% per person I’d love to measure, is that the bad things can re- when the city doubles) as people get busier. Aver- spond more quickly: it’s easy to increase the cost age walking speeds increase. Businesses, public of housing. It’s easy to increase crime. It’s easy to spaces, nightclubs, and public squares consume create congestion—it just happens.” more electricity. The city draws in more inven- The researchers are now looking into Bet- tors, artists, researchers, and financiers. Wealth tencourt’s intuition about the balance between increases, as does the cost of housing. negative and positive social behaviors. They are The superlinear growth rates also suggest also working to determine whether population that a city can grow indefinitely. When the re- and economic diversity increases with city size searchers plugged the scaling exponents into an and, if so, whether those factors can help protect urban growth equation, they found that cities a city from economic crisis. “What’s difficult,” driven by economies of scale were destined to adds Bettencourt, “is to create institutions that plateau, whereas those driven by innovation or promote very advanced learning and creativity.” wealth creation had the potential for unbounded Knowing more about cities just might make the growth. “Should a city have a finite size or task easier. t should it grow forever? How should it grow? You would argue about it forever if you hadn’t mea- Jenna Beck is a freelance journalist and a soon-to- sured,” Bettencourt says. be graduate of the Boston University Center for Sustaining that growth is the city’s chief chal- Science and Medical Journalism. She can be lenge. To grow indefinitely, a city has to peri- reached at [email protected].

Santa Fe Institute Bulletin SPRING 2008 research report the Revolutionary Lizard

“Who cares about lizards?” Ecology grants usually go to studies of dominant species—large animals that have enough charisma to merit an appearance on Animal Planet. So when SFI postdoc Lauren Buckley, along with a The group aims to come group of fellow ecologists, applied for funding to study the eastern fence lizard, up with a predictive model she found herself defending the quiet little crawlers. As a result, the humble liz- that could explain the ard may change the way ecologists think about global warming. Buckley and her collaborators are developing a new approach to one of the distribution of lizards, central objectives in ecology—modeling how wildlife responds to climate in hopes that their meth- change. Most of the current climate-change models fail to explain non-linear ods will eventually be migration patterns that have been observed so far, because they assume that as certain temperature zones move north or south, the flora and fauna that inhabit extended to other species. those zones will follow linearly. The models are based on the assumption that the climate an animal dwells in is the only climate in which it can survive. Buckley and her colleagues hope to come up with a better climate-change model that will predict where organisms can live based on how they forage, reproduce, and use energy. The new, “bottom up” approach has produced several models that account for non-linear responses to climate change, and Buckley’s collaboration is trying to test the niversity

U relative merits of three major contenders. The group aims to come up with a predictive model that could explain the distribution of lizards, in hopes that their methods will eventually be extended to other species. Eastern fence lizard (Sceloporus undulates) Will Cook; Duke In 2004, a study published in Nature triggered an avalanche of headlines in

Santa Fe Institute Bulletin SPRING 2008 temperatures they can gather the ecologists a basis for mod- traditional climate-envelope els for organisms as diverse as food and reproduce. Buckley eling the extent to which the models. Buckley illustrates the plants, fish, birds, and mam- calls it mechanistic modeling. lizards are able to adapt to their radical departure in a paper that mals. The group will meet And building a mechanistic environment. Buckley says will soon be published in Amer- twice a year at the National model requires not only a new that the physiological data is ican Naturalist. In the paper, she Center for Ecological Synthesis mathematical approach, but important because it indicates compares a mechanistic model in California and the National

es extensive data collection and the ability of an organism to with a climate-envelope model Evolutionary Synthesis Center c experimentation to ascertain change: “We’re trying to model by using each to “predict” the in North Carolina. the physiology of the animals in the process of evolution, but as current distributions of North They hope the distributional question. a first step, we are examining American lizards. The maps in models will be available to The physiological data for the how the variation in physiology the paper reveal a significant other ecologists by October lizard model will come from a across an organism’s distribu- difference between the old and 2009. “All the people who are laboratory at Indiana State Uni- tion allows it to live in a diver- new models, and higher overall independently working on versity. There, ecologist Michael sity of environments.” predictive success for the mech- mechanistic modeling are going iana university state photography servi

d Angilletta and his grad students The group aims not only to anistic model. Buckley explains to be in tune with one another, in lower fence lizards into special produce a good mechanistic that the paper demonstrates which I think is very cool,” the popular press. “Scientists and that if those climates shift chambers that measure oxygen model for how lizards respond that considering a lizard’s biol- Angilletta says. “If the exercise Physiological Predict Widespread Extinction or disappear, the species will consumption. They record the to climate change, but also to ogy is important when predict- turns out to be useful and can data collection by Global Warming” was the follow. Yet there is evidence that reptiles’ feeding rates, and heat get a sense of how many vari- ing its distribution. “A central show that these mechanistic allows ecologists to examine how headline that ran in The New species do not simply follow a and cool them to find the tem- ables a model needs in order to goal of the group’s research,” she models are more valuable than, the variation in York Times; “Climate Risk to single set of climactic condi- perature range at which they be predictive. Too few variables writes, “is to incorporate more say, a climate-envelope model, physiology across ‘Million Species’” on BBC News tions; data that lepidopterists remain active. Over the next make a model inaccurate, while biology, such as how organisms the way of thinking will spread an organism’s Online; and “By 2050 Warm- have collected over the past few years, Angilletta will collect too many make it impracti- interact, move, and evolve, into and other people will go out distribution ing to Doom Million Species, century indicates that certain such data from eight different cal. Using Angilletta’s data, the distribution models.” and tailor the math to their par- allows it to live Study Says” on National Geo- butterfly species in Europe have lizard populations. group will test three models of Buckley is also leading a ticular systems.”—Jenna Beck t in a diversity of graphic News. The study esti- remained stable in the face of a Once he knows whether varying degrees of complexity, broader, multidisciplinary col- environments. mated the future distributions mean temperature increase of lizards collected from different the practical and the predictive. laboration of about a dozen BELOW: Range predictions for the fence lizard vary when Here, Indiana State of plant and animal species 0.8 degrees centigrade, whereas environments have different All three of the models repre- researchers aiming to produce including population body size (green), population body size and University graduate using a correlative technique other species shifted northward. sensitivities to temperature, he sent radical departures from the mechanistic distribution mod- life history (blue), and with a 3° C temperature increase (red). student Joe called climate envelope mod- An analysis of fossilized mam- will present the data to Buckley, Ehrenberger loads eling (CEM). The technique mal remains similarly shows who will use it to model the a fence lizard into a infers species’ environmental that different species shifted in species’ environmental ranges. metabolic chamber. tolerance from the temperature, different directions and at dif- “It takes a lot more time than precipitation, and seasonality in ferent times during a climate gathering data from the litera- their present-day surroundings. change in the late Quaternary ture,” Angilletta explains. “The Though CEMs are the princi- period. CEMs can neither pre- mechanistic modeling is obvi- ple models ecologists use to predict nor explain such non-linear ously going to take more data dict how wildlife will respond behavior. because you can’t just get the to climate change, Buckley and Buckley’s team of ecologists data from museum records on- ago c

others criticize them for being pursues a new way of model- line. You have to collect data on hi c too simplistic. CEMs assume ing. They aim to predict species physiology, which is expensive that the climates in which spe- ranges based on how animals and time consuming.” niversity of cies currently live are the only metabolize energy and col- The time-consuming data U

10 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 11 temperatures they can gather the ecologists a basis for mod- traditional climate-envelope els for organisms as diverse as food and reproduce. Buckley eling the extent to which the models. Buckley illustrates the plants, fish, birds, and mam- calls it mechanistic modeling. lizards are able to adapt to their radical departure in a paper that mals. The group will meet And building a mechanistic environment. Buckley says will soon be published in Amer- twice a year at the National model requires not only a new that the physiological data is ican Naturalist. In the paper, she Center for Ecological Synthesis mathematical approach, but important because it indicates compares a mechanistic model in California and the National

10 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 11 research report Sleep to Repair

“What probing deep Has ever solved the mystery of sleep?” —Thomas Bailey Aldrich, Human Ignorance

The poet and novelist Thomas Bailey Aldrich died in 1907, just before the birth of modern sleep research. Since then, researchers have advanced the study of sleep with instruments such as electroencephalograms (EEGs), computerized axial tomog- raphy (CAT scans), and magnetic resonance imaging (MRI). Some have even begun to identify the molecular signals that accompany sleep and wakefulness. But Aldrich’s words still ring true, and the basic question of why we sleep remains a mystery. There are many hypotheses that offer explanations for the purpose of sleep, but in the absence of a convincing way to test and compare them, no con- sensus has emerged. SFI researchers Van Savage and Geoffrey West have narrowed down the quest for a sleep theory by extending their gaze beyond human beings or any single species. In a study published in the January 2007 issue of the Proceedings of the National Academy of Sci- ences USA, the researchers examined sleep times across 96 different mammals. They were able to test various hypotheses that connected sleep to functions in the body and/or brain, ruling out several with a thoughtful examination of sleep times, animal size, and metabolic rates. Having honed in on a quantitative connection between sleep times and the brain’s metabolic rate, the pair put forth a theory that suggests sleep’s primary function is to repair and reorganize the brain. The theory explains why a mouse sleeps 14 hours a day, an elephant sleeps less than four, and a human being falls in the middle with eight. It also provides a basis for further research into The Child and Fortune, Leon Bouillon (19th C./French) the nature of REM sleep, sleep and aging, and sleep times during

12 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 13 research report Sleep to Repair

“What probing deep Has ever solved the mystery of sleep?” —Thomas Bailey Aldrich, Human Ignorance

12 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 13

drawing onultra-thinmicrographsofsequentialbrainslices. signal toitstarget.To capturethismoment,GrahamJohnsonbased Deep insidethebrain,aneuronpreparestotransmit Santa Fe InstituteBulletin

SPRING 2008

graham johnson medical media / www.fivth.com theory wouldpredict thattheamountofwak theory constant forallanimals.The thermoregulatory body mass,thentheratiobetween them remains heated). Because bothbearthesamerelation to brain) andbrainsize (theamountofbrainbeing would dependonmetabolism(whichheatsthe were heatingandcoolingrates true, the theory in different animals.Theyspeculatedthatif for thedrasticallydifferent sleeptimesobserved onthegrounds thatitdidn’ting theory account West elegantlydebunkedtheheatingandcool of potentialexplanationsforsleep. Savage and living advocates ofnote,theystillcrop upinlists theorieshave few and whilethermoregulatory propounded by Hippocrates around the brainfrom overheating. Theideawasfirst lent tothenumberofcaloriesinadinnerroll. wakefulnessisroughly equiva hours ofsedentary ergy saved from eighthoursofsleepversus eight 15 percent duringsleep, andforhumans,theen much energy. Themetabolismonlyslows 10to skepticism becausesleepingdoesn’t conserve the bodyandbrain.Thisideahasbeenmetwith they encountered holdsthatsleepgives rest to sleep literature. the investigation, andtheybegantoread upon inbiologicalscaling andhisadvisor,pert tojoin ing subject.” Savage askedGeoffrey West, anex into it,themore Irealized thatsleepisafascinat different. It was really weird, andthemore Igot made mefascinatedfrom theperspective thatit’s for otherbiologicaltimes,” Savage says.“So that was acompletelydifferent patternthan you see “Sleep else—it scalestheopposite ofeverything pattern: theyactuallydecrease forlargeranimals. the dataandfoundthatsleeptimesdidn’t fitthe increasing size. Just tobesure, hewent tocheck and lifespan,increase inlengthfororganismsof that mostbiologicaltimes,suchasgestation time scales for animals of different sizes. He knew about when Van Savage wasponderingbiological Another hypothesissaysthatsleepprotects One ofthe better-known sleep hypotheses The ideaforacross-species sleepstudycame

400 BC, ------It seemedreasonable, then,thatthebodywould unlike other cells, are not replaced when they die. likelier candidatethanthebody becauseneurons, of thebodyorbrain?The brainseemeda the needforrepair. effectofmetabolism, suchas by somesecondary sleep ananimalrequires mightwell bedictated etc.), theresearchers surmisedthattheamountof in reverse (pachyderms sleeplessthanprimates, than pachyderms. Andsincesleeptimesscale than primates,whichhave quickermetabolisms colder” ones.Rodents have quickermetabolisms generally live lifeatafasterpacethan“largerand cal scalingisthat“smaller andhotter” animals Savage, oneofthewell-known factsofbiologi at metabolicratesacross species.According to wrought by metabolism,Savage and West looked torepair thedamage whether sleepmightserve that agecellsanddamageDNA. To findout releases free radicals—theinfamousmolecules keeps alive. When acellproduces energy, italso metabolism, whichslowlycellsit harmsthevery effectof brain. Cellulardamageisasecondary function istorepair damageinthebodyand/or offtheirlist. sleep theory Savage and West crossed thethermoregulatory differences insleeptimes acrossobserved species, mals. Since theprediction flatlycontradictedthe would cool,remain constantforallmam the brain,andsleepingtime,inwhichbrain ing time,inwhichthemetabolismwouldheat But how totelliftherepair occursat thelevel Yet anothersleephypothesisholdsthatsleep’s Ah, withwhatsubtilemeaningdidtheGreek “O peacefulSleep!untilfrompainreleased Call theethelessermystery atthefeast Whereof thegreatermystery isdeath!” I breatheagain uninterrupted breath! — Henry Wadsworth Longfello

Santa Fe InstituteBulletin - -

SPRING 2008 w 15

drawing onultra-thinmicrographsofsequentialbrainslices. signal toitstarget.To capturethismoment,GrahamJohnsonbased Deep insidethebrain,aneuronpreparestotransmit Santa Fe InstituteBulletin

SPRING 2008

Santa Fe InstituteBulletin - -

SPRING 2008 w 15

quickly and have more sensory input A History of Sleep Research in a given period of time than do large and slow ones. Because processing this sensory input might require Ca 350 BC—Aristotle wrote that 1935—Alfred Loomis used the elec- rotransmitter. Jouvet also located the sleep, neural reorganization fits the sleep results from warm vapors rising troencephalogram to detect different area in the brainstem wherein neurons metabolic sleep theory just as well as from the digestive tract. patterns in the brain activity of sleeping regulate REM and NREM sleep stages. neural repair. In the end, Savage and persons, hypothesizing that the various West did not choose between repair Ca 400 BC—In the Corpus 1968—Allan Rechtschaffen and An- EEG patterns corresponded to various z and reorganization. They suspect that Hippocratum, Hippocrates hypothe- thony Kales published a standardized states of consciousness. sleep may serve a dual function. “A sized that sleep protects the brain from method for interpreting EEG recordings. z overheating. 1953—The “Father of American Sleep certain amount of your metabolic en- 1974—Harold Zepelin and ergy goes into repair, a certain amount Research,” Nathaniel Kleitman, and his 1907—French scientists Rene Legen- Rechtschaffen published sleep times in z goes into reorganization,” West says. student Eugene Aserinsky reported dre and Henri Pieron kept dogs awake different mammalian species, mostly “At the level at which the theory was in the journal Science that rapid eye z for several days and extracted a toxin observed in zoos and laboratories. developed, it does not distinguish the movement (REM) occurs during sleep. from their blood serum. When they 1979—An analysis of data from the proportion.” To accompany their EEG measure- injected the toxin of the sleep-deprived Yet another intriguing finding that American Cancer Society found that ments, the researchers devised an elec- dogs into healthy dogs, the recipients fell out of the research was that the people who slept less than four, or © naturepl.com tro-oculogram (EOG) to record changes were induced to sleep. The book that fraction of time spent in rapid eye mo- more than ten hours per night had in electrical activity, hence, in move- invest in a daily repair process for the precious ment (REM) sleep is independent of body mass. the researchers published in the wake shorter life expectancies.z The results Hibernating ment, around the eyes. cells, and that such a brain-level process might Elephants sleep less than mice, but devote the of the experiment laid out a scientific dormouse were confirmed with a follow-up study require unconsciousness. The researchers’ quanti- same percentage of their total sleep time to REM approach to the study of sleep, and is 1958—William C. Dement, another of (Muscardinus between 1982 and 1988. avellanarius) curled tative tip-off to the connection between sleep and sleep. “We argue that total sleep time is this ef- generally thought to be the foundation- Kleitman’s students, connected REM 1989—By experimenting on rats, up asleep in nest, the brain was that sleep times, which generally fect of metabolism that you have to respond to in al text of modern sleep research. sleep to dreaming and found that both Rechtschaffen and colleagues conclud- Sussex, UK. decrease as body size increases, tended to scale some way,” Savage says. “But then there’s REM cats and humans sleep in REM and to brain size rather than body size. A horse is sleep and it’s not clear if it’s repair and reorgani- 1929—Constantin Von Economo, a z ed that sleep deprivation leads to death non-rapid eye movement (NREM) stag- z doctor who treated encephalitis lethar- 10,000 times larger than a mouse, but its brain is zation or something else.” Savage is now looking in two to three weeks. Deprivation of es. Scientists would later discover that gica patients in the wake of the 1918 only 1,000 times larger. And while a horse sleeps for clues about the function of REM by studying either REM or NREM sleep killed the ro- almost all mammals have REM sleep. (It flu, observed that patients who suffered less than a mouse, its nightly sleep time is not sleep times in babies, who sleep more than adults dents over a slightly longer time span. was once thought that platypuses and brain damage in the hypothalamus 10,000 times shorter. Savage and West calculated and spend a greater proportion of their sleep Today there are over 200 sleep centers dolphins did not have REM sleep. The a mass-specific brain metabolic rate: a rate that, time in REM. and preoptic area, succumbed to long and laboratories in the United States. notion was contradicted for platypuses when coupled with an animal’s brain size, would West points out that the study could also pro- periods of sleep and unconsciousness. A mass of experimental evidence docu- in 1999, when a group of researchers predict the amount of time the animal devoted vide a framework for further research into the He posited that sleep is regulated by a ments the cognitive and physiological constructed an elaborate platypusarium to sleep. The rate they calculated fit with rough connection between sleep and aging, both of specific location in the brain. effects of sleep deprivation, and there and discovered that platypuses actually z empirical measurements of brain metabolic rates, which tie into metabolism. “People my age seem is a widespread effort to research sleep so the brain-level repair hypothesis was still in to sleep less and I don’t know why that is,” West 1929—Hans Berger, a German pro- have more REM sleep than any other disorders such as narcolepsy, insom- the running. says. “I’m sure there are people studying it, but fessor of neurology and psychiatry, mammals. It is now thought that dol- nia, and sleep apnea. Researchers have Though the researchers had traced sleep’s pri- I’m not sure anyone understands why that should recorded his son’s brain waves by at- phins may spend a very small fraction shown that sleep is essential for some mary function to the brain, they didn’t have evi- be. Now when you get really old, you do sleep taching wires to the boy’s scalp. He of their sleep time in REM.) dence that the process was repair per se. Another more like cats, for 15 hours a day. There are a published a report on the “Elektren- forms of learning and processing mem- 1967—Michel Jouvet advanced the of sleep’s functions could be processing memories million reasons people sleep for longer times and kephalogramm,” or “electroencephalo- zories, and they continue to explore the modern biochemistry of sleep with the of the waking day. This cortical reorganization there are all sorts of factors that are neglected, so gram” (EEG) in English. nature of dreams and cognition during observation that the sleep stages in might also be governed by metabolism, because this study forms a baseline for trying to discover sleep.—Jenna Beck small animals with fast metabolisms move more what those factors are.”—Jenna Beck t cats depend on the presence zof a neu- 16 Santa Fe Institute Bulletin SPRING 2008 z z Santa Fe Institute Bulletin SPRING 2008 17

patterns of

Aaron Clauset has spent nearly three years

modeling the statistics of terrorism, but holds little hope that a mathematical model can predict whether a given man will walk a bomb into a given café on a given after- noon. He does believe that in large enough social systems, the capricious behaviors of individuals seem to fade in the face of collective patterns. “A classic question that many historians have asked over the years is, ‘Where does individual control end and statistical behavior take over?’” Clauset says. A physicist and computer scientist by training, he is pursuing that question. His work to date has led him to conclude that terrorist at- lamy

tacks conform to patterns, at least on a global scale. In Febru- A ary 2007, Clauset, a Santa Fe Institute postdoc, and his part- atthews /

ners Maxwell Young (now a graduate student at the University M of Waterloo) and Skrede Gleditsch (a reader at the University of Essex) published a study in the Journal of Conflict Resolution that made a novel claim: the frequency of severe terrorist attacks, when taken worldwide, seems to follow a remarkably simple equation. The statistical distribution fits severe events like 9/11 to the same curve as more common but less severe stockphoto.com; debris: © Jenny i ones that kill a dozen or so people. The pattern suggests that fire: © such rare and large events are not outliers, as was previously thought, but are somehow interconnected with the smaller at- Qana, South Lebanon: A wedding The authors claim that if an underlying connection exists, then taking measures to tacks. The authors claim that if an underlying connection exists, dress shop destroyed by bombing, then taking measures to discourage small-scale attacks might August 2006. discourage small-scale attacks might also prevent severe ones. also prevent severe ones.

18 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 19 research report

patterns of

Aaron Clauset has spent nearly three years

tacks conform to patterns, at least on a global scale. In Febru- A ary 2007, Clauset, a Santa Fe Institute postdoc, and his part- atthews /

18 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 19 The group also found that the global pattern ics and computation at the University of New breaks down at the regional scale. The distribu- Mexico. The professor, SFI’s Cris Moore had tion that described the global terrorist attacks did taught a unit on power laws just as Young and not hold for terrorist attacks within individual Clauset’s mutual interest in analyzing conflict led countries, possibly because political quirks, cul- them to look into conflict databases. When they tural and historical factors, and individual caprice found the database of the Memorial Institute for play a greater role in smaller regions. Clauset’s the Prevention of Terrorism (MIPT), they were group would like to understand how the simple satisfied that the data was copious and consistent an underwood/informatoin illustrated j

global pattern arises from the complex local con- enough for statistical analysis. Clauset and Young Deaths per million people, 1998-2007 flicts, and to that end they are currently studying looked at the 28,445 terrorist events that had The power law pre- <0.1 a single region of conflict. occurred since 1968, restricting their analysis to vailed as the best fit, 0.1 - 1 1 - 10 The study of social phenomena has enjoyed a the 10,878 events that resulted in one or more with the frequency 10 - 100 100 - 400 quantitative renaissance in recent years thanks casualties. When they plotted the severity of of events scaling as an to an infusion of methods from condensed mat- events against the frequency, the average num- inverse power of their ter physics. In 1998, Duncan Watts and Steve ber of deaths per terrorist attack (about 5) was a severity (P(x)~x–2.48 ). Strogatz employed a small-world network, which poor indicator of the overall distribution, which The power law suggested that severe and small and states. The paper also calls for new models to This map illustrates in mathematics and physics is a kind of graph, included events that killed tens and hundreds of terrorist events are interconnected, and that convincingly connect the observed distribution the level of “risk” to formalize a social theory commonly associ- people. The data did not fit a bell curve, as would counter-terrorism strategies that succeed on the to knowledge of the factors that mitigate or con- of terrorist attacks small scale might succeed in preventing larger tribute to terrorism. around the globe. events as well. In that spirit, the group is now embarking on “Risk” is defined as “At a large scale, no single person or group can control very well where the The challenge that remains is figuring out why a more detailed study of what governs terrorist the total number of terrorism events should follow a power law at all. attacks in a single region. Using the MIPT data deaths over a 10- conflict goes, and the statistics, or if you like, the physics, start to take over.” Skrede Gleditsch, a political scientist and stat- and public opinion polls, they are examining the year period divided istician, helped Young and Clauset frame their ongoing Israeli-Palistinean conflict. “Instead of by the nation’s analysis for a political science audience. looking at the macro level,” Gleditsch says, “we current estimated ated with “six degrees of separation.” Physicist typically distributed events such as heights of “A power law per se is a statistical distribution, are looking at individual groups so we can try population. Alberto-Laszlo Barabasi further popularized the people or speeds of cars on a highway. The severe and knowing that something fits that is maybe for a more integrative model.” The hope is that physics approach to social phenomenon in 1999 events, like 9/11 and a 1998 car bombing in not terribly interesting by itself,” Gleditsch says. learning about the behavior of the complex play- when he claimed to have found a pattern in the Nairobi that injured over 5,000, were drastically “But it might give you some ideas about what ers in the regional conflicts will lead to a better distribution of hyperlinks in the World Wide far from the mean. They made the distribution kind of processes could generate the distribu- explanation of the global pattern. Web. The scientific community took notice, and “heavy-tailed”—and in physics, a heavy tail is the tion.” If severe terrorist events have the same Clauset believes that their success in creating a as more scientists began looking for similarities calling card of a power law. underlying distribution as minor terrorist events, model will depend on the number of constraints between complex atomic systems and complex Power laws tend to be exciting to physicists can they be explained by scaled-up versions of a human system must have before statistics start social systems, “physics of society” papers began because they are often the direct consequences the same underlying processes? taking over. He says: “In terrorism, if you look claiming more space in the top journals. Phase of other physical laws. They also imply “scale The authors acknowledge the possibility, but at a small enough scale—say, attacks in Bagh- changes, which are typically used to describe the invariance,” which suggests that the small-scale do not claim to know what those processes are. dad—at that level it’s extremely strategic, and behavior of gasses, fluids, and solids, have been and large-scale events on the curve are generated “While there are many instances where complex individual decisions have a large effect on the sta- applied to traffic patterns. Power laws, the -in by the same mechanism. Clauset and Young had behavior can be plausibly explained by a relative- tistics. But the idea of looking at global conflict verse-square equations that describe gravitation to “curb their excitement,” in Young’s words, out ly simple generating mechanism,” Young wrote or an entire conflict between two sides is that the and electromagnetic force, are also thought to of recognition that “there are an infinite number in an email, “I personally hold out little hope for details matter less and striking statistical patterns fit distributions of wealth, populations of cities, of heavy-tailed distributions out there.” In the that in this case.” As a sample explanation, the emerge. In other words, at a large scale, no single and, now, terrorism. spirit of caution, the pair developed some new paper put forth a mathematical toy model that person or group can control very well where the The inspiration for the terrorism project came statistical tools for accurately characterizing and showed the power law distribution as resulting conflict goes, and the statistics, or if you like, the after Maxwell Young took a course on phys- testing power law distributions in empirical data. from a game-like competition between terrorists physics, start to take over.”—Jenna Beck t

20 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 21 The group also found that the global pattern ics and computation at the University of New breaks down at the regional scale. The distribu- Mexico. The professor, SFI’s Cris Moore had tion that described the global terrorist attacks did taught a unit on power laws just as Young and not hold for terrorist attacks within individual Clauset’s mutual interest in analyzing conflict led countries, possibly because political quirks, cul- them to look into conflict databases. When they tural and historical factors, and individual caprice found the database of the Memorial Institute for play a greater role in smaller regions. Clauset’s the Prevention of Terrorism (MIPT), they were group would like to understand how the simple satisfied that the data was copious and consistent an underwood/informatoin illustrated j

20 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 21

profile

Would the laws hold in a real ecosystem, comparing animals that live in direct Pablo competition with one another? Would they hold in different ecosystems?

By Marquet: Julie J. from all kinds of different animals in far-flung regions of held up just as Brown claimed. “It came out beautifully,” the world. Would the laws hold in a real ecosystem, com- Marquet says, “just like the prediction.” Guerilla Ecologist Rehmeyer paring animals that live in direct competition with one Brash kid that he was, he submitted the paper to another? Would they hold in different ecosystems? Science. After that, he applied for a Fulbright Fellowship Marquet figured he knew just how to find out. A pro- to be a doctoral student with Jim Brown. fessor of his was working in two stretches of coastline in Marquet’s prediction about his own success was as Chile that were very similar, except that in one, humans accurate as his prediction about the abundance of popula- weren’t allowed to fish or hunt, and in the other, they tions. A week after he arrived at the University of New were. Marquet realized that these areas were the perfect Mexico, the paper appeared on the cover of Science. “It testing ground for Brown’s ideas. was the best presentation any new graduate student could As a community ecologist, Marquet focused particu- ever have,” he says with a smile. larly on Brown’s claim about population density. The theory was almost like a magic trick: give Brown any The Biotic Bang SFI International Fellow ecosystem, anywhere in the world, and tell him a par- It wasn’t long before Marquet was driving through the Pablo Marquet ticular weight, say 10 grams. Then, Brown said, he could piñon and juniper highlands between Albuquerque and tell you the number of creatures you’d find per acre that Santa Fe to attend talks at the Santa Fe Institute. That Nature had called Pablo Marquet years ago was just 21 years old—a college kid, after all—and his weigh 10 grams. was the beginning of a life-long association with the by inspiring him to become an ecologist, but now it was English wasn’t very good. But now, who should step into So Marquet wanted to test Brown’s claim in his two place. Marquet has visited periodically ever since, recently calling him in a different way. He rushed through the halls the restroom but Jim Brown himself. A thought flashed Chilean ecosystems. As similar as the two regions were, as an International Fellow. of his crowded university building in downtown Santiago, through Marquet’s mind: “Look! I’m here, with this great they were a hard challenge for Brown’s theory. In the area driven less by his current need than by his excitement. A scientist!” Their eyes met, but Marquet couldn’t seem to with hunting, people did all kinds of things that changed scientist had just given a lecture that satisfied the frustra- untie his tongue. which animals prospered and which suffered. For ex- tion Marquet had been struggling with through college. Ideas were already forming in Marquet’s mind, how- ample, people collected giant sea snails to eat, so the area To Marquet’s eye, ecology was filled with all these little ever. He was formulating an experiment that would make with hunting had fewer snails than the undisturbed area. studies in little areas giving little results, with no grand Brown’s theory even stronger. And that didn’t just affect the snails. The mussels pros- theories tying any of them together. But Jim Brown, pered when the snails declined, because snails loved to Distinguished Professor at the University of New Mexico, Snails Versus Mussels: Predicting Populations eat mussels. The differences rippled throughout the entire nc I , had just shown how he was turning ecology into a quan- What Brown had done was to explore the ecological community of animals. s titative, predictive science, like physics or chemistry. consequences of a set of beautifully simple mathematical If Brown’s laws were really a universal property of living earcher s e

Marquet stepped into the restroom and looked out formulas that relate body mass to a remarkable number of things, then when the snails became rarer, some other R through the smoggy haze at the Andes towering above. He biological attributes: metabolic energy, life span, popula- creature with the same body size had to prosper. It might hoto had been even more amazed by what had happened after tion density, and more. He had also begun to explain why be something completely unrelated, like a little bird or P ner / Brown finished speaking. The great scientists assembled these formulas, called scaling laws, were popping up over a rodent, but the total number of creatures with that ss chmei in that room had attacked Brown’s theory as a meaning- and over again throughout biology. body size had to be the same. Similarly, when the mussels s less dead end. But Brown had fought back! He’d defended The scientists at the conference had laid into Brown’s prospered, some other small creature—perhaps a barnacle Steve G himself against the combined opposition of the greatest theories with a number of criticisms, and Marquet or even an insect—had to suffer. Marquet wanted to find ecologists in North and South America combined. thought one of them had some legitimacy. Brown and out if that was true. In his quest to uncover some of the secrets of how biodiversity develops,

Marquet had listened to the raging debate quietly. He others had found the scaling laws by pulling together data top: lauraware When Marquet crunched the numbers, the scaling laws Marquet uses bacteria, such as these rod-shaped E. coli bacilli.

22 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 23

profile

Would the laws hold in a real ecosystem, comparing animals that live in direct Pablo competition with one another? Would they hold in different ecosystems?

22 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 23

The International Program at the Institute was begun you forget, the bacteria will eat up all the food and then in 2000 to encourage multidisciplinary collaboration the population will crash. in countries in which little funding is available for such If you are really neglectful and never pour in more me- work. The program funds several researchers as Interna- dium, a surprising thing happens: a few bacteria continue tional Fellows each year, providing two years of financial to live. Despite the lack of food, the population continues support including up to two months in residence at the at a very low level. After a while, the population will even Institute, developing collaborations with SFI researchers. begin to grow again. But soon thereafter, it experiences The Institute also supports fellows in their home coun- another crash. This boom and crash cycle continues for tries by providing funds for them to organize workshops a while, until it somehow stabilizes into a continuing or host visitors. population of bacteria. Marquet’s presence at the Institute has provided just the Marquet explains that what happens is that after the earch.com kind of cross-fertilization of ideas and approaches that the bacteria consume the initial medium and mostly die off, s / foto program intends. “Pablo asks questions and approaches a mutant evolves with a remarkable innovation: it is able s things in a way that I don’t think any young scientist who to consume the waste products of the original bacteria grew up and was trained only in the U.S. or in Britain as food. “They’re eating the dead bodies of their fellows,” would have,” Brown says. “He is by far the most philo- Marquet says. © brand x picture sophical of all my former students. His work on ecology The new mutant thrives until it is outcompeted by a is integrated into a worldview that includes the broader new mutant better able to eat the waste from the previous La Portada, Antofagasta, Chile: When Marquet applied Jim Brown’s theory to a coastal region in Chile, the laws held up just as Brown claimed. science of complexity and a sense of what you might call mutant, and it repeats the cycle. The population stabilizes “It came out beautifully,” Marquet says, “just like the prediction.” human ecology, our place in nature.” when mutants coexist and a whole community of bacteria Marquet has used that philosophical outlook to tackle evolves. Marquet, Keymer, and Fuentes are developing theory-building, he also tackles an enormous variety of on the continent, Chilean scientists themselves develop perhaps the grandest question in ecology: why life first a mathematical model to understand this process more practical ecological problems. He is developing models to unusual patterns of thought. began to diversify. “The physicists try to understand the thoroughly, with hopes that the process may illustrate understand how climate change is likely to affect Chilean Marquet grew up during the years of Augusto Pinochet’s Big Bang,” Marquet says. “Ecologists need a theory to secrets behind the initial development of biodiversity. ecosystems. He is collaborating with archaeologists to reign, and he says that in many ways, Pinochet’s dictator- understand the Biotic Bang.” Marquet has undertaken this line of research while at understand why a hunter-gatherer society in Chile that ship had a very bad effect on science. It limited funding With SFI support, Marquet has organized workshops the Santa Fe Institute, and he sees an analogy in the work lived 8,000 years ago had enough energy and resources for science and created an atmosphere of fear in which in Chile, Santa Fe, and Santa Barbara that have gath- for the fruitfulness of the Institute. “At SFI, there are a lot to make elaborate mummies of their dead. He is also people were afraid to speak their minds. That was part of ered ecologists together to puzzle out an approach to the of mutants, so you can create innovation just like the bac- studying the patterns plants form in the Chilean Atacama why he was so astonished by Brown’s bold defense at the problem. He also worked with another SFI International teria,” he says. “Eventually, we can create a sort of food Desert, which is the driest desert on earth. conference in 1986. Fellow, David Storch of the Center for Theoretical Study web where my throw-away insights can be your resources Marquet has now returned to Chile, and he is commit- “The process of maturation in science is to find your in the Czech Republic, to organize a workshop in Prague. so we can produce something together. At the end of the ted to continuing both his practical and theoretical work own voice, finding yourself saying something that you be- Marquet, Storch, and Brown edited a book from the con- day, we are like a bacterial biofilm.” there. He has been tempted to move to the United States, lieve in,” he says. “That process is a combination of art and ference, entitled Scaling Biodiversity, which has just come Part of Marquet’s unique contribution to the SFI com- where he has more colleagues interested in the speculative science. It’s an act of creation, and I think that a dictator- out this year. munity, Brown says, comes from his breadth of interests work that he loves. But he notes: “I have a huge impact ship is not the best environment for creativity.” and ability to make surprising connections. “He reads in Chile in terms of students, connecting people to my But he notes that the dictatorship also bred an attitude One Bacterium’s Waste is Another’s Food a lot of literature, really philosophical literature. He’ll network of colleagues, changing how the system works, in the best Chilean scientists that he respects and embrac- In his own research, Marquet is working with his former sometimes say something that will change the whole bringing new perspectives, and making it easier for people es. “It’s guerilla science, with 100 percent passion. I really student Juan Keymer and SFI postdoc Miguel Fuentes tenor of conversation and send it in a new, more creative thinking in more theoretical ways. I would probably be love that spirit, feeling that what moved you to do this is to understand a simple experiment. In the process, they direction. Often it is by analogy, similar to the way that just one more scientist in the States or in Europe.” such a strong force that you will overcome any obstacle.” t hope to uncover some of the secrets of how biodiversity Isabel Allende uses alliteration in her novels.” He also recognizes the stamp of Chile in his own ways developed. Put a single species of bacterium into a flask of working. Chile, he says, is like an island, removed from Julie J. Rehmeyer was SFI’s very first undergraduate intern. in some medium that provides food and the bacteria will Science within a Dictatorship the rest of the world on one side by the Andes and on the She went on to do graduate work in mathematics at MIT reproduce madly. If you keep supplying medium for the Marquet’s scientific projects reflect that breadth of inter- other by the Pacific. Just as creatures on an island tend to and to teach at St. John’s College. She is now a freelance bacteria to eat, they will continue to prosper, but should est. At the same time as he works on his speculative evolve into unique forms that are different from creatures writer and the mathematics columnist for Science News.

24 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 25

24 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 25 Malware By Devon Jackson Wars Vulnerabilities versus Romanian), and an emerging market patches. Robustness versus phish- economy. Malware’s effects reach far ing. Botnets versus evolvability. And beyond computers, and some fear complexity versus spam. Some of that it threatens to drive the Internet, these terms and concepts are central as most people know it, to extinc- to the work and philosophy of the tion. That’s why the Santa Fe Institute Santa Fe Institute. Some, though, again agreed to host, for the second are more specific to the world of the straight year, a workshop on how to Internet—as complex a system as any deal with this potentially disastrous organic one. phenomenon. The predators and parasites of this This year’s workshop, like the system are known as malware. Usu- first, was organized by Matthew ally defined as software designed Williamson, a principal research sci- to infiltrate or damage a computer entist at Sana Security, and Eric Davis, system without the owner’s informed a senior policy specialist at Google. consent, malware (a combination of The two-day event, entitled “Fight- “malicious” and “software”) ranges ing Modern Malware II,” included from computer worms, viruses, and participants from academia, private identity theft to spam, spyware, and corporations, and the government.

adware—as well as botnets, distribut- And beyond its economic and so- © celia johnson / rappart.com ed denial-of-service attacks, phishing, cial impact, malware is the perfect lab pharming, and zombies. It’s a multi- rat for anyone interested in complex- billion-dollar criminal industry, with ity, interconnectedness, and evolu- its own language (called Eblish—an tion. It’s real, and it’s global. As SFI amalgamation of English, text mes- Vice President Chris Wood observed, saging-speak, email-speak and what- “Malware raises issues of evolvability, ever language happens to be native to robustness, and diversity. It is compu- the user, say, Nigerian, Mandarin or tation in the wild.”

26 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 27 Malware By Devon Jackson Wars Vulnerabilities versus Romanian), and an emerging market patches. Robustness versus phish- economy. Malware’s effects reach far ing. Botnets versus evolvability. And beyond computers, and some fear complexity versus spam. Some of that it threatens to drive the Internet, these terms and concepts are central as most people know it, to extinc- to the work and philosophy of the tion. That’s why the Santa Fe Institute Santa Fe Institute. Some, though, again agreed to host, for the second are more specific to the world of the straight year, a workshop on how to Internet—as complex a system as any deal with this potentially disastrous organic one. phenomenon. The predators and parasites of this This year’s workshop, like the system are known as malware. Usu- first, was organized by Matthew ally defined as software designed Williamson, a principal research sci- to infiltrate or damage a computer entist at Sana Security, and Eric Davis, system without the owner’s informed a senior policy specialist at Google. consent, malware (a combination of The two-day event, entitled “Fight- “malicious” and “software”) ranges ing Modern Malware II,” included from computer worms, viruses, and participants from academia, private identity theft to spam, spyware, and corporations, and the government.

26 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 27 In this fictional scenario, an attacker (green) hires bot herders (blue) to give instructions to zombies— According to a recent survey, “The getting into malware. “We’re dealing Given that 80 percent of all mal- computers infected by 2007 Malware Report: The Economic with a criminal enterprise equivalent ware attacks originate outside the their malware (white), Impact of Viruses, Spyware, Adware, to the Mafia, and one that has no U.S., American authorities often find which hit designated targets. Botnets, and other Malicious Code,” leader,” he added. themselves in dire need of coopera- conducted by Computer Economics, Malware’s engineers are moving tion from authorities in the country a monthly information technology at warp speed. But the guys in the from which the malware was initiat- newsletter, the worldwide economic white hats, as the anti-malware folks ed. And the situation would be easier impact of malware is in decline: sometimes call themselves (or, alter- if other countries had similar laws to dropping from a high of $17.5 bil- natively, the Jedi Council), have had those in the U.S., or any laws at all. lion in damages in 2004 to $14.2 their hands tied. “These bad guys ex- Many, however, have yet to even out- billion in 2005 to $13.3 billion in ist in a world without boundaries,” law malware, much less understand it. 2006. But Computer Economics’ edi- said former Secret Service agent Rob- Still, as difficult as it may be to prove tors cautioned against over-optimism: ert Rodriguez, a sentiment echoed that, say, someone in Latvia sent out “Although direct damages of malware by Google’s senior staff engineer, a virus, it’s Cox’s belief that the laws may be declining,” they wrote, “the currently in place are, for the most indirect or secondary damages are part, adequate. “We have the laws we Malware is the perfect lab likely increasing.” As pointed out by need,” says Cox. “What’s lacking is Stefan Savage, associate professor of rat for anyone interested attribution, number one, and getting computer science and engineering at in complexity, intercon- data from other countries. Beyond the University of California at San that, we also need more reporting of nectedness, and evolution. Diego and the director of the school’s Internet crimes—both from our own Collaborative Center for Internet It’s real, and it’s global. As private businesses and from our gov- Epidemiology and Defenses (a joint SFI Vice President Chris ernment agencies.” effort between UCSD and the Inter- Wood observed, “Malware And, as if to add emphasis to his important,” said Savage. “In malware, lions of compromised hosts. The we introduce them into the transac- national Computer Science Institute), point about the adequacy of exist- the problems change so quickly. Last bad guys can scale up faster because tion cost.” The goal is to put a drag raises issues of evolv- magazine “We really don’t know precisely how ing laws, Cox recently reported to year’s problems are not this year’s they have no laws, no rules, and no on the bad guys without imposing big this problem is, but we know it’s ablity, robustness, and the group the arrest of Alan Ralsky, wired problems.” The moving target of boundaries. The good guys—banks an equal amount of drag on the good large and growing.” diversity. It is computation the self-proclaimed “King of Spam.” malware makes it hard to figure out especially, agreed the Jedis—remain guys. Otherwise, the future looks Hence, a prevailing sense of urgen- Cox emphasized, “As I stated at the where to invest one’s efforts—viruses loath to share information with each rather Orwellian. “My fear is that in the wild.” cy (if not impending doom) among conference, the criminal justice pro- sent over email are not the problem other, with law enforcement, with in the effort to secure the industry, the workshop’s 18 participants. “My cess is not the first line of response in they once were. Instead, attacks sent their customers, or the public. Com- we’ll see people’s rights trampled,” fear is that the horses are already out Niels Provos, an expert on honey- addressing malware, but to the extent above : catalogtree.net for via browsers and malicious web pages petition frustrates cooperation, and said Vincent Weafer, a member of of the barn and it’ll be impossible to pots—computer systems set up as that malware is a form of computer are a growing threat. intellectual property laws, too, serve Symantec’s Security Response Team. get them back in,” said Howard Cox, traps for attackers. “The development crime, this case demonstrates that Which is why it’s just as important to suppress anti-malware innovations. “And that there’ll be these country- the Department of Justice’s Computer and acceleration of malware in China even kings are not out of reach.” to determine how to invest one’s ef- And the problem with most anti- wide firewalls enacted under the guise Crime Division assistant deputy chief. and elsewhere has gotten to the point fort. Given that exploiting a weakness malware innovations is that they tend of security.” “We’re in the third generation of where we can’t keep up,” he said. The Jedi Council Versus the is usually easier than patching it up, to come at the expense of overall ease: Whatever technological solutions this. The juveniles of yesterday—the “We’re just not adapting quickly Malwarts no wonder the containment, let alone one more lock on the door may slow arise out of the workshop, most of its hackers and the Defcon wan- enough,” added Pittsburgh-based So, in a world where everyone’s defeat, of malware seems a Sisyphean down an intruder but it also slows participants more or less agree that nabes—have turned into adults and FBI agent Michael McKeown, who, vulnerable but no one is account- task. “Our patches to fix up holes is down the owner when getting in or economics drives malware. Take the have figured out, ‘We can now make like Cox, sees a trend toward orga- able, who should be held liable? The like the whack-a-mole game, only in out—and that newfangled new lock incentives away, devalue them, or re- money at this thing.’” He noted that nized crime and a global network. browser? The user? The Internet technological terms,” said Savage. won’t come free, either. “If we intro- direct malware engineers toward be- the economic incentive makes it al- “We are doomed to be reactive,” service provider (ISP)? “The decision It’s a problem of scale, as well as duce frictions (anti-malware actions) neficent incentives, and the Internet most impossible to stop people from warned Savage. on where you invest your effort is speed—one software bug equals mil- into the equation,” said Savage, “then may survive. “My hope is that we can

28 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 29 In this fictional scenario, an attacker (green) hires bot herders (blue) to give instructions to zombies— According to a recent survey, “The getting into malware. “We’re dealing Given that 80 percent of all mal- computers infected by 2007 Malware Report: The Economic with a criminal enterprise equivalent ware attacks originate outside the their malware (white), Impact of Viruses, Spyware, Adware, to the Mafia, and one that has no U.S., American authorities often find which hit designated targets. Botnets, and other Malicious Code,” leader,” he added. themselves in dire need of coopera- conducted by Computer Economics, Malware’s engineers are moving tion from authorities in the country a monthly information technology at warp speed. But the guys in the from which the malware was initiat- newsletter, the worldwide economic white hats, as the anti-malware folks ed. And the situation would be easier impact of malware is in decline: sometimes call themselves (or, alter- if other countries had similar laws to dropping from a high of $17.5 bil- natively, the Jedi Council), have had those in the U.S., or any laws at all. lion in damages in 2004 to $14.2 their hands tied. “These bad guys ex- Many, however, have yet to even out- billion in 2005 to $13.3 billion in ist in a world without boundaries,” law malware, much less understand it. 2006. But Computer Economics’ edi- said former Secret Service agent Rob- Still, as difficult as it may be to prove tors cautioned against over-optimism: ert Rodriguez, a sentiment echoed that, say, someone in Latvia sent out “Although direct damages of malware by Google’s senior staff engineer, a virus, it’s Cox’s belief that the laws may be declining,” they wrote, “the currently in place are, for the most indirect or secondary damages are part, adequate. “We have the laws we Malware is the perfect lab likely increasing.” As pointed out by need,” says Cox. “What’s lacking is Stefan Savage, associate professor of rat for anyone interested attribution, number one, and getting computer science and engineering at in complexity, intercon- data from other countries. Beyond the University of California at San that, we also need more reporting of nectedness, and evolution. Diego and the director of the school’s Internet crimes—both from our own Collaborative Center for Internet It’s real, and it’s global. As private businesses and from our gov- Epidemiology and Defenses (a joint SFI Vice President Chris ernment agencies.” effort between UCSD and the Inter- Wood observed, “Malware And, as if to add emphasis to his important,” said Savage. “In malware, lions of compromised hosts. The we introduce them into the transac- national Computer Science Institute), point about the adequacy of exist- the problems change so quickly. Last bad guys can scale up faster because tion cost.” The goal is to put a drag raises issues of evolv- magazine “We really don’t know precisely how ing laws, Cox recently reported to year’s problems are not this year’s they have no laws, no rules, and no on the bad guys without imposing big this problem is, but we know it’s ablity, robustness, and the group the arrest of Alan Ralsky, wired problems.” The moving target of boundaries. The good guys—banks an equal amount of drag on the good large and growing.” diversity. It is computation the self-proclaimed “King of Spam.” malware makes it hard to figure out especially, agreed the Jedis—remain guys. Otherwise, the future looks Hence, a prevailing sense of urgen- Cox emphasized, “As I stated at the where to invest one’s efforts—viruses loath to share information with each rather Orwellian. “My fear is that in the wild.” cy (if not impending doom) among conference, the criminal justice pro- sent over email are not the problem other, with law enforcement, with in the effort to secure the industry, the workshop’s 18 participants. “My cess is not the first line of response in they once were. Instead, attacks sent their customers, or the public. Com- we’ll see people’s rights trampled,” fear is that the horses are already out Niels Provos, an expert on honey- addressing malware, but to the extent above : catalogtree.net for via browsers and malicious web pages petition frustrates cooperation, and said Vincent Weafer, a member of of the barn and it’ll be impossible to pots—computer systems set up as that malware is a form of computer are a growing threat. intellectual property laws, too, serve Symantec’s Security Response Team. get them back in,” said Howard Cox, traps for attackers. “The development crime, this case demonstrates that Which is why it’s just as important to suppress anti-malware innovations. “And that there’ll be these country- the Department of Justice’s Computer and acceleration of malware in China even kings are not out of reach.” to determine how to invest one’s ef- And the problem with most anti- wide firewalls enacted under the guise Crime Division assistant deputy chief. and elsewhere has gotten to the point fort. Given that exploiting a weakness malware innovations is that they tend of security.” “We’re in the third generation of where we can’t keep up,” he said. The Jedi Council Versus the is usually easier than patching it up, to come at the expense of overall ease: Whatever technological solutions this. The juveniles of yesterday—the “We’re just not adapting quickly Malwarts no wonder the containment, let alone one more lock on the door may slow arise out of the workshop, most of its hackers and the Defcon wan- enough,” added Pittsburgh-based So, in a world where everyone’s defeat, of malware seems a Sisyphean down an intruder but it also slows participants more or less agree that nabes—have turned into adults and FBI agent Michael McKeown, who, vulnerable but no one is account- task. “Our patches to fix up holes is down the owner when getting in or economics drives malware. Take the have figured out, ‘We can now make like Cox, sees a trend toward orga- able, who should be held liable? The like the whack-a-mole game, only in out—and that newfangled new lock incentives away, devalue them, or re- money at this thing.’” He noted that nized crime and a global network. browser? The user? The Internet technological terms,” said Savage. won’t come free, either. “If we intro- direct malware engineers toward be- the economic incentive makes it al- “We are doomed to be reactive,” service provider (ISP)? “The decision It’s a problem of scale, as well as duce frictions (anti-malware actions) neficent incentives, and the Internet most impossible to stop people from warned Savage. on where you invest your effort is speed—one software bug equals mil- into the equation,” said Savage, “then may survive. “My hope is that we can

28 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 29 find economic disincentives so that plained, exists when one party to a from an observer’s perspective than friction. Do they work? Will they? crime levels itself out? If you can iso- “What we’re trying to do,” said people don’t use malware or invest in transaction has more information an economist’s. He presented some Or does it just boil down to so much late these upper-tier forces—not the Gleichauf, “we’re trying to maintain it,” said Ari Schwartz, who works on than another party. Principles of in- of the mechanics of malware’s under- tilting at windmills? Or might it all ankle-biters—will they regulate them- the functionality of old systems.” privacy and government-information formation asymmetry can be used to ground economy. Malware has gone simply be part of some grander digital selves?” And later still, he speculated, Old, antiquated, not as robust as issues for The Center for Democracy force the more knowledgeable party from being a reputation economy— design that’s still evolving? “If we stop trying to improve the they need be. And doomed, perhaps, and Technology, a nonprofit advocacy to disclose what they know. This in which people hacked for kudos— Enter Lee Altenberg, an associate Internet, our products and all things though no one has yet given up. If group that helped various anti-spy- method can help reveal someone’s to a complex, stratified profit econ- professor in the Department of In- related to it, maybe they’ll stop.” anything, meetings such as this, and ware companies form a coalition. true intentions, for example, when omy, in which people are innovating formation and Computer Sciences at other SFI workshops, which encour- “We need to attack their revenue they want a party to read their mes- all the time. “They even try to phish the University of Hawaii who teaches Robustness to the Rescue? age collaboration and cross-pollina- streams,” declared Marshall Van Al- sage or install their applet. Two-sided each other and ruin each other’s repu- exclusively online and specializes in Malware is parasitic on the software, tion, infuse participants with a re- styne, an associate professor at Boston networks, he explained, are matching tations,” said Lance James, author of evolutionary theory and population hardware, systems, and users of the newed sense of purpose. In this case, University and visiting professor at markets with two distinct user groups Phishing Exposed and an expert on genetics. He thinks that programs be- Internet. How, then, have hosts the group will come up with a set of MIT who specializes in the econom- who provide each other with benefits. phishing and malware who heads up have enough like organisms that some fought off parasites in the past? In anti-malware action points. They will ics of information. An avowed skeptic Common examples include cardhold- Security Science Corporation’s Exter- lessons from nature might be appli- nature? In other economic systems? hold regular meetings, both real and ers and merchants on credit card plat- nal Threat Assessment Team. (Phish- cable to the Internet and malware. What’s the proper co-evolutionary virtual—that will include representa- forms, and doctors and patients on ing uses social engineering tactics as Altenberg gently coaxed the complex- response, if any? Hope that the host tives from the banking industry, the HMO platforms. Van a way to obtain access to user names, ity cat out of the malware bag. is robust enough, implied Gleichauf, insurance industry, U.S. CERT (the Alstyne, who helped passwords, identity information, A major contributor to the com- Savage, and others, to withstand a Computer Emergency develop theories of two- credit cards, and other personal and plexity discussion was Robert Glei- parasite as nasty and evolvable as Readiness Team), sided networks, pointed corporate data; it also relies heavily on chauf, a former Ph.D. candidate in malware. Gleichauf, for one, wanted Amazon, eBay, out that they often have botnets, software robots that run au- anthropology who is now the Chief more discussion on robustness. It’s Yahoo and/or sophisticated fraud- tonomously and automatically, usu- Technology Officer for Cisco’s Secu- his belief that email systems are ro- Earthlink, detection techniques ally on groups of zombie computers rity Technology Group. A realist as bust enough to survive. “But that’s and Microsoft. (similar to those used controlled remotely.) well as a ponderer, he is anything but not so in banking,” he said, “which is Beyond that, blasé about the effects and potential founded on trust, which doesn’t work they also want How have hosts fought off parasites in the past? of malware; nor does he see it as a electronically.” to establish a necessary evil, or think that the death Or do we latch on to a punctuating malware research that technology Hope that the host is robust enough, implied of the Internet is imminent. He does, event—some major shift or develop- institute. alone can combat Gleichauf, Savage, and others, to withstand a however, believe in leveling mecha- ment—mused Gleichauf, to escape “My hope,” said malware, Van Alstyne parasite as nasty and evolvable as malware. nisms, and keeps an eye out for events the parasites? “Is there a punctuated Davis, “is that there arrived with a set of very that can lead to large evolutionary event about to happen?” he asked. are livable boundaries. nontechological solutions. swings. “Yes, we need to minimize “Right now, that’s what we’re looking That there will be a malware crime In a quick overview of his paper, “An to catch credit card and insurance So, proposed Savage and James, the the impact of perturbations,” says for at Cisco: the browser versus the rate, that’s pretty much unavoidable. Economic Response to Unsolicited fraud), that can be applied to fight solution may be to attack the malware Gleichauf, “but, after all, there’s a life mobile market,” referring to the cur- But people will know what to do and Communication” (i.e., spam—which malware. Lastly, externalities may be market, as well as its still intact repu- cycle of information. Things tend to rent merging of computer technology what not to do.” That’s the hope costs about $50 billion a year in useful. An externality is an impact tation-based system. They also advo- find their stasis point.” with technology that’s mobile, such as anyway. t losses and makes up 92 percent of (positive or negative) on any party not cated attacking the malware industry “The question is,” he continued, phones and iPods. all email), Van Alstyne outlined involved in a given transaction. Van economically, disrupting its efforts to “when you factor in the losses against Eventually, the complexity-fu- Devon Jackson is a freelance writer a very cogent and rather elegant Alstyne showed how liability laws that launder its profits. the total amount of money going eled debate came back to the threat based in Santa Fe. He writes regularly three-pronged attack, based on the are currently in place for other types across the Internet, at what point is itself. “The rate of evolution for bad for Southwest Art, and has written for economic principles of information of cases might be applied with success Is Complexity Science the Solution? the pain of losses so high that you guys is so much higher,” said Savage. Smithsonian, Outside, The New York asymmetry, two-sided networks, and toward malware problems. Economic solutions. Technological take action?” He asked later on, dur- “Malware has such high evolvability, Times and many other magazines and externalities. Savage, too, dealt with the eco- solutions. Legal factors, industry facing one of the workshop’s tactical it may evolve to the point where the newspapers. He is also the author of Information asymmetry, he ex- nomics of malware, although more tors. Phishing, patching, spamming, sessions: “How much of malware Internet is no longer useable.” Conspiranoia!

30 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 31 find economic disincentives so that plained, exists when one party to a from an observer’s perspective than friction. Do they work? Will they? crime levels itself out? If you can iso- “What we’re trying to do,” said people don’t use malware or invest in transaction has more information an economist’s. He presented some Or does it just boil down to so much late these upper-tier forces—not the Gleichauf, “we’re trying to maintain it,” said Ari Schwartz, who works on than another party. Principles of in- of the mechanics of malware’s under- tilting at windmills? Or might it all ankle-biters—will they regulate them- the functionality of old systems.” privacy and government-information formation asymmetry can be used to ground economy. Malware has gone simply be part of some grander digital selves?” And later still, he speculated, Old, antiquated, not as robust as issues for The Center for Democracy force the more knowledgeable party from being a reputation economy— design that’s still evolving? “If we stop trying to improve the they need be. And doomed, perhaps, and Technology, a nonprofit advocacy to disclose what they know. This in which people hacked for kudos— Enter Lee Altenberg, an associate Internet, our products and all things though no one has yet given up. If group that helped various anti-spy- method can help reveal someone’s to a complex, stratified profit econ- professor in the Department of In- related to it, maybe they’ll stop.” anything, meetings such as this, and ware companies form a coalition. true intentions, for example, when omy, in which people are innovating formation and Computer Sciences at other SFI workshops, which encour- “We need to attack their revenue they want a party to read their mes- all the time. “They even try to phish the University of Hawaii who teaches Robustness to the Rescue? age collaboration and cross-pollina- streams,” declared Marshall Van Al- sage or install their applet. Two-sided each other and ruin each other’s repu- exclusively online and specializes in Malware is parasitic on the software, tion, infuse participants with a re- styne, an associate professor at Boston networks, he explained, are matching tations,” said Lance James, author of evolutionary theory and population hardware, systems, and users of the newed sense of purpose. In this case, University and visiting professor at markets with two distinct user groups Phishing Exposed and an expert on genetics. He thinks that programs be- Internet. How, then, have hosts the group will come up with a set of MIT who specializes in the econom- who provide each other with benefits. phishing and malware who heads up have enough like organisms that some fought off parasites in the past? In anti-malware action points. They will ics of information. An avowed skeptic Common examples include cardhold- Security Science Corporation’s Exter- lessons from nature might be appli- nature? In other economic systems? hold regular meetings, both real and ers and merchants on credit card plat- nal Threat Assessment Team. (Phish- cable to the Internet and malware. What’s the proper co-evolutionary virtual—that will include representa- forms, and doctors and patients on ing uses social engineering tactics as Altenberg gently coaxed the complex- response, if any? Hope that the host tives from the banking industry, the HMO platforms. Van a way to obtain access to user names, ity cat out of the malware bag. is robust enough, implied Gleichauf, insurance industry, U.S. CERT (the Alstyne, who helped passwords, identity information, A major contributor to the com- Savage, and others, to withstand a Computer Emergency develop theories of two- credit cards, and other personal and plexity discussion was Robert Glei- parasite as nasty and evolvable as Readiness Team), sided networks, pointed corporate data; it also relies heavily on chauf, a former Ph.D. candidate in malware. Gleichauf, for one, wanted Amazon, eBay, out that they often have botnets, software robots that run au- anthropology who is now the Chief more discussion on robustness. It’s Yahoo and/or sophisticated fraud- tonomously and automatically, usu- Technology Officer for Cisco’s Secu- his belief that email systems are ro- Earthlink, detection techniques ally on groups of zombie computers rity Technology Group. A realist as bust enough to survive. “But that’s and Microsoft. (similar to those used controlled remotely.) well as a ponderer, he is anything but not so in banking,” he said, “which is Beyond that, blasé about the effects and potential founded on trust, which doesn’t work they also want How have hosts fought off parasites in the past? of malware; nor does he see it as a electronically.” to establish a necessary evil, or think that the death Or do we latch on to a punctuating malware research that technology Hope that the host is robust enough, implied of the Internet is imminent. He does, event—some major shift or develop- institute. alone can combat Gleichauf, Savage, and others, to withstand a however, believe in leveling mecha- ment—mused Gleichauf, to escape “My hope,” said malware, Van Alstyne parasite as nasty and evolvable as malware. nisms, and keeps an eye out for events the parasites? “Is there a punctuated Davis, “is that there arrived with a set of very that can lead to large evolutionary event about to happen?” he asked. are livable boundaries. nontechological solutions. swings. “Yes, we need to minimize “Right now, that’s what we’re looking That there will be a malware crime In a quick overview of his paper, “An to catch credit card and insurance So, proposed Savage and James, the the impact of perturbations,” says for at Cisco: the browser versus the rate, that’s pretty much unavoidable. Economic Response to Unsolicited fraud), that can be applied to fight solution may be to attack the malware Gleichauf, “but, after all, there’s a life mobile market,” referring to the cur- But people will know what to do and Communication” (i.e., spam—which malware. Lastly, externalities may be market, as well as its still intact repu- cycle of information. Things tend to rent merging of computer technology what not to do.” That’s the hope costs about $50 billion a year in useful. An externality is an impact tation-based system. They also advo- find their stasis point.” with technology that’s mobile, such as anyway. t losses and makes up 92 percent of (positive or negative) on any party not cated attacking the malware industry “The question is,” he continued, phones and iPods. all email), Van Alstyne outlined involved in a given transaction. Van economically, disrupting its efforts to “when you factor in the losses against Eventually, the complexity-fu- Devon Jackson is a freelance writer a very cogent and rather elegant Alstyne showed how liability laws that launder its profits. the total amount of money going eled debate came back to the threat based in Santa Fe. He writes regularly three-pronged attack, based on the are currently in place for other types across the Internet, at what point is itself. “The rate of evolution for bad for Southwest Art, and has written for economic principles of information of cases might be applied with success Is Complexity Science the Solution? the pain of losses so high that you guys is so much higher,” said Savage. Smithsonian, Outside, The New York asymmetry, two-sided networks, and toward malware problems. Economic solutions. Technological take action?” He asked later on, dur- “Malware has such high evolvability, Times and many other magazines and externalities. Savage, too, dealt with the eco- solutions. Legal factors, industry facing one of the workshop’s tactical it may evolve to the point where the newspapers. He is also the author of Information asymmetry, he ex- nomics of malware, although more tors. Phishing, patching, spamming, sessions: “How much of malware Internet is no longer useable.” Conspiranoia!

30 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 31 finance Risk in Financial Markets— Learning from Nature

By John Whitfield

On July 19, 2007, the Dow Jones uncertainty. And one of the hottest questions in Industrial hit 14,000 for the first biology—and another focus of SFI research— time. Less than a month later, it was below is, what makes ecosystems robust, and what 13,000, and was still there three months later. causes them to collapse? Nor was this a local American difficulty: as of It certainly looks as if finance might have early December the Bank of England had spent something to learn from biology about stability. £30 billion ($60 billion) propping up the North- Author and hedge fund manager Rick ern Rock bank, one of the United Kingdom’s Bookstaber pointed out that despite the U.S. most notable lenders, which had invested, and economy as a whole being more stable, with lost heavily in the U.S. subprime mortgage fewer recessions and less variation in gross do- market. The European Central Bank, which is mestic product (GDP), financial markets have responsible for maintaining stability of the euro, become less so. This destabilization seems to spent $205 billion propping up markets across have come from within: Michael Mauboussin of the continent. Legg Mason Capital Management and Columbia Not surprisingly, the mortgage crisis and its University reported that four-fifths of the large secondary effects on other markets were on ev- movements in markets are uncorrelated with eryone’s mind at a forum in New York in Octo- outside events such as terrorist attacks, elections, ber 2007 on modeling risk in financial markets, and so on. co-hosted by the Santa Fe Institute and SAC One key consideration for both financial and Capital Partners. Why do such crashes happen? ecological systems is the link between their di- And how can we design and regulate markets to versity and their stability. In nature, changes in reduce the risk of them happening in the future? diversity—for example, when a species is lost Many of the speakers looked to biology for from a place—can trigger a cascade of secondary ideas, comparing the behavior of financial and extinctions among species that depended directly ecological systems, and hoping to learn from the or indirectly on the missing species, either for

left:veer.com © adaptations that life uses to deal with risk and food or to keep their own predators in check.

32 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 33 finance Risk in Financial Markets— Learning from Nature

By John Whitfield

left:veer.com © adaptations that life uses to deal with risk and food or to keep their own predators in check.

32 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 33 Why do such crashes happen? and the fund went out of business. LTCM had their friends, traders buy stocks that they know no exposure to Russian markets, just as Gold- little about, only to sell them at the first sign of And how can we design and man Sach’s Global Alpha Fund, one of those trouble, again creating a downward spiral in pric- regulate markets to reduce the worst hit during recent events (although Gold- es. The crowding in markets caused by this sort risk of them happening in man Sachs as a whole has done well), kept away of copying behavior is an important, but poorly from subprime mortgages. The problem was that understood influence on their behavior, he said. the future? in both cases, the funds that did hold such mort- In general, there was a sense at the meeting gages sold other assets that LTCM and Global that diversity is a good thing, that diverse systems the scent of potential mates whose immune- Alpha held, causing their value to fall. Highly should show smaller fluctuations and quicker system genes are different from their own; it’s leveraged funds, such as LTCM and Global Al- recovery times, and that loss of diversity augurs thought that this makes for healthier offspring. pha, were then forced to sell, which caused the trouble. For example, crop monocultures can be Queen honeybees mate with many males—at price of their holdings to fall, which forced them wiped out by a single pest or pathogen, and the the cost of increased physical wear and tear and to sell more, and so on. LTCM lost $4.6 bil- onset of groupthink in a market—when the wise exposure to predators—but the resulting increase lion in four months; Bloomberg, one of the top crowd becomes a Gadarene herd—is thought to in the genetic diversity of their brood makes for financial news sources, has reported that Global herald a crash. Some studies using agent-based a more able workforce and a better-function- Alpha might end 2007 $6 billion worse off than models, which simulate the behavior of large ing nest. And species in harsh and unpredictable it started, a 60 percent decline. groups of interacting individuals, seem to bear rew

D environments, such as annual desert plants and The network structure of the market may have out this view. rd a shrimp that breed in ephemeral pools, produce exacerbated the effects of the problems in the On the other hand, both Karp and Bookstaber ich R offspring that spread their germination or hatch- mortgage market. Ironically, the collateralized suggested that one source of instability in markets ing over a number of years. Many ecologists see debt options—where mortgage debt was par- AP Photo / this as a form of bet-hedging, a means of reducing celed out and sold on to other investors—that The yucca moth As humans increasingly dominate and manipu- the variance in their reproductive success. widened the impact of the subprime mortgage (Tegeticula Traders on the floor yuccasella) of the New York late wild ecosystems, understanding this process Such forms of diversity are analogous to one crisis were intended to reduce risk, by spread- is the yucca’s Stock Exchange, takes on a keen practical importance. of the central tenets of finance: keep a diverse ing the loans. But this seems to have increased only pollinator. Monday, August And yet what makes for robustness and stabil- portfolio. Fund managers try to ensure that the risk, by making the banks overconfident about Scientists find such 6, 2007. Stocks ity in ecology, and how this relates to diversity, is performance of their various holdings is not too who they loaned money to. Brown compared dependency in fluctuated in still unclear. It might be that the flow of insight closely correlated. They are paying to reduce subprime mortgages to a contagious person nature can provide early trading that can go both ways: the goldmine of data on deci- variance: their whole portfolio will not hit the infecting others with their disease. How such insight into what Monday following a sions, strategies, and their effects to be found in jackpot at once, but nor will it all go south at contagion spreads through markets is still poorly happens in financial sharp pullback the stock market transactions can tell us something once. “Moderation is key,” said Aaron Brown, a understood, as is the markets’ network structure markets. previous Friday. about how, in general, the behavior of many in- risk manager at AQR Capital Management. “The at all scales, from the transactions within a mar- dividuals—each pursuing a particular goal, be it idea is to do a little of a lot [of different things], ket to the way that, for example, the mortgage money or offspring—creates large-scale dynamics so that you do well in the good times and survive market affects the stock market. and emergent patterns. SFI researchers, working the bad times.” Market structure is not the only network that to take the comparisons between the two systems But, Brown added, you can’t tell which stocks affects how markets behave. The social network nc from the metaphorical to the concrete and quan- are synchronized until they all move at once. of dealers is also important. Jason Karp, director I titative, certainly hope so. A famous example is the 1998 collapse of the of research at CR Intrinsic Investors, a division of rchers, a

In both biology and finance, diversity can mean hedge fund Long Term Capital Management SAC Capital Advisors, told the New York meet- ese R lots of different things. Genetic diversity within (LTCM). The fund sought to risk-proof its in- ing that traders’ strategies are heavily influenced

individuals, and between the offspring of a single vestments by assuming a 30 percent correlation by their social contacts. “People make investment n / Photo a individual, seems to be a means of coping with between their performances, far higher than that decisions based on a note written on a napkin,” offm C .

external threats and environmental uncertainty. usually seen. But when the Russian economy Karp told the meeting. Such social connections M

For example, animals, including humans, prefer collapsed, this correlation went up to 70 percent, can also spread damage, he said: influenced by John

34 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 35 Why do such crashes happen? and the fund went out of business. LTCM had their friends, traders buy stocks that they know no exposure to Russian markets, just as Gold- little about, only to sell them at the first sign of And how can we design and man Sach’s Global Alpha Fund, one of those trouble, again creating a downward spiral in pric- regulate markets to reduce the worst hit during recent events (although Gold- es. The crowding in markets caused by this sort risk of them happening in man Sachs as a whole has done well), kept away of copying behavior is an important, but poorly from subprime mortgages. The problem was that understood influence on their behavior, he said. the future? in both cases, the funds that did hold such mort- In general, there was a sense at the meeting gages sold other assets that LTCM and Global that diversity is a good thing, that diverse systems the scent of potential mates whose immune- Alpha held, causing their value to fall. Highly should show smaller fluctuations and quicker system genes are different from their own; it’s leveraged funds, such as LTCM and Global Al- recovery times, and that loss of diversity augurs thought that this makes for healthier offspring. pha, were then forced to sell, which caused the trouble. For example, crop monocultures can be Queen honeybees mate with many males—at price of their holdings to fall, which forced them wiped out by a single pest or pathogen, and the the cost of increased physical wear and tear and to sell more, and so on. LTCM lost $4.6 bil- onset of groupthink in a market—when the wise exposure to predators—but the resulting increase lion in four months; Bloomberg, one of the top crowd becomes a Gadarene herd—is thought to in the genetic diversity of their brood makes for financial news sources, has reported that Global herald a crash. Some studies using agent-based a more able workforce and a better-function- Alpha might end 2007 $6 billion worse off than models, which simulate the behavior of large ing nest. And species in harsh and unpredictable it started, a 60 percent decline. groups of interacting individuals, seem to bear rew

external threats and environmental uncertainty. usually seen. But when the Russian economy Karp told the meeting. Such social connections M

For example, animals, including humans, prefer collapsed, this correlation went up to 70 percent, can also spread damage, he said: influenced by John

34 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 35 (although this view is controversial) is the increas- teracting species it becomes less, not more stable, flows within the food web might each be at work. ences the health of the market as a whole. ing number and complexity of financial instru- with species more likely to go extinct. Ecologists SFI Professor J. Doyne Farmer, who also spoke Wild fluctuations in markets are difficult to ments such as derivatives, which allow investors have spent the past quarter century trying to rec- at the conference, has been exploring the connec- explain with the dominant theory in financial to speculate on the performance of a market or oncile May’s model with their observations that tions between biology and finance for a decade. economics—the “efficient market hypothesis”— stock without actually buying that stock. And nature is full of large groups of species interacting For example, he compares different strategies— which assumes that the behavior of markets it is still not clear what diversity actually means, in complex ways, and with field and lab studies such as value investors, who look for companies reflects rational actors making the best use of suggesting that more complex, diverse ecosys- whose share price underestimates their true worth, perfect information. In this case, any volatility tems, in fact, show smaller fluctuations in their and trend-followers, who bet that a rising market reflects accurate adjustments to new knowledge. population sizes. will continue rising, and vice versa—to species. But there is increasing evidence that perfect ra- Ecologists are pursuing a variety of ideas, Farmer also draws parallels between the amount tionality cannot explain the behavior of markets, sometimes conflicting, about what gives rise to of money invested in each strategy to its popula- and that some, or much, of their volatility is stability, which—like diversity and complexity— tion size, and the flow of money through a market noise that impedes decision making. Understand- has been defined in several ways. One significant to the flow of energy through a food web. He has ing the sources of this noise should help reduce difference between real ecosystems and May’s found that the same mathematics used to describe it, and improve markets’ performance. model is that species interactions are not ran- how predators and prey affect each other’s popula- dom. Using data and models drawn from real-life tion dynamics—thought to be another contribu- food webs (networks describing who eats whom tor to the robustness, or otherwise, of ecological Of course, even the best-regulated market is in ecosystems), researchers including SFI research systems—can also describe the way that different vulnerable to human error; the subprime fellow Jennifer Dunne have found that they show trading strategies interact with one another. mortgage crisis was triggered by old-fashioned consistent non-random patterns in their degree Recently, Farmer and his colleagues obtained distribution (how the number of feeding links access to transaction data from exchanges in bad decisions—loaning money to people who per species is distributed across the whole set of London, Spain, New York, and Taiwan. The couldn’t pay it back. species). Dunne has shown that highly connected Taiwan data set, for example, has tens of billions © veer.com© food webs, with many links between the differ- of trades, with information on who made them. in financial terms. Here, ecology might provide ent species, are less likely to experience cascading This will allow the team to analyze financial mar- Of course, even the best-regulated market is Easily influenced by an insight, because it seems that it’s not species extinctions from loss of a particular species or set kets in unprecedented detail. In effect, they hope vulnerable to human error; the subprime mort- their friends—like diversity per se that counts, but what you do with of species, and thus are more robust than simpler, to fast-forward through the past 250 years of the gage crisis was triggered by old-fashioned bad de- sheep—some traders buy stocks it—robustness or volatility results as much from less well-connected webs. study of species, from basic taxonomy (identify- cisions—loaning money to people who couldn’t that they know the form taken by the network of interactions Models also show that food webs, like other ing and classifying the different strategies found pay it back. Most organisms and ecosystems have little about, only between the different players as from the number networks such as the Internet, are vulnerable to in the market), to cutting-edge ecology (under- been around long enough to have experienced to sell them at of players. loss of the most highly connected players. But it standing the interactions of different strategies, repeated cycles of good times and bad, and, as the first sign of Early theoretical ecologists also thought diver- turns out that in real systems, highly connected and their effect on the market as a whole). If one the food webs of Adirondack lakes show, to come trouble, creating a sity a good thing. In 1955, Robert MacArthur species are very unlikely to go extinct. In a study strategy increases its profits while another slumps, to terms with a variety of stresses, so that dealing downward spiral in hypothesized that the more links there were of the food webs of fish and plankton species for example, the former might be said to be prey- with risk is an important part of their makeup. prices. between species, the more stable the ecosystem, living in lakes in the Adirondacks, Dunne and ing on the latter. And over periods of months They know that bubbles burst and easy money as it reduced the destabilizing effects of any one colleagues found that the species most vulnerable to years, evolutionary changes in strategies and cycles don’t last forever. Perhaps the world of fi- species becoming dominant. A large number of to extinction are the ones that result in the few- composition of the market may also emerge. nance could learn from life that, in the long term, links, he suggested, is easier to achieve in more est secondary extinctions. This suggests, at least Ultimately, the team hopes to be able to reveal survival is a more realistic aim than victory. t diverse communities. But in 1972, Robert May in the absence of human manipulation, that the the market’s network structures, and to test the turned the field on its head—and launched the structure of ecosystems maximizes biodiversity idea that diverse markets function better than John Whitfield is a London-based science writer modern study of how diversity and complexity persistence. Researchers are now trying to de- homogenous ones. This, in turn, could reveal the and author of In the Beat of a Heart: Life, Energy, influences ecological stability—in a now-classic duce what shapes ecological networks into these effect that the rules governing exchanges, such as and the Unity of Nature. From September through paper in which he showed that if you add species robust configurations—forces such as natural se- the charges levied on different forms of trades, December 2007, he was science writer in resi- or links to a randomly structured network of in- lection or thermodynamic constraints on energy have on individual trades, and how this influ- dence at SFI.

36 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 37 (although this view is controversial) is the increas- teracting species it becomes less, not more stable, flows within the food web might each be at work. ences the health of the market as a whole. ing number and complexity of financial instru- with species more likely to go extinct. Ecologists SFI Professor J. Doyne Farmer, who also spoke Wild fluctuations in markets are difficult to ments such as derivatives, which allow investors have spent the past quarter century trying to rec- at the conference, has been exploring the connec- explain with the dominant theory in financial to speculate on the performance of a market or oncile May’s model with their observations that tions between biology and finance for a decade. economics—the “efficient market hypothesis”— stock without actually buying that stock. And nature is full of large groups of species interacting For example, he compares different strategies— which assumes that the behavior of markets it is still not clear what diversity actually means, in complex ways, and with field and lab studies such as value investors, who look for companies reflects rational actors making the best use of suggesting that more complex, diverse ecosys- whose share price underestimates their true worth, perfect information. In this case, any volatility tems, in fact, show smaller fluctuations in their and trend-followers, who bet that a rising market reflects accurate adjustments to new knowledge. population sizes. will continue rising, and vice versa—to species. But there is increasing evidence that perfect ra- Ecologists are pursuing a variety of ideas, Farmer also draws parallels between the amount tionality cannot explain the behavior of markets, sometimes conflicting, about what gives rise to of money invested in each strategy to its popula- and that some, or much, of their volatility is stability, which—like diversity and complexity— tion size, and the flow of money through a market noise that impedes decision making. Understand- has been defined in several ways. One significant to the flow of energy through a food web. He has ing the sources of this noise should help reduce difference between real ecosystems and May’s found that the same mathematics used to describe it, and improve markets’ performance. model is that species interactions are not ran- how predators and prey affect each other’s popula- dom. Using data and models drawn from real-life tion dynamics—thought to be another contribu- food webs (networks describing who eats whom tor to the robustness, or otherwise, of ecological Of course, even the best-regulated market is in ecosystems), researchers including SFI research systems—can also describe the way that different vulnerable to human error; the subprime fellow Jennifer Dunne have found that they show trading strategies interact with one another. mortgage crisis was triggered by old-fashioned consistent non-random patterns in their degree Recently, Farmer and his colleagues obtained distribution (how the number of feeding links access to transaction data from exchanges in bad decisions—loaning money to people who per species is distributed across the whole set of London, Spain, New York, and Taiwan. The couldn’t pay it back. species). Dunne has shown that highly connected Taiwan data set, for example, has tens of billions © veer.com© food webs, with many links between the differ- of trades, with information on who made them. in financial terms. Here, ecology might provide ent species, are less likely to experience cascading This will allow the team to analyze financial mar- Of course, even the best-regulated market is Easily influenced by an insight, because it seems that it’s not species extinctions from loss of a particular species or set kets in unprecedented detail. In effect, they hope vulnerable to human error; the subprime mort- their friends—like diversity per se that counts, but what you do with of species, and thus are more robust than simpler, to fast-forward through the past 250 years of the gage crisis was triggered by old-fashioned bad de- sheep—some traders buy stocks it—robustness or volatility results as much from less well-connected webs. study of species, from basic taxonomy (identify- cisions—loaning money to people who couldn’t that they know the form taken by the network of interactions Models also show that food webs, like other ing and classifying the different strategies found pay it back. Most organisms and ecosystems have little about, only between the different players as from the number networks such as the Internet, are vulnerable to in the market), to cutting-edge ecology (under- been around long enough to have experienced to sell them at of players. loss of the most highly connected players. But it standing the interactions of different strategies, repeated cycles of good times and bad, and, as the first sign of Early theoretical ecologists also thought diver- turns out that in real systems, highly connected and their effect on the market as a whole). If one the food webs of Adirondack lakes show, to come trouble, creating a sity a good thing. In 1955, Robert MacArthur species are very unlikely to go extinct. In a study strategy increases its profits while another slumps, to terms with a variety of stresses, so that dealing downward spiral in hypothesized that the more links there were of the food webs of fish and plankton species for example, the former might be said to be prey- with risk is an important part of their makeup. prices. between species, the more stable the ecosystem, living in lakes in the Adirondacks, Dunne and ing on the latter. And over periods of months They know that bubbles burst and easy money as it reduced the destabilizing effects of any one colleagues found that the species most vulnerable to years, evolutionary changes in strategies and cycles don’t last forever. Perhaps the world of fi- species becoming dominant. A large number of to extinction are the ones that result in the few- composition of the market may also emerge. nance could learn from life that, in the long term, links, he suggested, is easier to achieve in more est secondary extinctions. This suggests, at least Ultimately, the team hopes to be able to reveal survival is a more realistic aim than victory. t diverse communities. But in 1972, Robert May in the absence of human manipulation, that the the market’s network structures, and to test the turned the field on its head—and launched the structure of ecosystems maximizes biodiversity idea that diverse markets function better than John Whitfield is a London-based science writer modern study of how diversity and complexity persistence. Researchers are now trying to de- homogenous ones. This, in turn, could reveal the and author of In the Beat of a Heart: Life, Energy, influences ecological stability—in a now-classic duce what shapes ecological networks into these effect that the rules governing exchanges, such as and the Unity of Nature. From September through paper in which he showed that if you add species robust configurations—forces such as natural se- the charges levied on different forms of trades, December 2007, he was science writer in resi- or links to a randomly structured network of in- lection or thermodynamic constraints on energy have on individual trades, and how this influ- dence at SFI.

36 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 37 finance/sustainability

When Diversity Vanishes by Don Monroe

Complex systems, from ecolo- of extinction. “At least if we’re lucky, we’re not,” gies to economies, do interesting and he said. Nonetheless, “the crisis is real.” Erwin unexpected things. Much of this rich behavior emphasized that there are many types of diver- can be traced to the networks through which sity, which do not have the same impact. For ex- The Global Seed the underlying “agents” affect each other. Often, ample, individual species on different branches of Vault on the however, diversity of the agents themselves is es- the tree of life forms can become extinct without Norwegian island sential. If they act too similarly, the entire system substantial effect, but losing the same number of of Svalbard opened in February 2008. can cease to function. At the annual symposium species on a single branch could eliminate that Descending almost of the Santa Fe Institute Business Network, No- entire branch. 500 feet under the vember 1–3, 2007, an array of experts explored Global extinction reflects the combined chang- permafrost, it is this “diversity collapse” in contexts ranging from es in smaller, individual ecosystems around the designed to with- ecology and the food we eat, to finance and orga- world. Andrew Dobson of Princeton University stand earthquakes, nizational structure. described what he called “probably the best-stud- extreme temperature ied” ecosystem: the Serengeti National Park in fluctuations, and Ecological Collapse Tanzania. Established in 1951, this park and the bomb blasts. The most dramatic diversity collapses are mass surrounding areas provide “a natural example of extinctions, which have wiped out much of life what happens when we perturb an ecosystem,” he ox

five times in Earth’s history. Doug Erwin, of the observed. Outside the park, the ecology changes C predators, are more visible but provide less value, of healthy ecosystems that helps them to resist

National Museum of Natural History and SFI, dramatically because of farming and grazing. The n c.K. he said. They are also more sensitive to changes, disruption. In contrast, monocultures, such as a said that one of these, the end-Permian extinc- difference is most notable at the highest trophic so “monitoring these brittle species gives an early the endless fields of U.S. Midwestern corn, can

tion, wiped out “90 to 95 percent of everything levels in the food chain, Dobson observed. Above: de warning” of damage. succumb to a single pest. in the oceans, about 70 percent of everything Dobson contributed to the Millennium Eco- “We have a scarily short time scale to under- The important work of Robert May in the on land, and by all accounts was about the best system Assessment, which framed the contribu- stand how ecosystems collapse,” Dobson com- 1970s, however, showed that complexity actually thing that ever happened to life on Earth.” The tions of healthy ecosystems, at least in part, in mented. “Most large natural ecosystems will be reduces stability in some mathematical models. extinctions made room for later innovation—but terms of the economic “services” they provide destroyed in the next 30 to 50 years. The quality Ever since, Pascual said, ecologists have tried to not right away. “Eventually the diversity got big- to people. Almost half of the value, Dobson of human life on this planet is dependent on the understand “how more realistic structures lead to ger than before, but it took four million years to said, comes from the most basic level, includ- economic services supplied by those webs.” higher stability.” even get started.” ing bacteria, and another one third from plants. Historically, said Mercedes Pascual of the Uni- Instead of studying small perturbations as May In contrast to some observers, Erwin does These lower levels also tend to be more resilient. versity of Michigan and SFI, ecologists viewed did, Stefano Allesina, of NCEAS, and Pascual not believe that we are entering a “sixth wave” Higher trophic levels, including grazers and complexity in food webs as an essential feature looked at major shifts such as the disappearance

38 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 39 finance/sustainability

When Diversity Vanishes by Don Monroe

38 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 39 my a td / Al my L a / Al k l Press a oo C ry a © G © Pictori

of prey causing a predator to become extinct. Although the system as a whole seems static, Managed Ecosystems Cary Fowler of the Global Crop Diversity Trust Natural They also used food webs taken directly from individual species are not. “Individual popula- If the ecologists are right, natural ecosystems, asserted that this system depends on underappre- ecosystems, which have evolved com- ecological studies instead of mathematically gen- tions…are all going up and down like crazy,” which have evolved complex webs of interac- ciated diversity of plant varieties. plex webs of erated networks. They then determined which Pascual said. These fluctuations at one level may tions, are ideally “managed” by leaving them Even as agriculture has focused on fewer spe- interactions, are pathways are functional and which are redun- even enhance the stability at a higher level. The alone—when possible. In stark contrast, agri- cies over the past 12,000 to 15,000 years, Fowler ideally “managed” dant, from the perspective of secondary extinc- extinction of individual species may therefore be cultural crop management resembles control said, “diversity in some sense was increasing,” as by leaving them tion, and found that in real webs about 90 per- a misleading measure of the loss of diversity. theory—an engineering tool developed for much farmers in different regions selected variants with alone. The images cent of connections are functional, independent The question of the best level for gauging di- simpler systems—and features simplified eco- different traits. “There are about 120,000 differ- above and on the of the size of the ecosystem. versity also arose in work by Katia Koelle (Penn systems that depend heavily on external inputs. ent varieties of rice, each as distinct one from the facing page depict As a result, Pascual observed, “Even when sec- State), Sarah Cobey (University of Michigan), Much of the corn in the U.S. is grown with pe- other as a Great Dane from a Chihuahua,” he the contrast ondary extinctions are not observed, the loss of Bryan Grenfell (Penn State), and Pascual on the troleum-derived pesticides and fertilizer, and fed commented. between the species makes ecosystems more fragile to further evolution of flu. Genes evolve continuously, but to cattle whose excrement then becomes toxic Industrial agriculture, Fowler said, threatens pristine land of extinctions.” There may be little warning of an often with no effect on the “phenotype”: the sur- waste rather than nutrition for plants. this variation within species, although scientists the Serengeti approaching “tipping point,” in which the entire face proteins that determine immune response. An alternative was described by Joel Salatin, still cannot agree on how to measure it. In re- National Park in ecosystem collapses. The researchers modeled genetic evolution who recovered marginal land in Virginia by using sponse to new challenges, he wondered, “can we Tanzania and land outside the park, Pascual suggested that work by SFI External coupled with the prevalence in the human popu- cow manure to fertilize the grasses that are the continue to develop our agriculture without di- where grazing and Professor Ricard Solé may clarify the dynam- lation of immunity to particular variants. In this natural food for cattle. But Salatin complained versity?” His answer: “Obviously we can’t.” other uses have ics of interacting populations contributing to model, viruses multiply rapidly whenever they that the many regulations aimed at industrial- The expected global warming in coming prevailed. extinctions in these ecological networks. In his take on a new phenotype, quickly crowding out scale production present formidable barriers to decades makes these issues especially urgent. model, as species continually become extinct and other variants. “This pattern of boom and bust is small farms like his. “My guess is we are ill-prepared for this kind of new ones immigrate into a region, the food web explained by an interaction of genetic drift and In spite of such efforts, local production is change,” Fowler said. “But if we are prepared, it forms a self-organized state, with many of the selection, and not exclusively one or the other,” likely to be an anomaly in an industrial food will be because of the gene banks and the diver- features observed in real webs. Pascual said. system that prizes cheap and abundant food. But sity they contain.”

40 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 41 my a td / Al my L a / Al k l Press a oo C ry a © G © Pictori

40 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 41 similarly as other investors “My guess is we are ill-prepared for this kind of change,” Fowler said. “But if we are desperately sold the same prepared, it will be because of the gene banks and the diversity they contain.” assets (see “Risk in Finan- cial Markets” in this issue). “Diversity collapses are really Numerous gene banks have been storing seeds able to agricultural pests. Barrantes and Stepha- the source of inefficiencies in for crops and other plants around the world. nie Forrest, of the University of New Mexico and markets,” asserted Michael Unfortunately, Fowler said, many of them are SFI, have been exploring how artificial variability Mauboussin, Chief Invest- poorly funded and maintained. He stressed that in computer systems can slow the spread of ment Strategist at Legg Ma- for a modest cost—an endowment of some $250 malware. In any successful scheme, she stressed, son Capital Management and

million—“we can conserve the gene pool of our computers must still interoperate with similar an SFI trustee. hoto.com kp major crops in perpetuity.” As a start in this direc- performance and cost. Mauboussin reviewed three stoc i

tion, Fowler’s organization is funding a facility Diversity can be introduced at many levels, theories for how markets be- ms/ a above the Arctic Circle to provide a global seed re- Barrantes said. For example, the well-known come efficient, meaning that ulwilli

pository, sometimes called the “doomsday vault.” “buffer overflow” attacks rely on long data spill- prices reflect value. The first, pa © ing into areas of memory intended for programs. in which all investors behave Valuing Diversity Varying the locations of these segments can often rationally, is unrealistic. A second explanation, we cannot count on the right amount of diversity Complexity in food No such vault exists to preserve human culture. thwart the spread of infection between different which requires only that some investors exploit— existing,” Page said. webs helps them “The forces of homogenization are rampant,” said machines. This kind of artificial diversification and thereby remove—arbitrage opportunities, Page warned that merely recognizing the resist disruption. Suzanne Romaine of the University of Oxford. is “currently being used in two major operating “has failed us in critical junctures,” he asserted. advantages of cognitive diversity might not be In contrast, mono- She described the rapid extinction of languages, systems,” Barrantes said. Mauboussin contrasted these models with a enough to preserve it. “Diversity’s benefits may cultures, such as as large ones like Mandarin, Spanish, and English Diversity collapse in computer systems is prob- view of “markets as a complex adaptive system, be public goods that are not in any one’s interest the vast fields of spread at the expense of smaller ones. ably well ahead of that in ecosystems, Forrest where prices essentially emerge from the interac- to maintain.” U.S. Midwestern As a linguist, Romaine values languages as suggested. But she noted that “adding diversity tion of many agents.” In this view, also called One way to maintain diversity is through corn, can succumb data for her own work. But she sees linguistic back in is much easier than it would be in the “the wisdom of crowds,” three conditions assure time-varying selective pressures that prevent one to a single pest. extinction as part of a larger problem. “It’s not natural world.” efficiency: diversity among investors, an aggrega- idea from dominating. However, Page showed a just languages that are at stake, but forms of The dominance of Google, eBay, and others tion mechanism, and financial incentives. Of simple model in which such churn did not pre- knowledge,” she said. “They can’t be separated shows that “online niches are often winner-take these assumptions, he said, “the most likely to be vent uniformity. He suggested that maintaining from people, their identities, their cultural heri- all,” notes Virgil Griffith of the California Insti- violated is diversity,” which may decline imper- diversity also requires diverse selective processes tage, their well-being and their rights.” She also tute of Technology and SFI. But Griffith claimed ceptibly until it suddenly collapses. or richer networks, so that the criteria for picking stressed that language diversity and biodiversity that “promiscuous interoperability” can promote Scott Page of the University of Michigan and winners varies. often disappear together. diversity by allowing people to use data for new SFI, has compiled many ways that diverse groups The broad range of speakers at this symposium In a similar way, shared communication drives purposes. “When all-powerful monocultures outperform individuals. In expert judgment, for shows that the Santa Fe Institute is in no danger a homogenization in computer systems, said Ga- make data available, diversity flourishes,” Griffith example, as in diversified portfolios, the average of diversity collapse, although similar principles briela Barrantes, of the University of Costa Rica claimed. “The users diversify the monoculture, judgment of a group is always better. “This is apply in very different fields. Still, in the world and SFI. The dominance of Microsoft in per- not the other way around.” not a feel-good statement, this is a mathematical outside, increasing interconnectedness seriously sonal computer software is only the most visible theorem,” Page said. For problem solving, having threatens diversity in both human organizations example of this diversity collapse, she said. Diverse Perspectives multiple strategies can help a group evade road- and ecosystems. t stwood a e

a This uniform computing environment is sensi- In finance, diversification reduces risk by spread- blocks that hamper any one approach. ul

pa tive to threats, just as monocultures are vulner- ing money among assets that respond differently “In human systems,” Page said, “the thing that Don Monroe (www.donmonroe.info) writes on during market moves. But during the 1998 in- really works against cognitive diversity is selec- biology, physics and technology from Berkeley LEFT: Though much industrial agriculture favors cheap and ternational financial crisis, Long-Term Capital tion.” The increasingly global marketplace of Heights, New Jersey. Prior to 2003, he spent 18 abundant food, a great deal of diversity, such as these varieties Management suffered enormous losses when its ideas selects the current “best practices” at the years in basic and applied research at Bell Labo- of beans, can still be found. ostensibly diversified investments began to react expense of other approaches. “If the world is flat, ratories.

42 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 43 similarly as other investors “My guess is we are ill-prepared for this kind of change,” Fowler said. “But if we are desperately sold the same prepared, it will be because of the gene banks and the diversity they contain.” assets (see “Risk in Finan- cial Markets” in this issue). “Diversity collapses are really Numerous gene banks have been storing seeds able to agricultural pests. Barrantes and Stepha- the source of inefficiencies in for crops and other plants around the world. nie Forrest, of the University of New Mexico and markets,” asserted Michael Unfortunately, Fowler said, many of them are SFI, have been exploring how artificial variability Mauboussin, Chief Invest- poorly funded and maintained. He stressed that in computer systems can slow the spread of ment Strategist at Legg Ma- for a modest cost—an endowment of some $250 malware. In any successful scheme, she stressed, son Capital Management and

a This uniform computing environment is sensi- In finance, diversification reduces risk by spread- blocks that hamper any one approach. ul

42 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 43 finance

In this article we economist Leon Walras (1834–1910). For Wal- The NASDAQ look at some of the ras, the physical forces that could hold an object (National Asso- ideas explored during at rest, the so-called “balance point,” were ciation of Securities “Markets Week.”1 directly comparable to the “forces” of supply and Dealers Automated The markets are a demand that would eventually balance out to Quotations), the fantastic and data- create a market equilibrium price. largest electronic rich example of a As Smith pointed out, the analogy has many screen-based equity complex system. In flaws—chief of which is that it neglects the possi- securities trading fact, part of SFI’s bility for a physical system to move forever about market in the U.S., lists approximately beginnings can be an equilibrium point. To the audience’s delight, 3,200 companies traced to the bring- Smith introduced the term “analogon” to de- ing together of phys- scribe alluring but incorrect (or even dangerous) icists and economists scientific analogies.2 (at the behest and Although Walras’s analogizing of mechanics with backing from and markets was a misstep, it sowed the seeds then Citicorp CEO of an interesting idea. The study of markets John Reed) to un- through the lens of physics and the tools of rela- derstand the global tively sophisticated mathematics began to gain financial market as a adherents, and the metaphor of mechanics was complex system. soon replaced by a new physical analogy, in the form of classical thermodynamics. With its laws Every June, an effervescent and Beware the “Analogon” describing the behavior of a perfect gas—relat- diverse mix of graduate students and post- The compressed form of the CSSS lets paring temperature, pressure, and volume through doctoral fellows, plus a few junior faculty and ticipants see the interrelations among different equations of state—the analogy would be one of even a handful of business folk, come to the complex systems, be they ecologies, economies, a market driven toward an equilibrium assign- Economics Santa Fe Institute Complex Systems Summer societies, or brains. But good science requires ment of goods or an equilibrium price, like a gas School (CSSS). The CSSS is a four-week intel- more than analogies. Identification must lead to in a container compressed by a piston. As Smith lectual sprint designed to introduce a new gen- investigation, careful consideration, and brutally pointed out, this comparison also falls prey to and Markets eration of researchers to the ideas and techniques honest evaluation to see where an analogy holds, analogons, chief among which is the path-de- of complex systems. A four-week course can’t where it is useful, and where it falls apart. pendent nature3 of most economic processes. In as Complex Systems: possibly touch on every aspect of the discipline, Appropriately, Markets Week began with a recent work with External Faculty Member and so each week is given a theme. The 2007 CSSS cautionary tale of the use of scientific analogy in New School Professor Duncan Foley, Smith gives A Postcard from the 2007 Complex began with a week-long introduction to the fun- the study of economics and markets. “Neoclassi- the analogy more solid foundations, showing that Systems Summer School damentals of the subject, outlining its physical cal economics,” the modern paradigm of markets for “quasi-linear economies”4 there is a mapping roots in chaotic dynamics and the basics of agent- in which price emerges out of the competing or dictionary that essentially allows one to use based modeling. Week two provided a deep look pressures of supply and demand, has been inter- classical thermodynamics to draw rigorous or By Daniel Rockmore lamy

at the new science of networks and at ecological A preted in the light of contemporaneous physical formal conclusions regarding the market of inter- systems—all of which was good preparation for discoveries and theories. SFI Research Professor est. However, while the technical analogy makes the third week’s focus on financial markets. Week Eric Smith gave a wonderful thumbnail sketch sense, it is, Smith says, debatable as to whether or four touched on allometry and scaling laws in of this history—which becomes something of not these quasi-linear economies are a reasonable biology—for which SFI has become so well- a pre-history of econophysics. This dialogue is model of reality. The lesson is to apply a healthy

known—as well as epidemiology and evolution. above: © Cosmo Condina / generally thought to have begun with the French skepticism at all points of the investigation.

44 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 45 finance

known—as well as epidemiology and evolution. above: © Cosmo Condina / generally thought to have begun with the French skepticism at all points of the investigation.

44 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 45 situation, know something about history, and hope that this seed will grow into the dynamics have some simple update rule for making trading of the whole system. The art is not to simplify decisions and evolving their strategies. In outline, too much—remember Einstein’s admonition: this is similar to the many program trading for- “Make everything as simple as possible, but not mats widely used today—they are constrained by simpler.” Alternatively, we might also take the partial market information and have hard tempo- system as it is, consider it from the “top-down” ral constraints for executing their strategies. The and try to characterize the emergent phenome- Figure 1: A cartoon comparison of forces on a ball rolling down a hill that will result in Smith showed how even a rule-based marketplace agents’ (and programs’) strategies reflect traits non of interest. Once again, this involves choices. its equilibrium position at the bottom versus the “forces” of supply and demand that accompanied by the inevitable mismatches that such as degree of risk aversion, market depth, Understandings of a complex system are result in an equilibrium price for a good. occur in the random arrival of buyers and sellers and memory of past market behavior. The fun achieved when top-down and bottom-up ap- with random ranges of requirements and con- (i.e., surprise) in these models lies in the strate- proaches meet. In the case of an individual eq- Nevertheless, the thermodynamics analogy is al- straints (the “continuous double auction”) is able gies’ evolution—poorly performing strategies go uity, replicating various properties of the price luring, and closer to what you might call the com- to replicate many of the characteristic statistics the way of the dodo, while productive strategies series (long tails, long memory, correlated volatil- plex systems ethos, of an emergent structure born of the price dynamics of the actual marketplace persist and possibly adapt, finding trading niches. ity) in simulation hints that the model may be of relatively simple interactions among a multi- while simultaneously exposing various dependen- Questions abound: Do various populations and on the right track. A variety of basic statistical tude of individuals. In thermodynamics, these are cies of the dynamics of order flow.5 We may be update rules generate a market equilibrium? tools can measure the distance between simula- the trillions upon trillions of particles of varying closer to zero intelligence than we know... How sensitive are the dynamics to the changes tion and reality. But what happens when we turn energies, bouncing around some container accord- The dynamics and market effects of order flow in parameters? Are there recognizable regimes of our attention to an actual market—say, a col- ing to the laws of classical mechanics, collectively were also in part the subject of the lectures of behavior in the parameter space? lection of assets whose behaviors are necessarily giving rise to temperature and pressure. In the Smith’s colleague (and sometimes co-author), The CSSS was treated to four lectures on agent- correlated—where strategies must account for markets, it is a variety of many competing partici- SFI Research Professor J. Doyne Farmer. Farmer based markets by SFI External Professor and the interrelation of a wide range of complicated pants, from individuals to institutions, working in discussed the interplay of order flow and market Brandeis Professor of Economics, Blake LeBaron. assets. From 10,000 feet, a view of “the mar- the marketplace, which gives rise to price. impact—the effect that a trade has on the price LeBaron was one of the developers of the Santa ket” might look something like the cartoon in Of course, particles in a box have no agency, of the thing traded. He also explored what a Fe Institute Artificial Stock Market, which was Figure 2—in which capital flows between the and the degree of agency assigned to market par- science of finance and markets might look like, among the early agent-based markets. LeBaron different “submarkets” of equities, bonds, etc. ticipants is a critical component of any market pointing out where current theory (one that puts went over a good deal of the basics on how these The market is a network of large markets, model or simulation. The spectrum ranges from the paradigm of equilibrium at the center of all markets are constructed, and discussed some more which are in turn themselves networks of markets the “zero intelligence” model, which mirrors the considerations) falls down and explaining the im- recent work in which he and his colleagues have and so on and so on, until we arrive at some sort random motions of particles in a box, to the tra- portance of simulations based on more realistic built artificial markets with heterogeneous “inves- of tradable “atom.” Any bottom-up model would ditional “homo economicus”—the fully rational assumptions.6 tors.” These models generate market dynamics ultimately need to reproduce this kind of hierar- and omniscient actor capable of optimizing be- Both zero intelligence and full rationality that replicate much of the behavior usually associ- chical structure, a structure that seems to be both havior in light of total information. An interest- have the advantage of enabling (in many cases) ated with the time series of a given stock price, characteristic of financial markets, but also a gen- ing interpretation of “rationality” is to consider analytic solutions for their models. At intermedi- such as distributions with “fat tails”7 the “per- eral feature of many complex systems. how much of market dynamics derives from ate levels of rationality, mathematical proofs are sistence” of trading volume,8 the so-called “long How to articulate this structure in a math- smith institutional constraints as opposed to individual relatively rare. Instead, we can turn to the world c memory” of volatility (meaning that price fluctua- ematical way? This was the topic of three lectures preference. In this view, zero intelligence means of agent-based simulations and artificial markets. tions persist), and correlations between volatility by Greg Leibon of Dartmouth College. Leibon

a dynamics completely driven by the rules of the In these in silico markets, a large population of kmore and eri and volume. In particular, LeBaron finds connec- showed how tools from statistical learning can c marketplace, while homo economicus is a way participants (the agents) trades a generally small tions between persistence and the modification of analyze the correlation structure of the equities to get at a model in which purely internal (“per- number of assets. The agents have a range of individual strategies via imitation of known suc- market, or the “network of equities,” and reveal sonal”) concerns create social allocations without strategies—behavioral rules dependent on their cessful strategies in an order-driven market. the hierarchical structure described above. The institutions. Presumably, reality lies somewhere risk aversion and the state of the market—which core object of study was the weighted network ourtesy of daniel ro between these two extremes, but even zero intel- is some simple function of market observables. c Markets—A Top-Down Networks View of equities, in which two equities are connected ligence can provide a fair amount of insight into These agents generally have “bounded rational- Agent-based models are “bottom-up”: the sys- by an edge whose strength reflects the degree of

some aspects of price dynamics. In particular, ity”: they know something about the global all diagrams tem’s laws of interaction are specified, with the correlation in the time series of normalized daily

46 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 47 situation, know something about history, and hope that this seed will grow into the dynamics have some simple update rule for making trading of the whole system. The art is not to simplify decisions and evolving their strategies. In outline, too much—remember Einstein’s admonition: this is similar to the many program trading for- “Make everything as simple as possible, but not mats widely used today—they are constrained by simpler.” Alternatively, we might also take the partial market information and have hard tempo- system as it is, consider it from the “top-down” ral constraints for executing their strategies. The and try to characterize the emergent phenome- Figure 1: A cartoon comparison of forces on a ball rolling down a hill that will result in Smith showed how even a rule-based marketplace agents’ (and programs’) strategies reflect traits non of interest. Once again, this involves choices. its equilibrium position at the bottom versus the “forces” of supply and demand that accompanied by the inevitable mismatches that such as degree of risk aversion, market depth, Understandings of a complex system are result in an equilibrium price for a good. occur in the random arrival of buyers and sellers and memory of past market behavior. The fun achieved when top-down and bottom-up ap- with random ranges of requirements and con- (i.e., surprise) in these models lies in the strate- proaches meet. In the case of an individual eq- Nevertheless, the thermodynamics analogy is al- straints (the “continuous double auction”) is able gies’ evolution—poorly performing strategies go uity, replicating various properties of the price luring, and closer to what you might call the com- to replicate many of the characteristic statistics the way of the dodo, while productive strategies series (long tails, long memory, correlated volatil- plex systems ethos, of an emergent structure born of the price dynamics of the actual marketplace persist and possibly adapt, finding trading niches. ity) in simulation hints that the model may be of relatively simple interactions among a multi- while simultaneously exposing various dependen- Questions abound: Do various populations and on the right track. A variety of basic statistical tude of individuals. In thermodynamics, these are cies of the dynamics of order flow.5 We may be update rules generate a market equilibrium? tools can measure the distance between simula- the trillions upon trillions of particles of varying closer to zero intelligence than we know... How sensitive are the dynamics to the changes tion and reality. But what happens when we turn energies, bouncing around some container accord- The dynamics and market effects of order flow in parameters? Are there recognizable regimes of our attention to an actual market—say, a col- ing to the laws of classical mechanics, collectively were also in part the subject of the lectures of behavior in the parameter space? lection of assets whose behaviors are necessarily giving rise to temperature and pressure. In the Smith’s colleague (and sometimes co-author), The CSSS was treated to four lectures on agent- correlated—where strategies must account for markets, it is a variety of many competing partici- SFI Research Professor J. Doyne Farmer. Farmer based markets by SFI External Professor and the interrelation of a wide range of complicated pants, from individuals to institutions, working in discussed the interplay of order flow and market Brandeis Professor of Economics, Blake LeBaron. assets. From 10,000 feet, a view of “the mar- the marketplace, which gives rise to price. impact—the effect that a trade has on the price LeBaron was one of the developers of the Santa ket” might look something like the cartoon in Of course, particles in a box have no agency, of the thing traded. He also explored what a Fe Institute Artificial Stock Market, which was Figure 2—in which capital flows between the and the degree of agency assigned to market par- science of finance and markets might look like, among the early agent-based markets. LeBaron different “submarkets” of equities, bonds, etc. ticipants is a critical component of any market pointing out where current theory (one that puts went over a good deal of the basics on how these The market is a network of large markets, model or simulation. The spectrum ranges from the paradigm of equilibrium at the center of all markets are constructed, and discussed some more which are in turn themselves networks of markets the “zero intelligence” model, which mirrors the considerations) falls down and explaining the im- recent work in which he and his colleagues have and so on and so on, until we arrive at some sort random motions of particles in a box, to the tra- portance of simulations based on more realistic built artificial markets with heterogeneous “inves- of tradable “atom.” Any bottom-up model would ditional “homo economicus”—the fully rational assumptions.6 tors.” These models generate market dynamics ultimately need to reproduce this kind of hierar- and omniscient actor capable of optimizing be- Both zero intelligence and full rationality that replicate much of the behavior usually associ- chical structure, a structure that seems to be both havior in light of total information. An interest- have the advantage of enabling (in many cases) ated with the time series of a given stock price, characteristic of financial markets, but also a gen- ing interpretation of “rationality” is to consider analytic solutions for their models. At intermedi- such as distributions with “fat tails”7 the “per- eral feature of many complex systems. how much of market dynamics derives from ate levels of rationality, mathematical proofs are sistence” of trading volume,8 the so-called “long How to articulate this structure in a math- smith institutional constraints as opposed to individual relatively rare. Instead, we can turn to the world c memory” of volatility (meaning that price fluctua- ematical way? This was the topic of three lectures preference. In this view, zero intelligence means of agent-based simulations and artificial markets. tions persist), and correlations between volatility by Greg Leibon of Dartmouth College. Leibon

46 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 47 Bonds Currency analysis can automatically recognize some known like a member of a given sector. While such a determined. For example, historical “accidents” could very well play a part. In the physics of an ideal gas, it would be the fact that work (for dynamics of the market, such as the cyclic flow of “generative model” still doesn’t account for all example) is a path-dependent function of the state variables pressure, MBS capital between sectors. other kinds of market dependencies and interac- temperature, and volume. The analogons arise when we try to find While some of the clusters reflect or replicate tions, it still provides a more sensible model of analogies for the state variables in an economic process and then “derive” path dependence—or the lack thereof. Equities a traditional view of the market, comprising behavior for price series movement, given sector 4 A quasi-linear economy is one in which the utility function of the AMEX mainly equities from a single “traditional” sec- information. This is a first step toward some sort participants (how much they value any particular assignment of “goods,” tor, other clusters are highly mixed, suggesting of meso-scale artificial market working at a level including money) breaks up into a sum of the cash on hand plus a utility function of the non-cash goods alone. Commodities that new kinds of hybrid sectors have emerged above participants and strategies. NYSE 5 “Order flow” refers to the continual influx of buy and sell orders into or evolved in our era of highly diversified, mul- These lectures were the material that made up the marketplace for a particular equity. NASDAQ tinational, publicly traded entities. A new struc- Markets Week, but of course the subject mat- 6 See also Farmer, J. D., D. E. Smith, and M. Shubik. “Is Economics ture can suggest new markets, new derivative ter was discussed in other weeks too. The com- the Next Physical Science?” Physics Today 58(9) (2005): 37-42 products, new opportunities for investment, as pressed timescale of the CSSS causes the subjects 7 The notion of “fat tails” makes reference to an underlying probability Figure 2: Model of the market network in which capital flows between the distribution and the fact that there is a relatively large probability for the well as new strategies for ameliorating risk. to wash over one another, making participants different “submarkets.” occurrence of large deviations from the mean, in contrast to normally Each cluster can then be viewed as its own think about possible connections and relations. distributed (bell-curve) phenomena. With respect to the markets, this 9 is reflected in high volatility, so that the existence of fat tails makes risk close prices over 15 years. submarket, ready to be analyzed just like the For example, discussions of niche development, management much more difficult. Leibon’s analysis treats the equities network as original market. The process is then iterated innovation, and evolution carried over to discus- 8 Persistence (or hysteresis) refers to the ability to see the effects of a spring model, in which pulling or pushing on until one reaches a submarket that is effectively sions of emergence in markets. Could the new moving large amounts of a stock well beyond the time of the actual sale one piece can send shocks to other pieces. Given completely decorrelated. The result is a more work in food-web dynamics and extinction pat- or purchase—like a slowdown in traffic that persists well past the time of an accident. this mechanical toy composed of thousands of highly textured and geometric view of the corre- terns give insights into risk management—or 9 A direct analysis of the simple time series of close prices results in masses (equities) and almost 2.5 million springs lation structure of the market. It suggests a “null vice versa? Do the methodologies of functional all sorts of spurious identifications of correlation or the lack thereof. A (correlations), the goal is to try to uncover collec- model” of the correlation structure of the market neuroimaging and the hierarchical structure of standard first normalization is to consider the series of relative changes. In addition, we transform the data in order to take out market ef- tions of equities that generally move as one and that is much more interesting and sensible than a neuroanatomy map onto the measurements and fects—that is, general effects due to the movement of capital between then to try to understand the degrees to which strawman “random” assortment of correlations or structures of economies and markets? Analo- markets. 10 This is the content of a paper in preparation: G. Leibon, S. Pauls, subsets of the market are in phase or out. correlation matrices. gies or analogons? Only time and a lot of careful D. Rockmore, and R. Savell, “Hierarchical structures in the equities A similar problem confronts a chemist or mo- By analyzing the correlation structure of thought will tell. t market.” lecular biologist trying to understand (predict) the market as a network, market analysis has the properties of a complicated molecule. These been turned into a geometry problem (Leibon Daniel Rockmore is John G. Kemeny Parents fields use the mathematics of “spectral analysis”— trained as a geometer and topologist). And the Professor of Mathematics at Dartmouth Col- a subject that derives its name from studying the visual form of this network reduces the data of lege where he is also Chair of the Department of response of a substance to various kinds of radia- thousands of 1,000-dimensional time series to a Mathematics and Professor of Computer Science. tion—to reveal the properties of molecules from three-dimensional representation that contains a He is an External Professor at SFI and Director of the (slightly modified) network of connections good deal of the original information.10 the Complex Systems Summer School. This article that describe them. In a similar fashion, the same Identifying this clustered and hierarchical benefited from the helpful comments of Robert mathematics can reveal the structure of the equi- structure is in a sense a data-driven form of eq- Savell and Eric Smith. ties network. In both cases, the “large” modes of uities classification. In Leibon’s final lecture, he the molecule—the perturbations that cause the showed how “Bayesian nets” and more generally 1 Copies of the slides of almost all lectures from the 2007 CSSS as most dramatic resonances—indicate the existence “graphical models,” probabilistic tools for model- well as some related readings can be obtained at http://www.santafe. of large neighborhoods or clusters that effectively ing complicated (i.e., highly non-independent) edu/events/workshops/index.php/CSSS_2007_Santa_Fe-Readings move together. Finding the large modes lets multivariate distributions, could then use this 2 Smith and his co-author SFI External Professor Duncan Foley attri- bute “analogon” to a 1949 paper of J. Lisman. Coupled with “ridiculo- Leibon know how many big clusters to look for. understanding as training data to solve the prob- gram,” Mark Newman’s name for the indiscriminate use of obfuscating Figure 3: Network structure of the equities market after cluster analysis and dimension What is of interest here is that “neighborhoods” lem of classifying unknown or new data streams network diagrams, we now have a growing vocabulary for modes of bad reduction. Labels reflect dominant sector or classification (e.g., country). Node size science. of the market emerge in an unsupervised fashion, as (behaving like) members of a given sector (tra- is proportional to cluster size. Unfilled squares indicate highly mixed (i.e., not easily 3 In this context, path dependence for an economic process can be corresponding to labeled sectors such as technol- ditional or not). These tools can also be turned thought of as the idea that the equilibrium that is achieved (e.g., price) identifiable) clusters. Connections are made for the (relatively) ogy, energy, and so on. Also striking is that the upside down to generate time series that behave would depend on the manner (process) by which the price comes to be closest cluster centroids.

48 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 49 Bonds Currency analysis can automatically recognize some known like a member of a given sector. While such a determined. For example, historical “accidents” could very well play a part. In the physics of an ideal gas, it would be the fact that work (for dynamics of the market, such as the cyclic flow of “generative model” still doesn’t account for all example) is a path-dependent function of the state variables pressure, MBS capital between sectors. other kinds of market dependencies and interac- temperature, and volume. The analogons arise when we try to find While some of the clusters reflect or replicate tions, it still provides a more sensible model of analogies for the state variables in an economic process and then “derive” path dependence—or the lack thereof. Equities a traditional view of the market, comprising behavior for price series movement, given sector 4 A quasi-linear economy is one in which the utility function of the AMEX mainly equities from a single “traditional” sec- information. This is a first step toward some sort participants (how much they value any particular assignment of “goods,” tor, other clusters are highly mixed, suggesting of meso-scale artificial market working at a level including money) breaks up into a sum of the cash on hand plus a utility function of the non-cash goods alone. Commodities that new kinds of hybrid sectors have emerged above participants and strategies. NYSE 5 “Order flow” refers to the continual influx of buy and sell orders into or evolved in our era of highly diversified, mul- These lectures were the material that made up the marketplace for a particular equity. NASDAQ tinational, publicly traded entities. A new struc- Markets Week, but of course the subject mat- 6 See also Farmer, J. D., D. E. Smith, and M. Shubik. “Is Economics ture can suggest new markets, new derivative ter was discussed in other weeks too. The com- the Next Physical Science?” Physics Today 58(9) (2005): 37-42 products, new opportunities for investment, as pressed timescale of the CSSS causes the subjects 7 The notion of “fat tails” makes reference to an underlying probability Figure 2: Model of the market network in which capital flows between the distribution and the fact that there is a relatively large probability for the well as new strategies for ameliorating risk. to wash over one another, making participants different “submarkets.” occurrence of large deviations from the mean, in contrast to normally Each cluster can then be viewed as its own think about possible connections and relations. distributed (bell-curve) phenomena. With respect to the markets, this 9 is reflected in high volatility, so that the existence of fat tails makes risk close prices over 15 years. submarket, ready to be analyzed just like the For example, discussions of niche development, management much more difficult. Leibon’s analysis treats the equities network as original market. The process is then iterated innovation, and evolution carried over to discus- 8 Persistence (or hysteresis) refers to the ability to see the effects of a spring model, in which pulling or pushing on until one reaches a submarket that is effectively sions of emergence in markets. Could the new moving large amounts of a stock well beyond the time of the actual sale one piece can send shocks to other pieces. Given completely decorrelated. The result is a more work in food-web dynamics and extinction pat- or purchase—like a slowdown in traffic that persists well past the time of an accident. this mechanical toy composed of thousands of highly textured and geometric view of the corre- terns give insights into risk management—or 9 A direct analysis of the simple time series of close prices results in masses (equities) and almost 2.5 million springs lation structure of the market. It suggests a “null vice versa? Do the methodologies of functional all sorts of spurious identifications of correlation or the lack thereof. A (correlations), the goal is to try to uncover collec- model” of the correlation structure of the market neuroimaging and the hierarchical structure of standard first normalization is to consider the series of relative changes. In addition, we transform the data in order to take out market ef- tions of equities that generally move as one and that is much more interesting and sensible than a neuroanatomy map onto the measurements and fects—that is, general effects due to the movement of capital between then to try to understand the degrees to which strawman “random” assortment of correlations or structures of economies and markets? Analo- markets. 10 This is the content of a paper in preparation: G. Leibon, S. Pauls, subsets of the market are in phase or out. correlation matrices. gies or analogons? Only time and a lot of careful D. Rockmore, and R. Savell, “Hierarchical structures in the equities A similar problem confronts a chemist or mo- By analyzing the correlation structure of thought will tell. t market.” lecular biologist trying to understand (predict) the market as a network, market analysis has the properties of a complicated molecule. These been turned into a geometry problem (Leibon Daniel Rockmore is John G. Kemeny Parents fields use the mathematics of “spectral analysis”— trained as a geometer and topologist). And the Professor of Mathematics at Dartmouth Col- a subject that derives its name from studying the visual form of this network reduces the data of lege where he is also Chair of the Department of response of a substance to various kinds of radia- thousands of 1,000-dimensional time series to a Mathematics and Professor of Computer Science. tion—to reveal the properties of molecules from three-dimensional representation that contains a He is an External Professor at SFI and Director of the (slightly modified) network of connections good deal of the original information.10 the Complex Systems Summer School. This article that describe them. In a similar fashion, the same Identifying this clustered and hierarchical benefited from the helpful comments of Robert mathematics can reveal the structure of the equi- structure is in a sense a data-driven form of eq- Savell and Eric Smith. ties network. In both cases, the “large” modes of uities classification. In Leibon’s final lecture, he the molecule—the perturbations that cause the showed how “Bayesian nets” and more generally 1 Copies of the slides of almost all lectures from the 2007 CSSS as most dramatic resonances—indicate the existence “graphical models,” probabilistic tools for model- well as some related readings can be obtained at http://www.santafe. of large neighborhoods or clusters that effectively ing complicated (i.e., highly non-independent) edu/events/workshops/index.php/CSSS_2007_Santa_Fe-Readings move together. Finding the large modes lets multivariate distributions, could then use this 2 Smith and his co-author SFI External Professor Duncan Foley attri- bute “analogon” to a 1949 paper of J. Lisman. Coupled with “ridiculo- Leibon know how many big clusters to look for. understanding as training data to solve the prob- gram,” Mark Newman’s name for the indiscriminate use of obfuscating Figure 3: Network structure of the equities market after cluster analysis and dimension What is of interest here is that “neighborhoods” lem of classifying unknown or new data streams network diagrams, we now have a growing vocabulary for modes of bad reduction. Labels reflect dominant sector or classification (e.g., country). Node size science. of the market emerge in an unsupervised fashion, as (behaving like) members of a given sector (tra- is proportional to cluster size. Unfilled squares indicate highly mixed (i.e., not easily 3 In this context, path dependence for an economic process can be corresponding to labeled sectors such as technol- ditional or not). These tools can also be turned thought of as the idea that the equilibrium that is achieved (e.g., price) identifiable) clusters. Connections are made for the (relatively) ogy, energy, and so on. Also striking is that the upside down to generate time series that behave would depend on the manner (process) by which the price comes to be closest cluster centroids.

48 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 49 of understanding of how low-level visual inputs ships among these segments (e.g., spatial adja- Visual can be translated into high-level conceptual de- cency or color similarity) are encoded. scriptions has been termed the “semantic gap.” High-level vision is the process of translating Garrett Kenyon, a theoretical neuroscientist at these lower-level perceptions into the recogni- understanding Los Alamos National Laboratory (LANL), and tion of objects and their conceptual relationships, I, a computer and cognitive scientist, recently yielding a coherent description of the image as a By Melanie Mitchell organized a workshop on this topic at the Santa whole. This information processing is not wholly Fe Institute. The workshop, entitled “High-Level one way: it is generally believed that the different levels of processing continually communicate with one The mind quickly another in both a feed-for- translates an image ward and feedback manner. into abstract mean- Not only does information ing: These pictures from lower level process- of a humpback ing influence higher levels, whale with her calf, a human mother but higher-level processing, and child, and involving stored knowl- a swan and her cyg- edge, can guide lower-level nets easily conjure s e s perception in tasks such as the concept of segmentation. The ability to “mother and child, nterpri E

m s

l fluidly integrate the differ- or children.” ma ent levels is a major source Ani s l m of visual understanding in ma tockphoto.co s i /

s humans and other animals. S / Ani E

nner The talks and discussions a t , JAM tockphoto.co s s TT at the workshop concerned scientific and technological left: © big left: © leo left:WA © enigmas at different levels of visual processing. The first Consider the pictures above. roscientists and psychologists studying how the Perception and Low-Level Vision: Bridging the set of speakers—mainly computational neuro- What concept do all three images brain accomplishes this translation, as well as Semantic Gap,” brought together a small group scientists—described recent research on how the represent? for computer scientists who are trying to build of prominent neuroscientists, psychologists, com- brain performs low-level vision. Bartlett Mel of Of course there are any number of differ- programs that automatically determine semantics puter scientists, and theoretical biologists to dis- the University of Southern California and Ilya ent concepts these images represent, such as from pictures. Although science has uncovered cuss the semantic gap from an interdisciplinary Nemenman from LANL each spoke about the “animals,” “mammals,” “things with limbs,” many of the brain and perceptual mechanisms perspective and to see if a set of common princi- lowest level of the visual system: individual neu- and “multiple objects,” but most people would underlying low-level vision, very little is known ples of visual understanding could be discovered rons. Their common message was that neurons very quickly answer “mothers and babies,” or about how the brain accomplishes higher-level, from the collective knowledge of the participants. are surprisingly complex, both with respect to the something similar, perceiving that more abstract more abstract perception. Similarly, while mod- Visual perception can be roughly categorized information encoding and processing they (and concept to be the intended meaning of this ern computer vision systems have impressive into different levels of information processing. their subunits, such as dendrites) can perform juxtaposition. performance in some specific domains, there are In low-level vision, primitive visual features such and with respect to how information is commu- How does the mind so quickly translate an no systems able to recognize instances of visual as color, contrast, texture, edges, and contours nicated between individual neurons. In light of array of pixels of different light intensities and categories or understand the contents of visual are extracted from the collection of photons that data recently obtained from detailed probes into colors into an abstract meaning? Visual under- scenes with anywhere near the generality and ro- impinge upon the retina. Subsets of the image individual cells, novel brain imaging techniques, standing of this kind is a great mystery for neu- bustness of human perception. Our current lack are identified as separate “segments” and relation- and realistic computer models, classical theories

50 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 51 of understanding of how low-level visual inputs ships among these segments (e.g., spatial adja- Visual can be translated into high-level conceptual de- cency or color similarity) are encoded. scriptions has been termed the “semantic gap.” High-level vision is the process of translating Garrett Kenyon, a theoretical neuroscientist at these lower-level perceptions into the recogni- understanding Los Alamos National Laboratory (LANL), and tion of objects and their conceptual relationships, I, a computer and cognitive scientist, recently yielding a coherent description of the image as a By Melanie Mitchell organized a workshop on this topic at the Santa whole. This information processing is not wholly Fe Institute. The workshop, entitled “High-Level one way: it is generally believed that the different levels of processing continually communicate with one The mind quickly another in both a feed-for- translates an image ward and feedback manner. into abstract mean- Not only does information ing: These pictures from lower level process- of a humpback ing influence higher levels, whale with her calf, a human mother but higher-level processing, and child, and involving stored knowl- a swan and her cyg- edge, can guide lower-level nets easily conjure s e s perception in tasks such as the concept of segmentation. The ability to “mother and child, nterpri E

m s

l fluidly integrate the differ- or children.” ma ent levels is a major source Ani s l m of visual understanding in ma tockphoto.co s i /

s humans and other animals. S / Ani E

50 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 51 in neuroscience are being radically redrawn. Neu- spiking in neural populations; (2) the nature and roscientists are now rethinking paradigms that advantages of distributed and sparse representa- were once generally accepted; results from recent tions, in which sensory information is encoded experimental studies are undermining notions with only a small number of active neurons at any such as the static “receptive fields” of neurons, given point in time; (3) methods for measuring

the transmission of information among neurons statistical and information-theoretic properties of m via the simple counting of spikes, and the role natural visual inputs and inferring their implica- of dendrites as “passive” conductors of electri- tions for how the brain efficiently encodes visual m tockphoto.co s cal signals. As Tom Stoppard wrote in his play information; and (4) the role of feedback connec- i

Arcadia, “It’s the best possible time to be alive, tions from higher brain levels. tockphoto.co s teve geer/ when almost everything you thought you knew s is wrong.” For the theorists in the group, right: © big right: © Collective information processing by groups of the experimentalists’ talks drove home the dif- neurons was a key topic for other participating ficulty of obtaining and interpreting data about on?” He challenged the group with his assertion for something in vision—there are about 10 times Experiments brain scientists, including Garrett Kenyon, Pam the brain, which is perhaps the most complex of that “the fact that information flows all the way as many feedback as feed-forward connections in have shown that Reinagel (University of California, San Diego), all natural systems. The theorists, on the other up and down the reductionist hierarchy…has the visual cortex. Much discussion (some rather people can identify John George (LANL), Fritz Sommer (Redwood hand, underscored the increasing importance of significant implications for the way neuroscience heated) concerned the role of feedback informa- whether or not a Center for Theoretical Neuroscience), and David computer modeling in neuroscience. (and presumably other areas of biological infor- tion and what kinds of experiments could tease picture contains an m Field (Cornell). The question of how such- neu With respect to understanding sensory information processing) should be done.” The group out its function. animal in less that ronal groups represent information is currently mation processing or any other brain process, discussed some examples of information flow to Visual attention is one area where feedback 200 milliseconds.

being hotly debated in neuroscience. Particu- computational neuroscientist Tony Bell (Red- tockphoto.co and from levels as low as genetic regulation, and from higher levels seems essential. Psychologist s i : ©

larly significant controversies that were discussed wood Center for Theoretical Neuroscience) asked, w several in the group strongly disagreed that it is Todd Horowitz, from Harvard Medical School,

concern (1) the existence and role of correlated “What level of description should we be focusing belo important to include such levels when studying addressed the issues of what constitutes visual information processing in the brain. attention and whether it is needed to bridge Higher-level visual behavior was a second focus the semantic gap. In particular, when viewing a of the workshop. How do humans and other scene, what do we pay attention to and how does animals use vision in real life, and what can be that affect our ability to remember the scene later understood from studying human visual behavior on? Horowitz presented experimental results that from a psychophysical and psychological perspec- indicated that people require focused attention tive? Psychophysicist Simon Thorpe, from the to encode the “gist” of visual scenes in terms of Centre de Recherche Cerveau et Cognition in objects present and spatial layout, and in particu- France, described a surprising set of experimental lar, that focusing on the layout is more important results showing that humans are able to do some than focusing on the objects for remembering a complex visual recognition tasks much faster given scene. than had been previously thought. For example, Neuroscientist and psychophysicist Peter König Thorpe’s experiments have shown that most peo- (University of Zurich), zeroed in on a particu- ple can identify whether or not a picture contains lar form of visual attention: the control of eye an animal in less than 200 milliseconds. These movements in response to the information con- results call into question the role of feedback tent of input stimuli and “top-down” feedback. connections in visual recognition, since 200 mil- Computer scientist Dana Ballard (University of liseconds is not enough time for signals to propa- Texas at Austin), presented a computer model Visual perception can be categorized into different levels of information processing: Low-level vision takes in primitive visual gate to higher brain levels and for those higher for “multi-tasking” in visual attention, which his features such as color and texture, while high-level vision translates these perceptions into recognition of objects and their concep- levels to send feedback. However, it seems that group tested by having it navigate in a virtual tual relationships, yielding a coherent description of the whole. the brain clearly must use feedback connections reality simulation.

52 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 53 in neuroscience are being radically redrawn. Neu- spiking in neural populations; (2) the nature and roscientists are now rethinking paradigms that advantages of distributed and sparse representa- were once generally accepted; results from recent tions, in which sensory information is encoded experimental studies are undermining notions with only a small number of active neurons at any such as the static “receptive fields” of neurons, given point in time; (3) methods for measuring

larly significant controversies that were discussed wood Center for Theoretical Neuroscience) asked, w several in the group strongly disagreed that it is Todd Horowitz, from Harvard Medical School,

52 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 53 the rest of the body into account, psychologists studying abstract reasoning cannot ignore the effects of visual perception that “leak” into such reasoning. New approaches to classical problems in computer vision and pattern recognition were proposed by computer scientists Lakshman Prasad and James Theiler, both from LANL, and myself. Ideas from “left field” were presented by theoretical immunologist Tom Kepler, of Duke Uni- versity, who explored possible analogies between the visual system and the immune system, which itself must perform sophisticated, unpredictable, and ongoing tasks of pattern recognition

m and response. I described how my own work on high-level visual pattern recognition and analogy has been inspired by the immune system and tockphoto.co s other complex systems with pattern-recognizing © big abilities. People require focused attention to encode the “gist” of visual scenes. Interestingly, the In the last decade or so, research on both scene’s layout may be more important to memory than the objects themselves. natural and computer vision has become rather narrow and specialized; work on the fundamental Ballard, along with several other participants, problem of how different levels of vision are inte- argued for the importance of studying vision in grated has been largely neglected. In my view, the the context of the rest of the body. He stressed main benefit of the workshop was the opportu- that the visual system and motor system are nity for scientists studying widely diverse aspects tightly interconnected, and constrain one an- of vision to communicate to one another the other so as to make the computation of behavior recent major advances and controversies in their tractable. Psychologist Shimon Edelman, of Cor- areas. On the one hand, learning more about nell University, emphasized that vision scientists, the daunting complexity of the visual system in in modeling abilities such as object recognition, processing sensory data, juxtaposed with the ex- should not lose the phenomenological aspect of traordinary abilities of the mind to glean abstract vision—that is, visual phenomena that do not meaning, only made the gap seem wider. On the involve categorization and recognition, including other hand, the talks and discussions at the work- our first-person experience of what it feels like to shop gave all the participants a new appreciation “see.” Psychologist Rob Goldstone, from Indiana for the scope of the problem and its requirement University, described a different notion of em- for interdisciplinary collaboration, a sense of bodiment: the effects of visual perception on rea- where the most promising directions are, and (at soning and abstract problem solving. Goldstone least for me) plans for new collaborations and reported on experimental results that indicated many fresh ideas to take home and ponder. t that even the most abstract of tasks—mathematical reasoning—is affected by visual input, such as Melanie Mitchell is Professor of Computer Science the layout of the problem on the page. His point at Portland State University, and External Profes- was that, just as vision scientists need to take sor at the Santa Fe Institute.

54 Santa Fe Institute Bulletin SPRING 2008 the Evolution of Individuality and Sex

by don monroe s, Inc r che r cett / Photo Resea w © Don W. © Fa DonW.

“As far as we know, life began once, four the more striking when contrasted SEM (Scanning and a half billion years ago, and with our frail and mortal bodies. In Electron Microscope) since then has been passed down fact, eliminating inferior individuals of a single sperm like a family heirloom from parent is crucial to the biological evolution penetrating an egg at to offspring,” says Richard Michod, that drives new innovations. As an the point of fertiliza- head of the Department of Ecol- evolutionary biologist, Michod has tion: Researchers are ogy and Evolutionary Biology at the explored this process both theoreti- finding that sex has University of Arizona. Along the cally and experimentally. important biological way, life has repeatedly devised At a September SFI Business purposes beyond novel ways to organize and repro- Network meeting in Washington, procreation. duce itself as it evolved from primi- D.C., Michod spoke about “Coop- tive self-replicating molecules to eration and Conflict during the Evo- complex societies of multicellular lution of Individuality and Sex.” The organisms. talk came within the larger theme The persistence and enhance- of the meeting: “Conflict, Coop- ment of our genetic legacy is all eration and Creativity in Complex

Santa Fe Institute Bulletin SPRING 2008 55 Systems.” Organized by SFI Research some of which catalyze the replication fates of its constituents, Michod said, to devote resources to every aspect of those that form larger Fellow Jessica Flack, and William of DNA, but in the beginning there “it enforces cooperation on them,” survival, and addressing one task can colonies—display this Glenney, Chief of the Naval Opera- was probably only a single molecule. because cells with selfish mutants impede another. altruistic behavior. tions Strategic Studies Group, the The prime suspect is RNA, which can do worse than cells with cooperative “There are often tradeoffs between The researchers meeting focused on identifying the both replicate its internal structure genes. “The cell is the perfect example the effects of a trait on different also identified the types of conflict that lead to creative and act as a catalyst. of a conflict mediator.” necessary components of fitness,” molecular mecha- new solutions. Simply replicating is not enough, Michod only briefly mentioned Michod observed. In the alga Volvox nism by which Volvox Michod described a continuing however, because some errors always another milestone, the appearance of carteri, for example, cells cannot cells specialize. They cycle—more accurately a spiral—of happen during copying. Some three eukaryotic cells that comprise all mul- reproduce while they maintain the found that the ge- cooperation leading to new types of decades ago, theoretical biologists ticellular organisms as well as many hair-like flagella that let them move netic “switch” that conflict, which in turn lead to higher suggested that cooperative networks single-celled organisms. Biologists in response to their environment. An determines whether a levels of cooperation and the emer- of genes (presumably RNA at first) widely agree that the DNA-contain- individual cell cannot benefit from cell becomes a germ gence of novel traits. “Basic proper- that promote each other’s production ing mitochondria—the structures in having flagella without sacrificing -re or soma cell is adapt- ties of the living world emerge out could keep errors from accumulating. animal cells that provide energy— production. ed from a mechanism of this cycle of cooperation and con- As long as the error rate is below some and chloroplasts—which perform “Division of labor in the group is that already existed in flict,” he said. threshold, Michod said, such net- photosynthesis in plants—were once a way to break through this tradeoff single-cell algae. That He described several stages along works, known as heterocycles, “main- separate species. Even the nucleus, that governs the life of single cells,” mechanism—like hoto.com this spiral, starting with the innova- tain themselves stably through time.” the defining feature of eukaryotic Michod explained. In fact, many Vol- the calorie-restriction p stock tion of the cell. Drawing on his own Although free-floating genes can cells, may have arisen from such in- vox species form colonies with a few response that extends i ta / ta

experiments in colony-forming algae, cooperate in this way, “it’s a poorly terspecies cooperation. germ cells dedicated to reproduction life in many animal ri while other “soma” cells retain the fla- species—acts to delay ea lau

gella that let the entire colony move. reproduction in favor dr Although specialization clearly ben- of extended survival © an Michod described a continuing cycle—more accurately a spiral—of cooperation efits the colony as a whole, Michod when resources are Sex has an evolutionary cost, as evidenced in the extravagant display leading to new types of conflict, which in turn lead to higher levels of cooperation pointed out, the soma cells must act limited. of the male peacock and the emergence of novel traits. altruistically, because they forgo the There are limits chance to reproduce themselves. “Al- to this strategy for resolving conflict tied to reproduction. But other spe- truism is widely appreciated to be the between cells, Michod noted. Since cies, including large animals like some central problem of social behavior, the soma cells do not divide at all, lizards, reproduce asexually, while he clarified the conflicts leading to defined life, these cycles of genes,” Altruistic Cells and it’s also fundamental to the evo- they can never go on to make larger other species exchange genetic mate- multicellular life and to sexual repro- Michod observed. “Cheating was In the laboratory, Michod recapitu- lutionary transition to multicellular- structures. The way in which conflicts rial independently of reproduction, duction. This cycle of innovation illu- rampant.” A gene that undergoes a lates the more recent development of ity. It takes fitness from one level and are mediated, he noted, “can have ef- Michod noted. “The idea that sex is minates not only the nature of life on “selfish” mutation, so that it propa- multicellularity by studying colony- gives it to another, from the cell to fects on the future evolvability of the necessary for reproduction doesn’t Earth, but more general principles of gates at the expense of its partners, forming algae. Poised at the cross- the colony.” lineage.” In contrast, our own cells hold up.” how conflict and cooperation can fos- will eventually take over. “Pretty soon roads between a single-cell and multi- One evolutionary explanation of continue to divide. Even when they From an evolutionary perspective, ter innovative approaches to problems. there’s only taking, no giving, and the cellular lifestyle, these primitive plants altruism, known for decades, is based are not destined to become reproduc- Michod said, “the most obvious thing whole cycle goes extinct.” illustrate the conflicts that push life on kin selection: it makes sense for tive cells, they continue to specialize, about sex is its cost.” Extravagant dis- Cells as Conflict Mediators Faced with this conflict between its one way or the other. an individual to sacrifice its life for forming the various complex tissues plays like those of the male peacock In the primordial chemical soup, constituents, Michod said, “there’s a “A single cell has immediate and others, as long as those others share in our bodies. appear to be unnecessary and waste- Michod said, “the first individuals need for some kind of conflict me- effective interaction with its environ- enough of its genetic heritage. The ful, but help meet the challenge of were things like molecules and genes diation.” The membrane-enclosed ment,” Michod said, making it easy genetically identical algae cells clearly Sex: Evolution’s Raw Material finding a choosy mate. Sex also expos- replicating through some kind of cell solved this problem. “What the to get resources and get rid of wastes. meet this requirement. However, Finally, Michod discussed the evolu- es the body to the risk of infection. prebiotic chemistry.” Current organ- cell does is align the interests of all “The problem is simply that a single Michod’s team is still clarifying why tionary role of sexual reproduction. In “What are the benefits that offset isms use DNA to specify proteins, the members inside.” By linking the cell has to do everything.” A cell has only some Volvox species—generally many species, such as humans, sex is all these costs?” Michod asked.

56 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 56 Santa Fe Institute Bulletin SPRING 2008 57 Systems.” Organized by SFI Research some of which catalyze the replication fates of its constituents, Michod said, to devote resources to every aspect of those that form larger Fellow Jessica Flack, and William of DNA, but in the beginning there “it enforces cooperation on them,” survival, and addressing one task can colonies—display this Glenney, Chief of the Naval Opera- was probably only a single molecule. because cells with selfish mutants impede another. altruistic behavior. tions Strategic Studies Group, the The prime suspect is RNA, which can do worse than cells with cooperative “There are often tradeoffs between The researchers meeting focused on identifying the both replicate its internal structure genes. “The cell is the perfect example the effects of a trait on different also identified the types of conflict that lead to creative and act as a catalyst. of a conflict mediator.” necessary components of fitness,” molecular mecha- new solutions. Simply replicating is not enough, Michod only briefly mentioned Michod observed. In the alga Volvox nism by which Volvox Michod described a continuing however, because some errors always another milestone, the appearance of carteri, for example, cells cannot cells specialize. They cycle—more accurately a spiral—of happen during copying. Some three eukaryotic cells that comprise all mul- reproduce while they maintain the found that the ge- cooperation leading to new types of decades ago, theoretical biologists ticellular organisms as well as many hair-like flagella that let them move netic “switch” that conflict, which in turn lead to higher suggested that cooperative networks single-celled organisms. Biologists in response to their environment. An determines whether a levels of cooperation and the emer- of genes (presumably RNA at first) widely agree that the DNA-contain- individual cell cannot benefit from cell becomes a germ gence of novel traits. “Basic proper- that promote each other’s production ing mitochondria—the structures in having flagella without sacrificing -re or soma cell is adapt- ties of the living world emerge out could keep errors from accumulating. animal cells that provide energy— production. ed from a mechanism of this cycle of cooperation and con- As long as the error rate is below some and chloroplasts—which perform “Division of labor in the group is that already existed in flict,” he said. threshold, Michod said, such net- photosynthesis in plants—were once a way to break through this tradeoff single-cell algae. That He described several stages along works, known as heterocycles, “main- separate species. Even the nucleus, that governs the life of single cells,” mechanism—like hoto.com this spiral, starting with the innova- tain themselves stably through time.” the defining feature of eukaryotic Michod explained. In fact, many Vol- the calorie-restriction p stock tion of the cell. Drawing on his own Although free-floating genes can cells, may have arisen from such in- vox species form colonies with a few response that extends i ta / ta

56 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 56 Santa Fe Institute Bulletin SPRING 2008 57 “Sex repairs DNA,” he said, helping different arenas: policing of conflict individuals or species that use inferior of both atomic and thermonuclear to explain how “the cells in our body by dominant chimpanzees, elaborate strategies are killed off. Hopefully, weapons. Richard Rhodes, author of get old and die, but our germ line election procedures in Renaissance directly adopting the best practices of the Pulitzer Prize winning The Mak- goes on forever.” Venice, and co-option of “jumping others can help organizations avoid ing of the Atomic Bomb and many Further insight into sex comes genes” to build the adaptive immune this fate. other books, contrasted these two from species that can reproduce ei- system. In each case, she said, “either The other speakers at the meeting mega-projects, saying that, although ther sexually or asexually. There is the arms race between the compo- offered highly varied insights into the external conflict was a primary driver general agreement, Michod said, that nents or the implementation of the roles of conflict at the social level. for both, conflict within the project “in such organisms sex is induced by robustness mechanisms, resulted in Dean Simonton, Distinguished Pro- was much more destructive for the stress.” He suggested that the stress new problems, which in turn gener- fessor and Vice Chair of the Depart- hydrogen bomb. Aaron Clauset, a acts by affecting the balance between ated the evolution of new solutions ment of Psychology at the University postdoctoral fellow at SFI, discussed the oxidation and reduction reactions to maintain stability. So we get this of California at Davis, for example, more modern conflicts, analyzing that fuel life. “Stress universally upsets ratcheting up of complexity in all of extended evolutionary ideas to indi- patterns in terrorist attacks. He con- this balance,” he said, resulting in a these systems.” vidual creativity in two distinct ways. cluded that although the attacks of buildup of reactive oxygen species In the context of the whole meet- Creative geniuses are often misfits September 11, 2001 were unusually that can damage DNA. “Sex is a way ing, Michod’s research provides who do not reproduce, he said, so large, they are best seen as part of the of coping with this damage,” Michod intriguing examples of innovative their repeated appearance in the “long tail” of a power-law distribution commented, because it repairs DNA. mechanisms for mediating conflicts, population demands an evolution- of event sizes, rather than as histori- Indeed, he and collaborator Aurora which could also apply to human sit- ary explanation, perhaps like those cally unique events. Nedelcu of the University of New uations. Good managers, for example, that Michod described for altruism. The tension between conflict and ce

r Brunswick found that exposing algae well know the value of aligning the Simonton also described “secondary cooperation leads to innovation in

esou to antioxidants prevents their sexual interests of individuals with those of Darwinism,” the social and personal a variety of situations, and research- r t r response to heat stress. the whole team, as the cell does for its influences that encourage susceptible ers are still struggling to describe the Like other evolutionary changes, genetic networks. Similarly, division individuals to generate the wide-rang- many ways this happens. One day, onaux / a i

at sex helps to solve problems intro- of labor was applied in manufacturing ing ideas that underlie creativity. perhaps, they may be able to system- N duced by earlier innovations, but long before it was recognized in Vol- Uniquely creative individuals were atically analyze and even predict this usÈes M t

es it introduces new problems. These vox. Enlightened managers have also clearly critical in the development evolution. —Don Monroe d on i include biological and behavioral encouraged (non-sexual) “cross-fertil- manipulation of the mating process, ization” between teams to stimulate © RÈun such as genes that limit the resources new ways of thinking. Curiously, creative people, such as Isaac Newton (1642–1727), are often misfits who don’t reproduce, that males devote to offspring. “There Still, there are risks to using evo- raising questions about their repeated appearance in the population. are all kinds of conflicts that are set lutionary conflicts as a metaphor up because of sex.” Michod observed. for human behavior. For one thing, “Biologists differ strongly on what Sex also corrects genetic damage, “You solve one set of conflicts but biological evolution acts only through they think are the major benefits of Michod said. Damage to the DNA then create other sets of conflicts. It’s the persistence of genetic changes in sex, and there’s no real agreement of one partner can be repaired by the raw material that leads to contin- offspring, while cultural evolution on the whole problem.” They agree, undamaged DNA from the other ued evolution.” transmits new ideas rapidly between s, Inc however, that the mixture of traits partner. In this way, sex recovers an unrelated individuals and organiza- r che between individuals provides raw ma- error-correcting feature that had been Harnessing Social Conflict tions. Perhaps more sobering, bio- r terial for evolutionary change. present in early genetic networks. Other presenters at the meeting logical evolution works only because Like altruism, this explanation in- This feature was lost somewhat when identified the types of conflict that a / Photo Resea di volves a transfer of fitness between life began forming cells, because lead to creative new solutions. Flack Volvox alga, a kind of green algae, reproduces e M levels, in this case between individuals “damages and mutations get trapped set the stage by describing conflict- both asexually and sexually. Such behavior raises ology i and their species. on the inside,” Michod observed. mediation mechanisms in three very questions about the purpose of sex. B

58 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 59 “Sex repairs DNA,” he said, helping different arenas: policing of conflict individuals or species that use inferior of both atomic and thermonuclear to explain how “the cells in our body by dominant chimpanzees, elaborate strategies are killed off. Hopefully, weapons. Richard Rhodes, author of get old and die, but our germ line election procedures in Renaissance directly adopting the best practices of the Pulitzer Prize winning The Mak- goes on forever.” Venice, and co-option of “jumping others can help organizations avoid ing of the Atomic Bomb and many Further insight into sex comes genes” to build the adaptive immune this fate. other books, contrasted these two from species that can reproduce ei- system. In each case, she said, “either The other speakers at the meeting mega-projects, saying that, although ther sexually or asexually. There is the arms race between the compo- offered highly varied insights into the external conflict was a primary driver general agreement, Michod said, that nents or the implementation of the roles of conflict at the social level. for both, conflict within the project “in such organisms sex is induced by robustness mechanisms, resulted in Dean Simonton, Distinguished Pro- was much more destructive for the stress.” He suggested that the stress new problems, which in turn gener- fessor and Vice Chair of the Depart- hydrogen bomb. Aaron Clauset, a acts by affecting the balance between ated the evolution of new solutions ment of Psychology at the University postdoctoral fellow at SFI, discussed the oxidation and reduction reactions to maintain stability. So we get this of California at Davis, for example, more modern conflicts, analyzing that fuel life. “Stress universally upsets ratcheting up of complexity in all of extended evolutionary ideas to indi- patterns in terrorist attacks. He con- this balance,” he said, resulting in a these systems.” vidual creativity in two distinct ways. cluded that although the attacks of buildup of reactive oxygen species In the context of the whole meet- Creative geniuses are often misfits September 11, 2001 were unusually that can damage DNA. “Sex is a way ing, Michod’s research provides who do not reproduce, he said, so large, they are best seen as part of the of coping with this damage,” Michod intriguing examples of innovative their repeated appearance in the “long tail” of a power-law distribution commented, because it repairs DNA. mechanisms for mediating conflicts, population demands an evolution- of event sizes, rather than as histori- Indeed, he and collaborator Aurora which could also apply to human sit- ary explanation, perhaps like those cally unique events. Nedelcu of the University of New uations. Good managers, for example, that Michod described for altruism. The tension between conflict and ce

r Brunswick found that exposing algae well know the value of aligning the Simonton also described “secondary cooperation leads to innovation in

58 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 59 The Conflicted Brain By Jon Wilkins

When I played third base in Little tic and evolutionary conflicts within our brains. League, ground balls used to come at me fast. Brain-imaging studies suggest that different Invariably, once a year, one would hit me in the brain regions come into conflict with each other eye, or nose, or mouth. In those cases, I would over certain decisions. At the same time, many be positioned in front of the ball, but then, at the genes that are expressed in the brain show evi- last moment, I’d flinch and look away. The ball dence of having been in a long-term evolutionary would then bounce up and hit me in the face. conflict with each other. It is possible that when The problem was that different parts of my we feel as if we are of two minds, it is precisely body were trying to achieve two different goals. because different sets of our own genes have My legs placed my body in front of the bounc- effective control over different regions of our ing ball—with the goal of catching it. My face, brains, and these different brain regions are exert- on the other hand, was trying to get out of the ing antagonistic influences on the decision-mak- way. If my entire body had committed to getting ing process. away from the ball, I could simply have moved The idea of conflict in the mind is old. What is aside and let the ball through. Better yet, if my new is that we are beginning to understand some entire body had committed to catching the ball, of the mechanisms through which these conflicts I would have kept watching it, and would have play. This new understanding undermines our had my glove in the right position to catch it; preconceptions about human intelligence and keeping my eye on the ball metaphorically would our notions of the “self.” In fact, it may turn out have saved me from doing it literally. to be misleading to talk about the notion of indi- Most people recognize this type of conflict: the viduals having a single “self” at all. feeling of being pulled internally in two directions. It’s the tiny angel and tiny devil sitting on Battle of the Neurons our shoulders, urging us to take disparate ac- On the basis of recent neuroeconomic experi- tions—we want to lose weight, but we also want ments, researchers have suggested that difficult to eat that donut. Recent research suggests that decisions might be made by a type of “voting” this sensation may have a basis in real mechanis- process within the brain. Typically, an experimen-

60 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 61 The Conflicted Brain By Jon Wilkins

60 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 61 tal subject performs some behavioral task while these two responses are localized in distinct sites that favoring B, choice A is fol- his or her neural activity is monitored. This type within the brain. It is as if one set of neural cir- lowed, and vice versa. of procedure has been used for decades to identi- cuits is screaming out, “You must act! The lives This is like the principle Medial Parietal Lobe fy regions of the brain that are particularly active of the five people on the track outweigh a single upon which the Anglo-Ameri- Frontal Gyrus during specific cognitive or behavioral tasks. life!” while another set of circuits screams, “It is can criminal justice system is STRIATUM Anterior One recent set of studies focuses on decision- wrong to kill this man! The ends do not justify based. In principle, the goal is Cingulate Cortex Posterior making in difficult moral dilemmas. Subjects face the means!” The decision that is ultimately made to reliably determine guilt or Cingulate Gyrus a hypothetical situation: a train is rushing toward depends on which of the two sets of circuits innocence on the basis of avail- five people who are tied to the track. The only screams louder. able evidence. The mechanism way to save those five people is to kill a different by which we attempt to reach Hippocampus person (for example, by throwing him or her in Why can’t we all just get along? an unbiased verdict, however, Hypothalamus front of the train). On the one hand, people sense Why might decision-making in the brain be is an antagonistic interaction that five lives are more valuable than one. On structured like this? Is decision-making through between biased advocates. One stockphoto.com the other hand, there is something morally and competition adaptive, or is it a maladaptive party is charged with gather- i emotionally distressing about the idea of actively byproduct of an evolutionary process? Most ing and presenting all of the akobchuk/ participating in the death of the one person. evolutionary explanations fall into one of two evidence that the defendant Y In the study, the subjects show increased activ- categories. These explanations assume (often im- is guilty. Another party col- ity in a region of the brain called the anterior cin- plicitly) either (1) that this conflict is an adapta- lects the evidence that would tion, or (2) that it is a historical artifact. While exonerate the defendant. A brain scan:Vasiliy © This new understanding undermines our preconcep- there may well be some truth to either or both third entity—the judge and/or of these assumptions, both are rooted in a naïve jury—assesses which of the two has presented a regions that have evolved more recently, regions tions about human intelligence and our notions of White text understanding of natural selection that fails to more compelling case and rules accordingly. with their own set of heuristic rules. Sometimes, indicates locations the “self.” In fact, it may turn out to be misleading to capture the nuances of the evolutionary process. The second common explanation is that this the old and new rules contradict each other. In of brain regions After briefly explaining these two simple classes conflict is an artifact of the evolutionary history this case, the conflict is not conceived as adaptive, talk about individuals having a single “self” at all. mentioned in the of explanation, I will introduce a third possible of our brains. Our brains evolved through modi- but rather as an unfortunate limitation resulting neuroeconomics explanation, one based on recent advances in fication of an earlier primate brain, which was from the historical path followed by evolution. studies. Orange gulate cortex, associated with conflict resolution. molecular biology and evolutionary genetics: derived from an earlier mammalian brain, and Now I’ll suggest a third possible class of ex- text locates the Subjects also show increased neural activity in that this apparent neural conflict reflects a genu- so on. The human brain is necessarily descended planation—that the conflict is in some sense human structures other regions. Some of these regions (such as the ine evolutionary genetic conflict. from a long line of brains, each of which had to real. What I mean is that the conflict represents corresponding to medial frontal gyrus and the posterior cingulate The first class of explanation assumes that com- function well enough in its own environment to the direct outcome of natural selection, but that the mouse brain gyrus) are associated with the emotional response petition is a powerful and robust way to make allow its bearer to survive and reproduce. Thus we natural selection is acting differently on different regions mentioned to the situation. Other regions (including the choices in a noisy and complicated world. Imag- have inherited a neural architecture that evolved parts of the brain. The apparent conflicts -be in the genetic chimera experi- parietal lobe) are associated with higher cognitive ine that you face two choices, A and B. The brain in a very different context. It may simply be that tween regions of the brain are a manifestation of ments. functions—the “reasoning” part of the brain. In- receives a constant stream of information from when a complex organ is constructed in this way, an underlying genetic conflict. terestingly, the subjects who say that they would the environment, most of which is irrelevant to by layering new functions on old, certain conflicts sacrifice the one to save the five show relatively the decision at hand. It must collect and evaluate and incompatibilities will inevitably arise. Imprinted Genes higher levels of activity in this “rational” part of the relevant information and follow the better of This second class is being invoked when re- We now understand that a conflict exists between the brain. The subjects who would refuse to act the two choices. One possible solution is to es- searchers talk about conflicts between the “reptil- the maternally and paternally inherited sets of show relatively more activity in the more “emo- tablish one apparatus that filters this stream and ian” and “mammalian” or the “emotional” and genes within each of us. About one percent of tional” regions of the brain. gathers all the evidence favoring choice A. A sec- “cognitive” parts of the brain. In this scenario, genes in mammals (including humans) are sub- We are a long way from a complete under- ond apparatus would be dedicated to collecting our ancestors had a brain with certain heuristic ject to genomic imprinting. In the case of these standing of this type of decision process. How- evidence favoring B. Each would then produce rules that it used to navigate the world. Our imprinted genes, the gene copy, or allele, that ever, one possible interpretation is that the situ- a signal proportional to the weight of this evi- modern brains contain regions that are homolo- came from your father functions differently from ation elicits two conflicting responses, and that dence. If the signal favoring choice A outweighs gous to those ancestral brains. We also have other the allele that came from your mother, even if the

62 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 63 tal subject performs some behavioral task while these two responses are localized in distinct sites that favoring B, choice A is fol- his or her neural activity is monitored. This type within the brain. It is as if one set of neural cir- lowed, and vice versa. of procedure has been used for decades to identi- cuits is screaming out, “You must act! The lives This is like the principle Medial Parietal Lobe fy regions of the brain that are particularly active of the five people on the track outweigh a single upon which the Anglo-Ameri- Frontal Gyrus during specific cognitive or behavioral tasks. life!” while another set of circuits screams, “It is can criminal justice system is STRIATUM Anterior One recent set of studies focuses on decision- wrong to kill this man! The ends do not justify based. In principle, the goal is Cingulate Cortex Posterior making in difficult moral dilemmas. Subjects face the means!” The decision that is ultimately made to reliably determine guilt or Cingulate Gyrus a hypothetical situation: a train is rushing toward depends on which of the two sets of circuits innocence on the basis of avail- five people who are tied to the track. The only screams louder. able evidence. The mechanism way to save those five people is to kill a different by which we attempt to reach Hippocampus person (for example, by throwing him or her in Why can’t we all just get along? an unbiased verdict, however, Hypothalamus front of the train). On the one hand, people sense Why might decision-making in the brain be is an antagonistic interaction that five lives are more valuable than one. On structured like this? Is decision-making through between biased advocates. One stockphoto.com the other hand, there is something morally and competition adaptive, or is it a maladaptive party is charged with gather- i emotionally distressing about the idea of actively byproduct of an evolutionary process? Most ing and presenting all of the akobchuk/ participating in the death of the one person. evolutionary explanations fall into one of two evidence that the defendant Y In the study, the subjects show increased activ- categories. These explanations assume (often im- is guilty. Another party col- ity in a region of the brain called the anterior cin- plicitly) either (1) that this conflict is an adapta- lects the evidence that would tion, or (2) that it is a historical artifact. While exonerate the defendant. A brain scan:Vasiliy © This new understanding undermines our preconcep- there may well be some truth to either or both third entity—the judge and/or of these assumptions, both are rooted in a naïve jury—assesses which of the two has presented a regions that have evolved more recently, regions tions about human intelligence and our notions of White text understanding of natural selection that fails to more compelling case and rules accordingly. with their own set of heuristic rules. Sometimes, indicates locations the “self.” In fact, it may turn out to be misleading to capture the nuances of the evolutionary process. The second common explanation is that this the old and new rules contradict each other. In of brain regions After briefly explaining these two simple classes conflict is an artifact of the evolutionary history this case, the conflict is not conceived as adaptive, talk about individuals having a single “self” at all. mentioned in the of explanation, I will introduce a third possible of our brains. Our brains evolved through modi- but rather as an unfortunate limitation resulting neuroeconomics explanation, one based on recent advances in fication of an earlier primate brain, which was from the historical path followed by evolution. studies. Orange gulate cortex, associated with conflict resolution. molecular biology and evolutionary genetics: derived from an earlier mammalian brain, and Now I’ll suggest a third possible class of ex- text locates the Subjects also show increased neural activity in that this apparent neural conflict reflects a genu- so on. The human brain is necessarily descended planation—that the conflict is in some sense human structures other regions. Some of these regions (such as the ine evolutionary genetic conflict. from a long line of brains, each of which had to real. What I mean is that the conflict represents corresponding to medial frontal gyrus and the posterior cingulate The first class of explanation assumes that com- function well enough in its own environment to the direct outcome of natural selection, but that the mouse brain gyrus) are associated with the emotional response petition is a powerful and robust way to make allow its bearer to survive and reproduce. Thus we natural selection is acting differently on different regions mentioned to the situation. Other regions (including the choices in a noisy and complicated world. Imag- have inherited a neural architecture that evolved parts of the brain. The apparent conflicts -be in the genetic chimera experi- parietal lobe) are associated with higher cognitive ine that you face two choices, A and B. The brain in a very different context. It may simply be that tween regions of the brain are a manifestation of ments. functions—the “reasoning” part of the brain. In- receives a constant stream of information from when a complex organ is constructed in this way, an underlying genetic conflict. terestingly, the subjects who say that they would the environment, most of which is irrelevant to by layering new functions on old, certain conflicts sacrifice the one to save the five show relatively the decision at hand. It must collect and evaluate and incompatibilities will inevitably arise. Imprinted Genes higher levels of activity in this “rational” part of the relevant information and follow the better of This second class is being invoked when re- We now understand that a conflict exists between the brain. The subjects who would refuse to act the two choices. One possible solution is to es- searchers talk about conflicts between the “reptil- the maternally and paternally inherited sets of show relatively more activity in the more “emo- tablish one apparatus that filters this stream and ian” and “mammalian” or the “emotional” and genes within each of us. About one percent of tional” regions of the brain. gathers all the evidence favoring choice A. A sec- “cognitive” parts of the brain. In this scenario, genes in mammals (including humans) are sub- We are a long way from a complete under- ond apparatus would be dedicated to collecting our ancestors had a brain with certain heuristic ject to genomic imprinting. In the case of these standing of this type of decision process. How- evidence favoring B. Each would then produce rules that it used to navigate the world. Our imprinted genes, the gene copy, or allele, that ever, one possible interpretation is that the situ- a signal proportional to the weight of this evi- modern brains contain regions that are homolo- came from your father functions differently from ation elicits two conflicting responses, and that dence. If the signal favoring choice A outweighs gous to those ancestral brains. We also have other the allele that came from your mother, even if the

62 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 63 conflict is the fact that natural selection acts to neocortex. These regions (evidently favored by First, we expect to find an escalatory “arms race” maximize the number of allele copies that are maternally inherited alleles) are involved in many between different brain regions. If maternally de- passed on to future generations. These copies can higher cognitive functions, such as planning and rived genes are expanding one region to bias deci- be passed on directly through the survival and problem solving. sions in a particular way, paternally derived genes reproduction of the individual organism carrying In the other type of chimera, the androgenetic will counter by augmenting regions with the op- those genes, or through the reproductive success cells are overrepresented in different regions, in- posite effect. Eventually, we might expect this to of related individuals who carry an identical copy cluding the mediobasal forebrain and the hypo- produce an increase in overall brain size. In fact, of the allele. Since the maternally and paternally thalamus. The regions favored by these paternally this may have happened: over the past hundred derived alleles are related to different individuals inherited alleles participate in behaviors such as million years, the size of the mammalian brain has (e.g., cousins on your mother’s side versus those food seeking, mating, social aggression, and the increased disproportionately relative to body size. on your father’s), the strategy that maximizes the expression of emotions. We also expect this conflict to increase fragil- number of copies passed down differs between Another set of observations concerns a genetic ity. Different sets of genes are pulling hard in the two alleles. disorder called Turner syndrome. Girls with opposite directions; a mutation in any one of Turner syndrome inherit only one copy of the these genes can result in a dramatic shift in the Parental Conflicts in the Brain X chromosome (as opposed to the normal two). system. Imprinted genes have been linked to Our understanding of imprinted genes in the These girls suffer from a variety of problems, and many human behavioral dysfunctions, including nc I brain is still in its infancy. However, observations generally perform less well than chromosomally schizophrenia, ADHD, autism, and bipolar dis- involving genetic chimeras (made up of two sets normal individuals on various cognitive tasks. order. These disorders may be much more severe of cells with different genomes) suggest that our Relevant here is that there are significant differ- esearchers, R maternally and paternally derived genes are in ences in the extent to which different cognitive It is perhaps time to stop thinking of the human brain hoto P conflict over how large particular brain regions skills are impaired, depending on the parental should be. origin of the single X chromosome. If the X as evolution’s crowning achievement and the physical

annah Gal / Normal mammalian development requires chromosome is paternally, rather than maternally,

H embodiment of the “self.” Rather, our brains are © the presence of both maternally and paternally inherited, the girls exhibited better verbal ability, casualties of millions of years of internal conflict. two alleles have identical DNA sequences. These derived complements of alleles. In mice, it is pos- social cognition, and behavioral inhibition. This DNA functional differences are the result ofepigenetic sible to make gynogenetic embryos (which contain These are only two of the experiments that sug- autoradiogram superimposed over (meaning “on top of the genes”) modifications to two maternally derived sets of genes) or andro- gest that imprinted genes affect the development and/or more common than they would be in the a child’s face shows the DNA that are established in the germ line, genetic embryos (which contain two paternally of different brain regions, and the development absence of genomic imprinting. a pattern of bands and propagated through developments. derived sets of genes) by removing the nucleus of different sets of skills. Other experiments sug- As humans, we routinely engage in a wide vari- produced during During spermatogenesis (the production of (containing the DNA) from one cell and inject- gest conflicts over how much care to provide ety of self-destructive behaviors. We cheat on our DNA analysis. sperm), imprinted genes are marked with paring it into another. These uniparental embryos for offspring and how to value risk. If the neu- diets. We don’t exercise. We smoke and gamble It may represent ticular chemical modifications. Duringoogenesis fail to develop beyond the first few rounds of cell roeconomists are right, some of our decisions and get addicted to a wide range of substances. genetic individuali- (the production of eggs), these same genes are division, since they have inherited two inactive may be determined by comparing the relative It is perhaps time to stop thinking of the human ty, paternity testing, marked with different modifications. The- off copies of many imprinted genes (and are receiv- intensity of activity in two or more regions of the brain as evolution’s crowning achievement and or biometric identi- spring therefore inherits two functionally dif- ing double the normal dose from the rest). brain. Imprinted genes might effectively tip the the physical embodiment of the “self.” Rather, fication. ferent alleles. When this individual grows into However, it is possible to combine these cells scales to favor one type of decision over another our brains are casualties of millions of years of in- adulthood and has offspring of its own, these with cells from a normal, biparental embryo. by influencing the relative size of these different ternal conflict. Every decision we make is argued marks will be erased and reset on the alleles that These mixed, or chimeric, embryos develop rela- brain regions. out by at least two distinct evolutionary “selves.” it passes on to the next generation. tively normally. The inclusion of gynogenetic and We may eventually discover that multiple per- Why would a complicated mechanism like this androgenetic cells appear to have complementary What Does it All Mean? sonality disorder is simply the most extreme exist? The most prominent explanation suggests effects on the development of the brain. When a Based on the type of effects we see in other sys- manifestation of a dynamic that governs even the that genomic imprinting is the outcome of an chimera contains gynogenetic cells, those cells are tems affected by genomic imprinting, we can most mundane behaviors in each of us. t evolutionary conflict between these maternally overrepresented in particular brain regions, in- speculate about how this conflict might have af- and paternally derived alleles. The key to this cluding the striatum, the hippocampus, and the fected our brains. Jon Wilkins is an SFI professor.

64 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 65 conflict is the fact that natural selection acts to neocortex. These regions (evidently favored by First, we expect to find an escalatory “arms race” maximize the number of allele copies that are maternally inherited alleles) are involved in many between different brain regions. If maternally de- passed on to future generations. These copies can higher cognitive functions, such as planning and rived genes are expanding one region to bias deci- be passed on directly through the survival and problem solving. sions in a particular way, paternally derived genes reproduction of the individual organism carrying In the other type of chimera, the androgenetic will counter by augmenting regions with the op- those genes, or through the reproductive success cells are overrepresented in different regions, in- posite effect. Eventually, we might expect this to of related individuals who carry an identical copy cluding the mediobasal forebrain and the hypo- produce an increase in overall brain size. In fact, of the allele. Since the maternally and paternally thalamus. The regions favored by these paternally this may have happened: over the past hundred derived alleles are related to different individuals inherited alleles participate in behaviors such as million years, the size of the mammalian brain has (e.g., cousins on your mother’s side versus those food seeking, mating, social aggression, and the increased disproportionately relative to body size. on your father’s), the strategy that maximizes the expression of emotions. We also expect this conflict to increase fragil- number of copies passed down differs between Another set of observations concerns a genetic ity. Different sets of genes are pulling hard in the two alleles. disorder called Turner syndrome. Girls with opposite directions; a mutation in any one of Turner syndrome inherit only one copy of the these genes can result in a dramatic shift in the Parental Conflicts in the Brain X chromosome (as opposed to the normal two). system. Imprinted genes have been linked to Our understanding of imprinted genes in the These girls suffer from a variety of problems, and many human behavioral dysfunctions, including nc I brain is still in its infancy. However, observations generally perform less well than chromosomally schizophrenia, ADHD, autism, and bipolar dis- involving genetic chimeras (made up of two sets normal individuals on various cognitive tasks. order. These disorders may be much more severe of cells with different genomes) suggest that our Relevant here is that there are significant differ- esearchers, R maternally and paternally derived genes are in ences in the extent to which different cognitive It is perhaps time to stop thinking of the human brain hoto P conflict over how large particular brain regions skills are impaired, depending on the parental should be. origin of the single X chromosome. If the X as evolution’s crowning achievement and the physical

annah Gal / Normal mammalian development requires chromosome is paternally, rather than maternally,

64 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 65 Cherry Picking the Periodic Table: A New View of Life By James Trefil

The periodic table in an orderly way. Read from of the elements is a left to right in any row and the beautiful thing. It was elements progressively increase first written down in 1869 by in atomic number. Look at the Russian chemist Dimitri the entries in any column and Mendeleyev, a professor in St. you find elements with similar © bill frymire / masterfile.com Petersburg. Mendeleyev had an chemical properties. Men- the Santa Fe Institute are right, everything around us, includ- second asks why those particu- This contemporary interesting life—born as one deleyev arrived at this way of the periodic table may turn out ing living systems. Each bin is lar elements are used and not version of the peri- of 14 children to a family in organizing the elements after to play an important role in ad- labeled with the name of an ele- others. Let’s start with the first odic table has the Siberia, he was such a preco- years of trying to make sense dressing another fundamental ment—hydrogen, carbon, ytter- part. In the table below, we see elements arranged cious student that the entire of the seemingly chaotic assort- question—the question of how bium, and so on. Some of the some common chemical ele- in blocks set in family moved to St. Petersburg ment of chemical elements that life evolved on our planet. To bins, like hydrogen, are huge, ments, together with the per- columns. so that he could attend the uni- had turned up in the 19th cen- see what Morowitz is getting at, holding a significant amount of centage (by weight) in which versity. Later, after his work had tury. He had no idea why the we have to remember that, at all the material in the universe. these elements are found in the achieved international atten- table seemed to make sense, or bottom, life is based on chemis- Other bins, like ytterbium, universe at large, in the Earth’s tion, he went through a divorce why it predicted the existence of try, so that when we talk about are small, and represent less crust, and in the human body and remarriage. Technically, then unknown elements such as evolution, we are really talking abundant substances. But large (which we will take as a proxy this made him a bigamist in the germanium and scandium (they about how things arranged or small, what’s in these bins is for living systems in general). LIBRARY

OTO eyes of the Orthodox Church, showed up as gaps in his orderly themselves so that certain atoms the stuff from which the entire Two things leap out at us PH

CE a situation which is supposed arrangement). Understanding attached themselves to each universe is made. The question from this table. The first is N

SCIE to have prompted the Czar to why the table is the way it is other and interacted with other that Morowitz wants to ask that a massive winnowing of / OSTI

V declare, “Mendeleyev may have would have to wait for a half groups of atoms. To know how is simple: Why does life seem elements took place when the O N two wives, but I have only one century—for the discovery of life developed, in other words, to use some building materi- Earth formed—the mix of ele- RIA © Mendeleyev!” the structure of the atom and we have to start by knowing als more than others? Why, in ments on our planet isn’t much

This is the first version of the periodic table, drawn in 1869 by The periodic table is one of the development of modern how atoms and molecules came other words, does life seem to like the mix in the universe at Dmitri Ivanovich Mendeleyev. He left gaps in the table for new those organizational schemes quantum mechanics. Neverthe- to interact with each other as require more from some “bins” large. More importantly for this elements, which were indeed later discovered, vindicating his that demonstrates the fact that less, the table gives us a list of they do in living systems. in the periodic table than from discussion, though, we can see theory. the universe has an underlying the basic building blocks from And that’s where the periodic their neighbors? that yet another winnowing simplicity, despite its appar- which the universe is made. table comes in. Think of it as There are actually two parts took place when living systems ent complexity. It lists all of And today, if new ideas put a kind of giant Home Depot, to this question. The first part formed, because elements the known chemical elements forward by Harold Morowitz of with bins that contain all the asks what elements actually found there (carbon and iron, (118 and counting at this time) George Mason University and materials necessary to build appear in living systems, the for example) are not particu-

66 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 67 Cherry Picking the Periodic Table: A New View of Life By James Trefil

OTO eyes of the Orthodox Church, showed up as gaps in his orderly themselves so that certain atoms the stuff from which the entire Two things leap out at us PH

CE a situation which is supposed arrangement). Understanding attached themselves to each universe is made. The question from this table. The first is N

SCIE to have prompted the Czar to why the table is the way it is other and interacted with other that Morowitz wants to ask that a massive winnowing of / OSTI

66 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 67 larly common even on Earth. swer is obvious. For example, ther that each piece of fruit has and form multiple chemical Keshan’s syndrome is

BELOW: Like a oto.com

In fact, scientists have long one of the basic requirements little Velcro patches that allow bonds easily. In terms of our caused by a lack of the ph grocer adjusting element universe earth crust human body stock

known that living systems are of life is that there has to be it to attach to others. analogy, they come in big boxes element. And oddly i his pile of fruit to / v e suit his customers’ made almost entirely of a few some way to transfer informa- In the beginning, you might and have big Velcro patches. enough, there are even h H 74 1.5 10

glic Hydrogen elements—carbon, hydrogen, tion from one generation to the just put together a pile from But once we get past these a few obscure bacteria A

needs, nature has, is n e over time, shifted nitrogen, oxygen, phosphorus, next—the job done by DNA the biggest box—all oranges, simple atoms, Morowitz’s “fine in which experimenters D He 24 0 0 the mix of atoms and sulfur (a convenient mne- in living systems on Earth. for example. You would quickly tuning” begins to operate with have found that sele- Helium ; body: © in living systems to monic is CHNOPS). To these Carbon is an atom that can learn, however, that your a vengeance. In the chemical nium can be replaced L /JP C .005 0.01 18 make them better we add some elements common be formed into long, complex customers wanted more than maelstrom we call the cell, tiny by tellurium, the next ASA Carbon competitors. in seawater, like sodium and chains, and hence can be used oranges, and you would start differences in the efficiency of element in that column O .01 46 65 calcium, and a few trace ele- to form the backbone for in- arranging your piles of oranges a chemical reaction can have in the periodic table. Oxygen ments, and that’s pretty much formation-carrying molecules. so that you could fit in grape- a huge effect on the ability of This kind of experiment Fe .001 6.2 <.05 core: courtesty of N

it. So why, Morowitz asks, are Go down one row in the peri- fruit and lemons. Later, you the cell to reproduce. And just may eventually tell us h Iron these the chosen “bins”? odic table and we find silicon, might learn that the Velcro on as a slight customer preference something about how Ca .0007 4.6 1.5 Calcium

For some elements, the an- an atom that has properties the ordinary grapefruits was for lemons in our analogy will life works its way down ebula, eart similar to carbon and, one too strong, so that customers quickly produce more lemons the periodic table. n would think, could serve as the had a hard time taking them off in the citrus pile, a slight advan- The necessity of trace to incorporate it.” Common chemical basis for an alternate form of the pile, but that the Velcro on tage in reaction rate will allow elements in living systems can This way of looking at life elements together “DNA.” Indeed, silicon-based pink grapefruit didn’t create this the cell that has it to outcom- sometimes lead to surprising and the periodic table is new— with the percent- life has been a staple of science problem. Over time, then, your pete, and eventually eliminate situations. For example, a num- as you read this, SFI scientists age (by weight) fiction for years. Unfortunately, pile would come to have only its competitors. ber of years ago in Australia, have been thinking about it for in which they when you look at the properties pink grapefruit. Eventually, Morowitz points to an inter- sheepherders were puzzled only a few months. Thus, it is are found in the of silicon in detail, it just won’t driven by the desire to maxi- esting example of this effect. when large numbers of their difficult to imagine where it Earth’s crust, and work. For one thing, the bonds mize your sales, the pile might It turns out that all mammals flocks sickened and died when will lead. We know that each in the human body. formed between silicon atoms come to have proportions of require selenium as a trace grazing in a particular area. atom, each obscure element, It’s evident that a tend to be weak, and even citrus quite different from the element (it forms part of com- Investigators eventually found has a story to tell us about how massive winnow- when chains are formed, they proportion in the boxes from plex molecules that govern that the soil in that area was it came to be incorporated into ing of elements tend to be unstable in the pres- which the pile is made. the processes by which toxic severely depleted in cobalt, living systems. At the very least, took place when ence of water and oxygen. This In the same way, Morowitz materials are removed from an element that plays a role when we know these stories the Earth formed, and yet another makes them poor choices for an argues, over billions of years cells, increasing the efficiency in the chemistry of vitamin we will have filled in another took place when alternate biochemistry. living systems have fine-tuned of those processes). Selenium B12. Plants can thrive without piece of the marvelous tapestry living systems For other elements, however, their molecules in response to is below oxygen and sulfur in cobalt, but mammals cannot, of evolution. And maybe—just formed. the reason for the choice is less their environment, driven by the the periodic table, which means as the sheepherders learned to maybe—in uncovering these obvious. Morowitz proposes a inexorable pressure of natural it has similar chemical proper- their cost. (Feeding the sheep stories, we’ll find something simple answer to this question: selection. Like the grocer adjust- ties, but it is relatively rare (it’s cobalt supplements eventually completely new and unexpect- fitness, or, in his words, “fine ing his pile in response to cus- concentration in the Earth’s solved the problem.) ed. That’s the beauty of basic tuning.” To understand how tomer demand, nature has, over crust is only one ten-millionth “What we learn from the research. this works, think of a simple geological time, shifted the mix that of oxygen). Nevertheless, periodic table,” Morowitz says, Stay tuned! t analogy. Imagine that you are of atoms in living systems to because some ancestral organ- “is that life is always conducting a grocer who wants to make make them better competitors. isms containing selenium were chemical experiments, trying James Trefil is Clarence J. a pile of citrus fruit for your That life should be based on slightly better at surviving then to find that small advantage. Robinson Professor of Physics customers. You start with crates the CHNOPS atoms isn’t too those that were not, today we It doesn’t matter how difficult at George Mason University and of lemons, oranges, grapefruits, hard to understand—they are all need it. In fact, a rare heart it is to find the element you a member of the FIBR project. aula eastwood p

68 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 69 larly common even on Earth. swer is obvious. For example, ther that each piece of fruit has and form multiple chemical Keshan’s syndrome is

BELOW: Like a oto.com

glic Hydrogen elements—carbon, hydrogen, tion from one generation to the just put together a pile from But once we get past these a few obscure bacteria A

68 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 69 2008 SFI

Wednesday, May 14, 7:30 pm; Armory for the Arts Community Peter Norvig, Director of Research, Google Inc. Practice Makes Perfect: How Billions of Examples Lead to Better Models of Language, 4 Pictures, and Other Things 1

/ A computer might not learn in the same way that a person does, but it can use massive amounts of data to

5 perform selected tasks very well. In this talk, Norvig explores the way that a computer can correct spelling mistakes, translate from Arabic to English, and recognize celebrity faces about as well as an average human— Lectures and can do it all by learning from examples rather than by relying on programming.

SFI has exciting lectures lined up for this year. The March talk explores how genetic Monday, June 2, 7:30 pm; James A. Little Theater aspects of the brain influence our behavior. April and May presentations focus on Liz Bradley, Professor, Computer Science, University of Colorado, Boulder cutting-edge topics in information and computing. Summer events are devoted David Capps, Professor, Dance, Hunter College to art and science—dance, the visual arts, and music. In the autumn we consider 2 Con/cantation: Chaotic Variations various aspects of human social behavior. / Bradley and Capps present a dialog between computer and dancer in a performance piece revolving around 6 Los Alamos National Bank provides major underwriting for this program. a simple movement phrase, its mathematical refraction, and its reconstitution into a structure inspired by the theme and variation form. They describe the piece as “an unfurling of the possibilities of order, kinetic logic, and cause and effect.” This lecture is underwritten by the National Dance Institute of New Mexico. Wednesday, March 19, 7:30 pm; James A. Little Theater

Jon Wilkins, Professor, Santa Fe Institute Wednesday, July 16, 7:30 pm; James A. Little Theater

9 Devil or Angel: Genetic Conflicts in Brain and Behavior Charles Falco, UA Chair of Condensed Matter Physics and Professor of Optical Sciences,

1 The genes that we inherit from our mothers and fathers favor two different sets of behavior, and clash both University of Arizona /

during brain development and in real-time decision-making. This talk will discuss the causes and consequences 6 3

of this conflict, as well as what these discoveries mean for our understanding of the human mind and our notion 1 Art, Optics and Human Vision of the unified “self.” / Falco—working with painter David Hockney—has identified a variety of optical evidence within a number of 7 paintings demonstrating that artists as early as Jan van Eyck used optical projections as aids for producing portions of their images. While making these discoveries, Falco and Hockney developed new understanding The following lectures in April and May are underwritten by RedfishGroup, Santa Fe. Both lectures take place at about how more recent artists, such as Monet, Pissarro, and others, have created some of their iconic images. the Armory for the Arts, 1050 Old Santa Fe Trail. This lecture is underwritten by Gerald Peters Gallery.

Wednesday, April 23, 7:30 pm; Armory for the Arts Wednesday, August 13, 7:30 pm; James A. Little Theater Seth Lloyd, Professor, Mechanical Engineering, and Principal Investigator, Research Laboratory of Electronics, Massachusetts Institute of Technology; author, Programming the Universe: A Quantum Computer Dimitri Tymoczko, Arthur Scribner Bicentennial Preceptor and Assistant Professor, Music, 3 Scientist’s Takes On the Cosmos Princeton University 2 3 / 1

4 Programming the Universe The Geometry of Consonance: Music and Mathematics / Is the universe a giant computer? If so, how does it compute? It has been known since the nineteenth century Musical chords live in interesting geometrical spaces called “orbifolds”—spaces that contain unusual twists 8 that every atom in the universe carries with it bits of information, and that every time two atoms collide, and strange “singularities,” analogous to the black holes of general relativity. Tymoczko provides an accessible, those bits flip. The universe is carrying out a computation at the most microscopic level. This talk discusses multimedia introduction to this new way of thinking about music. the implications of the computing universe, ranging from quantum computation and quantum gravity, to the problem of free will and the ultimate future of life.

70 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 71 2008 SFI

/ A computer might not learn in the same way that a person does, but it can use massive amounts of data to

Jon Wilkins, Professor, Santa Fe Institute Wednesday, July 16, 7:30 pm; James A. Little Theater

9 Devil or Angel: Genetic Conflicts in Brain and Behavior Charles Falco, UA Chair of Condensed Matter Physics and Professor of Optical Sciences,

1 The genes that we inherit from our mothers and fathers favor two different sets of behavior, and clash both University of Arizona /

during brain development and in real-time decision-making. This talk will discuss the causes and consequences 6 3

70 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 71 2008 SFI Community Lectures

Tuesday, Wednesday and Thursday, September 16, 17, 18 7:30 pm; James A. Little Theater

Samuel Bowles, Professor, Santa Fe Institute; Professor, Economics, University of Siena Stanislaw Ulam Memorial Lectures: A Cooperative Species—How We Got to Be Both Nasty and Nice

8 Humans are remarkably cooperative animals. We frequently engage in joint projects for the common benefit

1 on a scale extending beyond the family to include total strangers. We do this even when contributions to , the project are costly and yield little private benefit. Examples include upholding social norms even when a

7 transgression would not be noticed, engagement in warfare, and actions to preserve the natural environment.

1 Lecture 1. A cooperative species (or are we just afraid someone may be looking?) , Drawing on archaeological, genetic, climatic, and other information about the conditions under which our 6 distant ancestors lived, Bowles will show why standard explanations of human cooperation are inadequate. 1

/ Lecture 2. Altruism, parochialism, and war: Rambo meets Mother Teresa

9 Bowles uses computer simulations to generate artificial histories of humanity over tens of thousands of years, tracing alternative trajectories that could explain how we got to be both nasty and nice. Lecture 3. Machiavelli’s mistake: why policies designed for “wicked men” fail The final lecture will show why taking account of our ethical dispositions and the conditions necessary to both enhance and empower cooperative motivations is essential if we are to face the challenges of environmental sustainability, control of epidemic disease, the governance of the information-based economy, and political violence.

Tuesday, October 7, 7:30 pm; James A. Little Theater

Frans B. M. de Waal, Director, Living Links, Yerkes Regional Primate Research Center; Professor, Psychology Department, Emory University

7 Our Inner Ape: On the Possibility of Empathy in Other Animals / De Waal explores expressions of empathy in animals, especially nonhuman primates, and presents a “Russian 0 doll” model of how animals perceive others. The model comes from a core emotional linkage arising from a 1 direct mapping of another’s behavioral state onto the subject’s representations. This “Perception-Action Model” provides the basis for an increasing distinction between self and other, so that the other is recognized as the source of felt emotions. This lecture is underwritten by Wayne and Barbara Coleman.

Wednesday, November 5, 7:30 pm; James A. Little Theater

Daniel Gilbert, Harvard College Professor of Psychology at Harvard University and Director of Harvard’s Hedonic Psychology Laboratory; author, Stumbling on Happiness 5 / Stumbling on Happiness 1

1 People want to be happy. To achieve this they must do two things. They must predict how they will feel in a variety of possible futures, and they must act to bring about the best of these and avoid the worst. Gilbert describes what science has to teach us about why we seem to stumble on (and not upon) happiness.

All lectures begin at 7:30 p.m. Most lectures are Wednesday nights—with the exception of June, September, and October. (See schedule above.) April and May lectures take place at the Armory for the Arts, 1050 Old Pecos Trail, Santa Fe; all other lectures take place at James A. Little Theater, 1060 Cerrillos Road, Santa Fe. Admission is free but seating is limited. For a more expanded explanation of the talks, visit www.santafe.edu/events/talks-public-lectures.php or call 505/984-8800.

72 Santa Fe Institute Bulletin SPRING 2008 TRANSITIONS

External Faculty of the early universe” and explored other fields and collaborating with The driving force of SFI’s scientific novel applications of information en- other researchers to tackle parallel life is its network of external re- tropy with the Santa Fe Institute’s J. problems across different disciplines. searchers, affiliated with universities Doyne Farmer. He is “thrilled to be affiliated with the and research institutions throughout A science writer by hobby, Bais fre- Santa Fe Institute,” as he considers it the world. Here are the most recent quently contributes to Dutch newspa- to be “the leading place in the world additions: pers and magazines. He has authored for interdisciplinary interaction in the two books for a general audience, one sciences.” David Ackley received a Ph.D. that walks the reader through a hand- In his free time, Bergstrom studies from Carnegie Mellon University, and ful of physics equations that “marked the art of bonsai. then went on to be a member of the turning points in our thinking about Cognitive Science Research Group at nature” and a pictorial guide to rela- Luis Bettencourt works for the Bellcore. Currently, he is professor of tivity that was published this year. theoretical division of Los Alamos computer science at the University of Bais lives with his family in the National Laboratory. He says it’s “a New Mexico. His ongoing research in- Netherlands, where he works as a great thrill to be involved with SFI, terests center on artificial life models professor at the University of Amster- which has pioneered the scientific and real artificial life; current research dam’s Institute for Theoretical Physics. environment essential for progress in emphases include genetic algorithms interdisciplinary research areas. “ and programming, distributed and Carl Bergstrom An expert on Bettencourt’s professional interest social computing, robust self-aware information and evolution, Carl lies in discovering themes in com- systems, and computer security. Bergstrom has modeled the spread plex systems where new data and of antibiotic resistance in hospitals, mathematical theory can test classic Sander Bais From quantum field mapped the flow of information ideas and develop predictive insights. theory to string theory, Sander Bais’s through scientific communities, and His current projects include analyz- research focuses primarily on prob- co-authored a paper with his father ing data from living neural networks, lems in theoretical high-energy phys- on parent-offspring conflict. He is an working with epidemiological data ics. He has, however, “made interest- associate professor of biology at the streams to predict the evolution of ing excursions to adjacent fields like University of Washington. emerging infectious diseases, and

top: © kip malone / santa fe institutecondensed matter theory and physics Bergstrom enjoys learning about understanding human social dynam-

Santa Fe Institute Bulletin SPRING 2008 73 ics through the study of urban or- Chile. As a member of the External ecosystem) of these rules/constraints. is currently putting the finishing entitled The Nature of Com- ganization. He is also conducting an Faculty, she welcomes the opportuni- In order to address these critical is- touches on a book about social and putation, and hopes that SFI empirical investigation of Santa Fe ty to further contribute to the vitality sues, the lab uses both theoretical, economic networks. He also co-edits will become his intellectual coffee shops. and intellectual energy of SFI. computational, biophysical, and a monograph series and a journal of home for the next couple of An avid rock climber and aspiring physiological and ecophysiological games and economic behavior. High- years. UNT Raissa D’Souza Trained as a blue water sailor, D’Souza intends approaches. Research in the lab can lights from his past studies include statistical physicist, applied math- someday to count scaling El Cap and be summarized into four distinct yet examining the incentives for countries Maya Paczuski found- euroscience, N ematician, and theoretical computer sailing to French Polynesia among interrelated areas: (1) The evolution to go to war and exploring a system ed the Complexity Science

scientist, Raissa D’Souza is currently her accomplishments. of form and functional diversity; (2) by which Andorran farmers mutually group at the University of etwork N a professor of engineering at UC Da- The origin of allometric relationships insure each other against fires. Calgary in 2006. Through vis, where she is helping launch their Brian Enquist is a broadly trained (how characteristics of organisms Jackson “looks forward to work- collaboration, the group enter for Complex Systems Center. Her cur- plant ecologist and associate profes- change with their size) and the scaling with other researchers at SFI and seeks to combine theory, C rent interest is building mathematical sor at the University of Arizona. His ing of biological processes from cells studying how social network struc- technological advances, models of feedback and interaction in lab investigates how functional and to ecosystems; (3) The evolution of ture impacts beliefs, decision making, and the best available data layered networked systems. physical constraints at the level of life-history and allocation strategies; and behavior.” to models of distributed in- rof. Guenterrof. Gross, D’Souza has been a regular visitor the individual (anatomical and physi- and (4) Community ecology and mac- formation networks (such as P to SFI since 1996, when she attended ological) influence larger scale eco- roecology. Jon Machta, a statistical physi- global computing), structure the Complex Systems Summer logical and evolutionary patterns. In Enquist first came to the Santa Fe cist, works with phase transitions, of complex networks (such courtesy of School as a Ph.D. student studying particular, the lab focuses on two core Institute in 1997 as a student in the non-equilibrium systems, pattern for- as protein interaction), cellular automata. She recently came areas: (1) Highlighting and deducing Complex Systems Summer School. mation and the computational power complex dynamical sys- Luis Bettencourt’s work on neural networks: Analysis of full circle when she attended the general principles, scaling rules, and of physical processes. He is interested tems, and social and eco- patterns of electrophysiological activity in cortical tissue China Complex Systems Summer the physical constraints influencing Duncan Foley is a non-main- in the interface between computer nomic phenomena. grown in vitro provides a window into collective information School in 2006 and 2007 as a lecturer the evolution of organismal form, stream economist who came to SFI in science and statistical physics both In September 2007, Pac- processing mechanisms in the brain. and helped organize the SFI-spon- function, and diversity; and (2) Un- 2001 for a workshop on the manage- from the practical angle of algorithm zuski presented some of her sored residency month at the Institute derstanding the larger scale ramifica- ment of common-property resources. development and for the insights that transdisciplinary research on signal- disease, and contriving a theory to for Complex Systems in Valparaiso, tions (ecological, evolutionary, and He has since been thinking about the computational complexity theory pro- ing-metabolic networks in a lecture explain why human infants don’t look range of ways physics concepts and vide in understanding and character- for the ENRAGEing Ideas workshop. like their parents. The reason, he says, methods can be applied to economic izing the emergence of complexity in The lecture addressed the challenges is “they have evolved to avoid detec- problems, the general area some- physical systems. of developing cancer treatments, and tion in the event that the domestic times called “econophysics.” He be- Machta is a professor at the Univer- how network theory and nanoscience father is not the biological father.” lieves that the Santa Fe Institute “can sity of Massachusetts, where he has could combine to address them. Interestingly, he is married with two provide a forum where physicists been head of the Physics Department Paczuski has committed to “in- children. interested in economic systems can for the past five years. He has greatly crease communication between learn more about economic history, enjoyed his visits to SFI and “looks different scientific disciplines and Sidney Redner, a Boston Uni- the history of economic thought, and forward to the closer connections as between science and the general versity physics professor, cut class what is known about the actual func- an External Faculty member and new public.” for the entire 2004–2005 academic tioning of economic institutions.” arenas in which to work.” year. He spent his truancy at the Los Foley, Leo Model Professor of Eco- Mark Pagel, editor-in-chief of the Alamos Center for Nonlinear Studies nomics at the New School for Social Stephan Mertens Trained journal Evolutionary Bioinformatics as the distinguished Ulam Scholar.

Davis Research, also has a strong interest in as a theoretical physicist, Stephan and a professor of biological sciences This jaunt in the Southwest brought UC the relation of evolutionary thinking Mertens claims an identity as a “mi- at the University of Reading, hopes to him to the Santa Fe Institute, where ouza / S to economic behavior, in statistical grant who wanders back and forth collaborate with SFI faculty in stud- he was introduced to like-minded re- theory and information, and in the between physics, mathematics, and ies of “assembly rules” for protein searchers. aissa D’ R complex neurological system that computer science.” He has worked on networks, on ideas of redundancy As a practitioner of statistical me- gives rise to circadian rhythms. He phase transitions in computational and evolvability, and on studies of chanics and network theory, he ap-

courtesy of hopes his work in these areas will in- problems, algorithmic complexity, language and cultural evolution. preciates that SFI is “a place where tersect with the thinking at SFI. pseudorandomness, and on parallel His past work includes statistically his brand of research is in the focus Raissa D’Souza’s research on emergent patterns in a model of flow and jamming: Bands of red computing. inferring the structure of dinosaur ge- rather than in the margins.” particles moving eastward and blue particles moving northward self-organize and delay the onset of Matthew Jackson, a professor Mertens is currently working with nomes, studying the evolution of the Redner has modeled aging and the jamming phase transition. of economics at Stanford University, SFI professor Cris Moore on a book prion proteins that cause mad cow immortality in cell proliferation, the

74 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 75 ics through the study of urban or- Chile. As a member of the External ecosystem) of these rules/constraints. is currently putting the finishing entitled The Nature of Com- ganization. He is also conducting an Faculty, she welcomes the opportuni- In order to address these critical is- touches on a book about social and putation, and hopes that SFI empirical investigation of Santa Fe ty to further contribute to the vitality sues, the lab uses both theoretical, economic networks. He also co-edits will become his intellectual coffee shops. and intellectual energy of SFI. computational, biophysical, and a monograph series and a journal of home for the next couple of An avid rock climber and aspiring physiological and ecophysiological games and economic behavior. High- years. UNT Raissa D’Souza Trained as a blue water sailor, D’Souza intends approaches. Research in the lab can lights from his past studies include statistical physicist, applied math- someday to count scaling El Cap and be summarized into four distinct yet examining the incentives for countries Maya Paczuski found- euroscience, N ematician, and theoretical computer sailing to French Polynesia among interrelated areas: (1) The evolution to go to war and exploring a system ed the Complexity Science

74 Santa Fe Institute Bulletin SPRING 2008 Santa Fe Institute Bulletin SPRING 2008 75 hopes to use his physics background Andreas Wagner heads a labo- to broaden the concept of statistical ratory at the University of Zurich’s De- mechanics in order to make it more partment of Biochemistry. The Wagner suitable for predicting complex sys- Lab pursues the study of evolution tems. Since his first invitation to the at all levels, from individual genes to Santa Fe Institute in 2000, Thurner has communities of organisms. Wagner analyzed magnetic fields of human himself is examining the evolution brain activity, developed a statistical of molecular networks. method for calculating gene expres- As a member of the SFI External sion, and studied banking networks Faculty, he hopes “to advance the

. ciliberti and financial asset price dynamics. frontier of our understanding of S He believes that “any progress in evolution by interacting with the handling complex systems resides in best minds that study complex

courtesy of finding ways of treating their intrinsi- systems.” t cally large number of variables si- Andreas Wagner’s work with the evolution of multaneously,” and that “shamelessly molecular networks: Each circle corresponds to a generalizing and advancing statistical Robert O. Anderson transcriptional regulation network, a network of mechanics is a promising starting d 1917–2007 genes that can turn each other on and off. point for a better understanding of Robert O. Anderson, Chairman social and economic systems that are of the SFI Board of Trustees effect of zealotry on elections, and way too complicated for the classic, in the early 1990s, died in No- can tell you exactly why soccer and reductionist approach to science.” vember 2007. baseball are more competitive than In his free time he raises funds for In addition to his leadership football. a school-building project in South role at SFI, Anderson served America. on the Board of Regents of Cosma Shalizi came to the Santa New Mexico Tech and was Fe Institute in 1998 to work in the Constantino Tsallis Born in Distinguished Professor of Pe- Evolving Cellular Automata Project; Athens but currently living in Brazil, troleum Engineering there. The the Computation, Dynamics and In- physicist Constantino Tsallis special- Robert O. Anderson School of ference group; and the Dynamics of izes in complexity and nonextensive Management at the University Learning group. With a background statistical mechanics. He is credited of New Mexico was named in statistical physics of complex sys- with introducing the notion of what as a tribute to him in 1974. He tems, most of his current work draws is known as “Tsallis entropy” and was a Life Trustee of the Cali- on information theory, which he con- “Tsallis statistics” in his influential fornia Institute of Technology, siders to be “an invaluable tool for 1988 paper “Possible generalization the University of Chicago, and proving probabilistic results.” of Boltzmann-Gibbs statistics” pub- the International Institute for He says his name really is Cosma, lished in the Journal of Statistical Environmental Development in and swears that it is “one of a small Physics. Tsallis recently spent two full London. number of Italian masculine names years at SFI as a visiting researcher, Besides being active in the ending in A.” When not writing books which he describes as “a wonderful oil business, Anderson’s other or teaching at Carnegie Mellon Uni- period in my personal and scientific business interests included versity, Shalizi likes to contemplate life when I had the opportunity of in- cattle raising and feeding op- Talking Heads lyrics and optimize the teracting with great scientists.” erations, mining and milling, information on his website for reten- As a member of the External Fac- and general manufacturing. tion in the pineal gland. ulty, he would like to “participate Anderson received numer- in all possible manners of the SFI ous awards for his tireless ef- Stefan Thurner, a “wannabe scientific life.” He hopes to visit SFI forts in support of public, chari- concert clarinetist” and the head of again, collaborate with various resi- table, and political affairs. the Medical University of Vienna’s dent members, and possibly organize Complex Systems Research Group, workshops.

76 Santa Fe Institute Bulletin SPRING 2008

SANTA FE INSTITUTE Non-Profit Organization 1399 HYDE PARK ROAD U.S. Postage SANTA FE, NEW MEXICO 87501 U.S.A. PAID Permit No. 49 Lubbock, Texas