Pushing Networks to the Limit

remedy this initial failure, the government re- set of SES variables across resource systems are 19. R. Wade, Village Republics: Economic Conditions for Collective
opened the fishery but divided the coastal area needed to enable scholars to build and test Action in South India (ICS, San Francisco, CA, 1994).
20. J.-M. Baland, J.-P. Platteau, Halting Degradation of
into more than 50 sectors, assigned transferable theoretical models of heterogeneous costs and Natural Resources (Oxford Univ. Press, New York, 2000).
quotas, and required that all ships have neutral benefits between governments, communities, and 21. J. M. Acheson, J. A. Wilson, R. S. Steneck, in Linking Social
observers onboard to record all catches (32). individuals and to lead to improved policies. and Ecological Systems, F. Berkes, C. Folke, Eds. (Cambridge
Furthermore, the long-term sustainability of Univ. Press, Cambridge, 1998), pp. 390–413.
References and Notes 22. P. N. Wilson, G. D. Thompson, Econ. Dev. Cult. Change
rules devised at a focal SES level depends on 41, 299 (1993).
1. E. Pennisi, Science 302, 1646 (2003).
monitoring and enforcement as well their not 2. R. B. Norgaard, Conserv. Biol. 22, 862 (2008). 23. E. Mwangi, Socioeconomic Change and Land Use in
being overruled by larger government policies. The 3. National Research Council, The Drama of the Commons Africa (Palgrave MacMillan, New York, 2007).
long-term effectiveness of rules has been shown (National Academies Press, Washington, DC, 2002). 24. E. Schlager, W. Blomquist, S. Y. Tang, Land Econ. 70,
4. L. Pritchett, M. Woolcock, World Dev. 32, 191 (2004). 294 (1994).
in recent studies of forests in multiple countries to 25. A. Agrawal, in People and Forests: Communities,
5. F. Berkes et al., Science 311, 1557 (2006).
depend on users’ willingness to monitor one an- 6. E. Ostrom, R. Gardner, J. Walker, Rules, Games, and Institutions, and Governance, C. C. Gibson, M. A.
other’s harvesting practices (15, 31, 33, 34). Larger- Common-Pool Resources (Univ. of Michigan Press, Ann McKean, E. Ostrom, Eds. (MIT Press, Cambridge, MA,
scale governance systems may either facilitate Arbor, MI, 1994). 2000), pp. 57–86.
7. F. Berkes, C. Folke, Eds., Linking Social and Ecological 26. P. B. Trawick, Hum. Ecol. 29, 1 (2001).
or destroy governance systems at a focal SES level.
Systems (Cambridge Univ. Press, Cambridge, 1998). 27. E. Ostrom, Understanding Institutional Diversity
The colonial powers in Africa, Asia, and Latin 8. M. A. Janssen, Complexity and Ecosystem Management (Princeton Univ. Press, Princeton, NJ, 2005).
America, for example, did not recognize local (Edward Elgar, Cheltenham, UK, 2002). 28. J. A. Brander, M. S. Taylor, Am. Econ. Rev. 88, 119 (1998).

Downloaded from http://science.sciencemag.org/ on May 16, 2016

resource institutions that had been developed 9. J. Norberg, G. Cumming, Eds., Complexity Theory for a 29. X. Basurto, J. Soc. Nat. Resour. 18, 643 (2005).
over centuries and imposed their own rules, which Sustainable Future (Columbia Univ. Press, New York, 30. H. Nagendra, Proc. Natl. Acad. Sci. U.S.A. 104, 15218 (2007).
2008). 31. E. Ostrom, H. Nagendra, Proc. Natl. Acad. Sci. U.S.A.
frequently led to overuse if not destruction (3, 7, 23). 103, 19224 (2006).
10. S. A. Levin, Ecology 73, 1943 (1992).
Efforts are currently under way to revise and 11. R. Axelrod, M. D. Cohen, Harnessing Complexity 32. C. W. Clark, The Worldwide Crisis in Fisheries: Economic
further develop the SES framework presented (Free Press, New York, 2001). Models and Human Behavior (Cambridge Univ. Press,
here with the goal of establishing comparable 12. E. Ostrom, Proc. Natl. Acad. Sci. U.S.A. 104, 15181 (2007). Cambridge, 2006).
13. J. Wilson, L. Yan, C. Wilson, Proc. Natl. Acad. Sci. U.S.A. 33. G. C. Gibson, J. T. Williams, E. Ostrom, World Dev. 33,
databases to enhance the gathering of research 273 (2005).
104, 15212 (2007).
findings about processes affecting the sustain- 14. W. A. Brock, S. R. Carpenter, Proc. Natl. Acad. Sci. U.S.A. 34. E. Coleman, B. Steed, Ecol. Econ. 68, 2106 (2009).
ability of forests, pastures, coastal zones, and water 104, 15206 (2007). 35. Supported in part by NSF grants BCS-0624178 and
systems around the world. Research across dis- 15. A. Chhatre, A. Agrawal, Proc. Natl. Acad. Sci. U.S.A. 105, BCS-0601320. I thank T. K. Ahn, R. Axtell, X. Basurto,
13286 (2008). J. Broderick, E. Coleman, C. Eavey, B. Fischer, C. A. González,
ciplines and questions will thus cumulate more E. Jameson, B. de Leon, D. Porter, M. Schlueter, D. Sprinz,
16. R. Meinzen-Dick, Proc. Natl. Acad. Sci. U.S.A. 104,
rapidly and increase the knowledge needed to 15200 (2007). and J. Walker for comments and suggestions.
enhance the sustainability of complex SESs. 17. G. Hardin, Science 162, 1243 (1968).
Quantitative and qualitative data about the core 18. T. Dietz, E. Ostrom, P. Stern, Science 302, 1907 (2003). 10.1126/science.1172133

mutual interactions, be it trade, ownership, R&D

PERSPECTIVE
alliances, or credit-debt relationships. Different
agents may have different behaviors under the
Economic Networks: same conditions and have strategic interactions
(1). These evolving interactions can be represented
The New Challenges by network dynamics that are bound in space and
time and can change with the environment and
coevolve with the agents (2). Networks are formed
Frank Schweitzer,1* Giorgio Fagiolo,2 Didier Sornette,1,3 Fernando Vega-Redondo,4,5 or devolve on the basis of the addition or deletion
Alessandro Vespignani,6,7 Douglas R. White8 of either agents or the links between them.
The socioeconomic perspective has empha-
The current economic crisis illustrates a critical need for new and fundamental understanding of the sized understanding how the strategic behavior
structure and dynamics of economic networks. Economic systems are increasingly built on of the interacting agents is influenced by—and
interdependencies, implemented through trans-national credit and investment networks, trade relations, or reciprocally shapes—relatively simple network
supply chains that have proven difficult to predict and control. We need, therefore, an approach that architectures. One common example is that of a
stresses the systemic complexity of economic networks and that can be used to revise and extend star-spoke network, like a very centralized or-
established paradigms in economic theory. This will facilitate the design of policies that reduce conflicts
between individual interests and global efficiency, as well as reduce the risk of global failure by making 1
ETH Zurich, D-MTEC, Kreuzplatz 5, 8032 Zurich, Switzerland.
economic networks more robust. 2
Laboratorio di Economia e Management (LEM), Scuola Superiore
Sant’Anna, Piazza Martiri della Liberta 33, 56127 Pisa, Italy. 3Swiss
he economy, as any other complex sys- can be traced back to the structural properties Finance Institute, c/o University of Geneva, 40 Boulevard Du Pont

T tem, reflects a dynamic interaction of a

large number of different agents, not just
a few key players. The resulting systemic be-
of the underlying interaction network.
Research examining economic networks has
been studied from two perspectives; one view
d’Arve, 1211 Geneva 4, Switzerland. 4Economics Department,
European University Institute, Via della Piazzuola 43, 50133
Firenze, Italy. 5Instituto Valenciano de Investigaciones Economicas,
Calle Guardia Civil, 22 esc. 2 no 1, 46020 Valencia, Spain. 6School
of Informatics and Pervasive Technology Institute, Indiana
havior, observable on the aggregate level, often comes from economics and sociology; the other
shows consequences that are hard to predict, as originated in research on complex systems in University, 919 East 10th Street, Bloomington, IN 47408, USA.
7
Institute for Scientific Interchange, 10133 Torino, Italy. 8Institute
illustrated by the current crisis, which cannot be physics and computer science. In both, nodes of Mathematical Behavioral Sciences, University of California,
simply explained by the failure of a few major represent the different individual agents, which 3151 Social Science Plaza, Irvine, CA 92697, USA.
agents. Thus, we need a more fundamental in- can represent firms, banks, or even countries, and *To whom correspondence should be addressed. E-mail:
sight into the system’s dynamics and how they where links between the nodes represent their [email protected]

422 24 JULY 2009 VOL 325 SCIENCE www.sciencemag.org

 SPECIALSECTION
ganization, in which a central “hub” channels all conflict between individual incentives and aggre- examine how these rules might be grounded
communication among agents. In this “micro” gate welfare, or their impact on the overall ef- on the basis of the economic incentives of the
perspective we focus on the individual system ficiency in the performance of the network at agents. Thus, instead of focusing on understand-
elements and their detailed network of rela- large. Furthermore, this problem is exacerbated if ing the endogenous behavior of individual agents,
tions. In contrast, for large setups, one adopts a the underlying environment is subject to persistent the complex-systems approach centers on under-
“macro” perspective that focuses on the statis- volatility, which may, for example, be due to the standing how the network-formation rules sys-
tical regularities of the network as a whole. Each intrinsic ephemerality of innovation (9), and if tematically affect the emerging link structure (4).
approach has its advantages and disadvantages. agents are out of equilibrium, as in most real- Networks generated with different stochastic
Previous work on the micro perspective was world situations. If this is the case, it is reasonable algorithms, such as random, scale-free or small-
strongly rooted in oversimplifying assumptions to posit that agents follow simple bounded- world networks, have been compared with real
on both the structure of the network and on rational rules that are modified in light of their complex networks including those in biology,
agents’ behaviors (3). For example, the micro ap- experiences. However, under such conditions, i.e., metabolic and genetic networks; infrastruc-
proach may have emphasized agent incentives agents are unable to attain efficient configura- ture, i.e., road networks and power grids; com-
in the development of informal links within firms tions, despite their continuous efforts to adapt to munication, i.e., internet and mobile phone; and
and may have failed to successfully predict real- an ever-changing situation. Additionally, even social interaction, i.e., collaborations (1, 2, 10).
istic dynamic outcomes. The macro approach small changes in environmental volatility can Comparing network structures across these dif-
better accounts for the large-scale system proper- have drastic consequences in the overall configu- ferent disciplines suggests that economic net-

Downloaded from http://science.sciencemag.org/ on May 16, 2016

ties, but fails in linking these to the economic ration of the system [e.g., (Fig. 1)]. works may also reflect a similar universality (11).
motivation of individual agents (4). The inability of previous approaches to re- Indeed, the connections of banks in an interbank
In recent micro approaches, economic networks produce statistical regularities that have been ob- network (12, 13), show the fat tail, characteristic
were often viewed as the result of a network- served empirically in network structures justifies of a scale-free system, that indicates that only a
formation game among competing and cooper- our pursuit of a complex-systems approach that few banks interact with many others. In this ex-
ating agents. In this regard, agents include firms may provide predictions for large-scale networks. ample, banks with similar investment behavior
that collaborate in joint R&D projects (5) or These predictions are made from the testing of will cluster in the network. Similar regularities also
workers who share information on job oppor- stochastic rules that affect link formation and can be traced for many examples including the
tunities (6); their links are added or deleted as that take into account, in addition to some sort international trade network (ITN) (14, 15) and
the consequence of purposeful decisions attempt- of randomness, the characteristic features of the regional investment or ownership networks (16).
ing to maximize their payoffs. In this context, agents, such as their degree of connectivity (num- In the complex-network context, “links” are
agents must rely on (and be able to) anticipate ber of links) or their centrality, as measured on the not binary (existing or not existing), but are
what others may do (in a generally imperfect and basis of the importance of a node—which, in weighted according to the economic interaction
asymmetric manner); use information about their turn, can be affected by its links to other nodes. under consideration [for example, in a network
environment (which may be limited); frame the However, the complex-systems approach pos- of major financial institutions worldwide shown
problem within some necessarily bounded time tulates rules exogenously and does not explicitly in (Fig. 2)]. Furthermore, links represent traded
horizon; and learn from the past, which volumes, invested capital, and so on,
may create a biased experience if sim- and their weight can change over time.
ilar situations are encountered later. Distinguishing networks at different
1 N = 100
These considerations tended to re- N = 1000
levels where we consider directed or
sult in a dramatically large number of N = 5000 undirected and weighted or unweighted
options that agents must choose from 0.8 links helps illuminate the evolution of
Network efficiency

on the basis of limited information. their topological properties.

The micro analysis of economic net- When the foreign direct invest-
0.6
works relies on game theory, which ments (FDI) among European firms
aims at identifying Nash equilibria (i.e., are presented as a directed network,
situations that are strategically stable in 0.4 power-law scaling is observed. This
the sense that no agent has an incentive scaling depends on the number of em-
to deviate). It can also rely on opera- ployees in both the investing and the
0.2
tions research, where algorithms for firm invested in, and on the number of
searching and optimizing have been incoming and outgoing investments of
developed. As the number of nodes 0 both firms (16) This allows single time-
0 0.2 0.4 0.6 0.8 1
and possible links scales up, however, point predictions about the investments
such problems become very difficult Environmental volatility that regions will receive or make, on
to solve, and classical approaches are the basis of the activity and connec-
Fig. 1. The structure and efficiency of an equilibrium network strongly
unsatisfactory. tivity of their firms.
depend on the conditions under which myopic agents can form links. We
The game-theoretic literature has Similar structural transitions can
show computer simulations that assume that agents prefer to connect to
highlighted the crucial role of incen- the neighbors of their neighbors that have a higher centrality, which also be detected in the ITN. By weight-
tives in the endogenous and induced creates local shortcuts. Network efficiency is measured on the basis of ing a country’s centrality in terms of
behavior of socioeconomic networks the aggregate centrality of agents. Environmental volatility measures the the likelihood that any given additional
(3, 7, 8). However, this micro approach risk that if any single agent is exposed to an exogeneous shock, it will dollar traded in the world reaches that
has not typically been integrated with force the deletion of one link. If the loss of links pushes the network country by following existing links
macro approaches that can identify the efficiency down and environmental volatility up past some critical level, with a probability proportional to its
complex systemic forces at work. With- the strongly homogeneous network structure will break down into a weight, the relative changes in central-
out this information, we cannot fully sparse, hierarchical structure, similar to a core-periphery structure and is ity over time show trends for different
understand important issues, such as the accompanied with a breakdown in network efficiency. countries that predict divergence in

www.sciencemag.org SCIENCE VOL 325 24 JULY 2009 423

 Pushing Networks to the Limit
Generali Mitsubishi UFJ Time and space. By allowing a time-dependent
Sumitomo resolution of the properties of economic net-
Royal Bank Scotland Lloyds TSB
works, we will be able to move beyond a single-
HBOS snapshot approach. This will allow the researcher
Bank of to identify the evolutionary path of networks
Gen.Electric Bank Nova America
Scotia Prudential Fin. through the combination of complementary infor-
Bear Stearns mation sources. A good illustration of this is pro-
Intesa–Sanpaolo Goldman Sachs vided by the R&D networks in the field of human
UBS ING biotechnology (20), which follow a predictable
Santander
life cycle related to the timing of the exchange
Aberdeen Soc .Generale and integration of knowledge.
Morgan Stanley
Commerzbank Structure identification. Extracting network
Capital Group topology from reported data, in particular for ag-
Mediobanca
gregated economic data, is very difficult. This is
FMR Corp
Unicredito particularly true for the banking sector, where
Sumitomomitsui HSBC Citigroup detailed accounts of debt-credit relations are not
Key Corp publicly available, although theoretical decom-
Friends Provident Barclays

Downloaded from http://science.sciencemag.org/ on May 16, 2016

JP Morgan positions of aggregated data have been studied
Cr.Suisse Chase BNP Paribas
Fidelity Mng. (13). Even then, analyses may resemble reading
tea leaves and reveal only previously known or
IFI Nomura predicted information. Statistical regularities in eco-
nomic networks can be identified through large-
Deutsche Bank scale data sets, but difficulties in assessing the
Wellington Mng.
Franklin Res. relevance of the various measures remain. In an
Merrill Lynch
evolving economic network, we require informa-
tion about agents’ roles, their function and their
Fig. 2. A sample of the international financial network, where the nodes represent major financial influence (23). New methods are needed to iden-
institutions and the links are both directed and weighted and represent the strongest existing relations tify patterns, and new concepts are needed to
among them. Node colors express different geographical areas: European Union members (red), North quantify both direct and indirect influence (e.g.,
America (blue), other countries (green). Even with the reduced number of links displayed in the figure,
through ownership). The identification in the ITN
relative to the true world economy, the network shows a high connectivity among the financial in-
of such roles based on similar positions in the net-
stitutions that have mutual share-holdings and closed loops involving several nodes. This indicates that
the financial sector is strongly interdependent, which may affect market competition and systemic risk work suggests that promising steps have begun to
and make the network vulnerable to instability. identify functional roles played by interactive agents
that relate to specific patterns in the link structure
regional integration within the global economy networks of interactions. We should then be able of their multirelational interaction network (24).
and do so better than traditional international trade to predict and propose economic policies that Mapping a large network as a homologous small
and macroeconomic statistics (17). This is also favor networks structures that are more robust to one, with statistically optimal sets of distinctive roles,
because the latter do not account for the entire economic shocks and that can facilitate integra- gives a statistical correspondence in the case of the
network topology but only consider bilateral di- tion or trade. Below, we briefly describe what is ITN world alignments for New World–Afro–East
rect trade links. For example, between 1980 and needed to tackle this endeavor. Asian versus North and Central Eurasia alignments
2005 East Asian countries experienced huge in- Massive data analysis. Our ability to obtain by cross-cutting each of their interconnected cores,
creases in their centrality scores, but the centrality more and better quality data will foster the tran- semiperipheries, and peripheries, as in world sys-
ranking of most Latin American countries fell. The sition from a qualitative to a quantitative and tem models, but with much greater precision.
trade statistics of these regions, however, displayed evidence-based science. As computational power Beyond simplicity. All economic networks are
similar patterns. In other words, these astonish- increases, large-scale network data can be gath- heterogeneous with respect to both their agents and
ingly different development records were not well ered for different levels of the economy (e.g., interaction strength and can also strongly vary in
tracked by international trade and macroeconom- firms, industries, and countries), and models can time (25). Previous studies of efficient (i.e., not
ics statistics. Thus, network-based approaches may be tested through the generation of large, syn- wasteful) and equilibrium (or strategically stable)
provide a more powerful way to manage, monitor, thetic, data sets. New processing methods should networks assumed homogeneity. However, as the
and govern complex economics systems. open a wide range of business data and internet differences between weighted and unweighted
However, a focus on centrality or other such communication that can be exploited. It will then network properties indicate for the ITN case (14),
properties of networks can only provide a first- be possible to gather individualized data on spe- any prediction of phase transitions may fail under
order classification that emphasizes the role of cific interactions over time such as employee flows, these simplified assumptions. In fact, heterogene-
fluctuations and randomness and cannot predict R&D collaborations, and so on within a business ities of agents can turn out to prevent phase tran-
the underlying dynamics of the agents, whether or firm-bank credit market interactions. These large sitions, i.e., become a source of stability.
they are firms or countries. We anticipate that the data streams require more powerful tools to digest Systemic feedbacks. Simple amplification
next generation of research will be able to mea- and manipulate the huge scale of available infor- mechanisms (such as herding) can dominate the
sure any deviations from universality and will mation reflecting agent interactions and network network dynamics at large, despite the best in-
allow us to identify the idiosyncrasies associated properties. Databases containing this information tentions of the agents. Economic networks are
with individual agent dynamics and their decision- may therefore complement both theoretical eco- subject to amplifications that may result from the
making processes. This new wave of research nomic network experiments (18, 19) and empirical redistribution of the load if one node fails (e.g.,
should begin to merge the description of indi- economic network studies (20, 21) and provide electricity in a power grid or credit debt in a bank-
vidual agents strategies with their coevolving large-scale observations in real-time (22). ing network). If a single node fails, it may force

424 24 JULY 2009 VOL 325 SCIENCE www.sciencemag.org

 SPECIALSECTION
other nodes to fail as well, which may eventually appropriate description of economic agents and 15. D. Garlaschelli, M. I. Loffredo, Phys. Rev. Lett. 93,
lead to failure cascades and the breakdown of the their interactions, and a systemic perspective be- 188701 (2004).
16. S. Battiston, J. F. Rodrigues, H. Zeytinoglu, Adv. Complex
system, denoted as systemic risk. This applies in stowing a new understanding of global effects as Syst. 10, 29 (2007).
particular to financial networks where links coming from varying network interactions are 17. J. Reyes, S. Schiavo, G. Fagiolo, Adv. Complex Syst. 11,
represent standing debts and claims between needed. We predict that such studies will create a 685 (2008).
connected financial institutions. However, it is more unified field of economic networks that ad- 18. M. Kosfeld, Rev. Netw. Econ. 3, 20 (2004).
19. S. Callander, C. Plott, J. Public Econ. 89, 1469 (2005).
not well understood how the structure of a vances our understanding and leads to further insight. 20. W. Powell, D. White, K. Koput, J. Owen-Smith,
financial network affects the probability of a We are still far from a satisfactory identification Am. J. Sociol. 110, 1132 (2005).
systemic failure. Although a topical subject, most and integration of the many components, but the 21. B. Kogut, G. Walker, Am. Sociol. Rev. 66, 317 (2001).
theoretical and empirical methods are not suited recent advances outlined suggest a promising start. 22. D. Sornette, F. Deschatres, T. Gilbert, Y. Ageon, Phys. Rev.
Lett. 93, 228701 (2004).
to predicting cascading network effects. The 23. T. A. Snijders, G. G. van de Bunt, C. E. Steglich, Soc.
References and Notes
mainstream view assumes that a denser network 1. F. Vega-Redondo, Complex Social Networks (Econometric Networks, in press; published online 26 March 2009
allows for a better diversification of the individ- Society Monographs, Cambridge Univ. Press, Cambridge, 2007). (10.1016/j.socnet.2009.02.004).
ual failure risk (26). However, systemic risk has 2. A. Barrat, M. Barthelemy, A. Vespignani, Dynamical 24. J. Reichardt, D. White, Eur. Phys. J. B 60, 217 (2007).
Processes on Complex Networks (Cambridge Univ. Press, 25. A. Kirman, J. Evol. Econ. 7, 339 (1997).
been shown to increase, depending on the cou-
Cambridge, 2008). 26. F. Allen, D. Gale, J. Polit. Econ. 108, 1 (2000).
pling strength between nodes (27). Furthermore, 3. M. O. Jackson, A. Wolinsky, J. Econ. Theory 71, 44 (1996). 27. S. Battiston, D. Delli Gatti, M. Gallegati, B. Greenwald,
most stable dynamic network models account 4. R. Albert, A.-L. Barabasi, Rev. Mod. Phys. 74, 47 (2002). J. Stiglitz, J. Econ. Dyn. Control 31, 2061 (2007).

Downloaded from http://science.sciencemag.org/ on May 16, 2016

for only the addition or removal of a single agent 5. J. Hagedoorn, Res. Policy 31, 477 (2002). 28. We would like to thank M. König, C. J. Tessone, S. Battiston,
to or from the network at each instance of time. 6. M. Granovetter, Getting a Job: A Study of Contacts and and S. Vitali (ETH Zurich) for aid with figures and S. White
Careers (Univ. of Chicago Press, Chicago, 1995) for commentary on the text. Data for Fig. 2 were provided by
However, the addition or removal of whole groups Orbis Database (end of 2007), Bureau Van Dijk. F.S. and
7. V. Bala, S. Goyal, Econometrica 68, 1181 (2000).
of agents to or from the network (e.g., as part of 8. M. D. König, S. Battiston, M. Napoletano, F. Schweitzer, D.S. acknowledge financial support from the ETH Competence
a systemic failure) may result in a larger, less Netw. Heterog. Media 3, 201 (2008). Center, “Coping with Crises in Complex Socio-Economic
predictable, and less stable system. 9. M. Marsili, F. Vega-Redondo, F. Slanina, Proc. Natl. Acad. Systems” (CCSS), through ETH Research Grant CH1-01-08-2.
Sci. U.S.A. 101, 1439 (2004). F.V-R. gratefully acknowledges financial support from the
In summary, we anticipate a challenging re- Spanish Ministry of Education under grant SEJ2007-62656.
10. S. P. Borgatti, A. Mehra, D. J. Brass, G. Labianca, Science
search agenda in economic networks, built upon 323, 892 (2009). A.V. acknowledges funding from NIH, DTRA, the EC-FET
a methodology that strives to capture the rich pro- 11. R. M. May, S. A. Levin, G. Sugihara, Nature 451, 893 (2008). program and the Lilly Foundation. D.W.’s work is supported
cess resulting from the interplay between agents’ 12. G. Iori, G. De Masi, O. Precup, G. Gabbi, G. Caldarelli, by external faculty funding at the Santa Fe Institute and
J. Econ. Dyn. Control 32, 259 (2008). anonymous nonprofit contributions to the University
behavior and the dynamic interactions among of California at Irvine faculty group in Social Dynamics
13. M. Boss, H. Elsinger, M. Summer, S. Thurner, Quant.
them. To be effective, however, empirical studies Finance 4, 677 (2004). and Complexity.
providing insights into economic networks from 14. G. Fagiolo, S. Schiavo, J. Reyes, Phys. Rev. E Stat. Nonlin.
massive data analysis, theory encompassing the Soft Matter Phys. 79, 036115 (2009). 10.1126/science.1173644

huge libraries of historical meteorological patterns

PERSPECTIVE
into large-scale computational simulations. Al-
though we often complain about the accuracy of
Predicting the Behavior of daily weather forecasts, we must remember that
numerical weather models and predictions allow
Techno-Social Systems us to project the path and intensity of hurricanes,
storms, and other severe meteorological occurrences
and, in many cases, to save thousand of lives by
Alessandro Vespignani anticipating and preparing for these events.
Given the success that has been achieved in
We live in an increasingly interconnected world of techno-social systems, in which infrastructures
weather forecasting for decades, why haven’t we
composed of different technological layers are interoperating within the social component that drives their
achieved the same success in the quantitative pre-
use and development. Examples are provided by the Internet, the World Wide Web, WiFi communication
diction of the next pandemic spatio-temporal pat-
technologies, and transportation and mobility infrastructures. The multiscale nature and complexity of
tern or the effects over the next decade of connecting
these networks are crucial features in understanding and managing the networks. The accessibility of new
billions of people from China and India on Internet
data and the advances in the theory and modeling of complex networks are providing an integrated
growth and stability? The basic difference is that
framework that brings us closer to achieving true predictive power of the behavior of techno-social systems.
forecasting phenomena in techno-social systems
starts with our limited knowledge of society and
odern techno-social systems consist of driven by human behavior. To predict the be- human behavior rather than with the physical laws

M large-scale physical infrastructures (such

as transportation systems and power
distribution grids) embedded in a dense web of
havior of such systems, it is necessary to start
with the mathematical description of patterns found
in real-world data. These descriptions form the
governing fluid and gas masses. In other words,
though it is possible to produce satellite images of
atmospheric turbulence, we do not yet have large-
communication and computing infrastructures basis of models that can be used to anticipate scale worldwide, quantitative knowledge of human
whose dynamics and evolution are defined and trends, evaluate risks, and eventually manage fu- mobility, the progression of risk perception in a
ture events. If fed with the right data, computational population, or the tendency to adopt certain social
modeling approaches can provide the requested behaviors. In recent years, however, tremendous
Center for Complex Networks and Systems Research, School level of predictability in very complex settings. progress has been made in data gathering, the dev-
of Informatics and Computing, and Pervasive Technology
Institute, Indiana University, Bloomington, IN 47408, USA;
The most successful example is weather forecast- elopment of new informatics tools, and increases in
and Institute for Scientific Interchange, Turin, Italy. E-mail: ing, in which sophisticated supercomputer infra- computational power. A huge flow of quantitative
[email protected] structures are used to integrate current data and data that combine the demographic and behavioral

www.sciencemag.org SCIENCE VOL 325 24 JULY 2009 425

 Economic Networks: The New Challenges
Frank Schweitzer, Giorgio Fagiolo, Didier Sornette, Fernando
Vega-Redondo, Alessandro Vespignani and Douglas R. White (July
23, 2009)
Science 325 (5939), 422-425. [doi: 10.1126/science.1173644]

Editor's Summary

Downloaded from http://science.sciencemag.org/ on May 16, 2016

This copy is for your personal, non-commercial use only.

Article Tools Visit the online version of this article to access the personalization and
article tools:
http://science.sciencemag.org/content/325/5939/422

Permissions Obtain information about reproducing this article:

http://www.sciencemag.org/about/permissions.dtl

Science (print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week
in December, by the American Association for the Advancement of Science, 1200 New York
Avenue NW, Washington, DC 20005. Copyright 2016 by the American Association for the
Advancement of Science; all rights reserved. The title Science is a registered trademark of AAAS.