Prof. Dr. Dirk Helbing Chair of Sociology, in particular ... · PDF Phone: +41 44 633 27 01...
6
www.ccss.ethz.ch Phone: +41 44 632 8880 ● ● e-mail: [email protected] www.soms.ethz.ch ETH Competence Center Coping with Crises in Complex Socio-Economic Systems (CCSS) Prof. Dr. Dirk Helbing Chair of Sociology, in particular of Modeling and Simulation Traffic theory has explained the relationships between fundamental indicators, such as speed, density and traffic flow. The most common relationship is called fundamental diagram, charactering the regimes of traffic flow (free or congested) in a specific road location. Such laws, however, are not sufficient to describe the entire complexity of traffic congestion in an urban road network. To obtain a better under- standing of city traffic, we follow a simulation-based approach , as data availability from cities is limited. We have developed innovative mode- ling techniques, which are able to reproduce the variability of urban congestion even for the same traffic volume. Moreover, we have discovered the spatial variability of density as a key variable of urban traffic flow. It reveals surprisingly clear functional relation- ships rather than producing large data clouds for congested traffic,as previous approaches did. (see figure below) In our computer simulations, a 6km 6km city-center is repre-sented by a lattice-like road network reminding of traffic flow in Manhattan or Barcelona. of social and economic systems can suddenly break down, and the balance between altruism and sel- fishness is therefore critical for the ultimate survival of societies. Let us consider, for example, the problem of climate change. Will society remain cohesive and make a joint effort to reduce Co2 emissions? Or on the contrary, will solidarity fall apart, resulting in uncoordinated individual behaviors? To address this issue, we have developed a model which allows us to pose the problem quantitatively. It basically consists of a population where individuals have the option to contribute or not to contribute to public goods. Moreover, they can also migrate, thus deciding how many people and whom they relate with. The figure below illustrates the kind of results we obtain with this model. Each panel displays different measures showing the degree of cooperation and aggregation in the emerging society. Our main conclusion is that the success of society in staying together and cooperatively working to a common end depends very much on the level of greed, of individuals. Only intermediate levels result in high levels of social cooperation and cohesiveness, which has strong impli- cations in terms of education and public policy.Too high greediness can drive a social system to instability, as it has been observed for the financial system. In another project, we investigate the crucial question of social cohesion. While social cohesion in human societies is largely dependent on high levels of cooperation and social ties, both features are steadily challenged by individual self-interest in social dilemma situations. As outbreaks of civil wars and the recent crisis of financial markets illustrate, the stability Evolution of a model society for two levels of greed: (a) 0.3 and (b) 0.8. The outcome clearly depends on this parameter. (a) For moderate greediness, society is stable, with high levels of aggregation and cooperation. (b) For high greediness, instability results with random behavior and general dissatis- faction. a) b)
Transcript of Prof. Dr. Dirk Helbing Chair of Sociology, in particular ... · PDF Phone: +41 44 633 27 01...
www.ccss.ethz.ch
Phone: +41 44 632 8880 ● ●e-mail: [email protected] www.soms.ethz.ch
ETH Competence CenterCoping with Crises in Complex Socio-Economic Systems (CCSS)
Prof. Dr. Dirk HelbingChair of Sociology,in particular of Modeling and Simulation
Traffic theory has explained the relationships betweenfundamental indicators, such as speed, density andtraffic flow. The most common relationship is calledfundamental diagram, charactering the regimes oftraffic flow (free or congested) in a specific roadlocation. Such laws, however, are not sufficient todescribe the entire complexity of traffic congestion inan urban road network. To obtain a better under-standing of city traffic, we follow a simulation-basedapproach , as data availability fromcities is limited. We have developed innovative mode-ling techniques, which are able to reproduce thevariability of urban congestion even for the same trafficvolume. Moreover, we have discovered the spatialvariability of density as a key variable of urban trafficflow. It reveals surprisingly clear functional relation-ships rather than producing large data clouds forcongested traffic,as previous approaches did.
(see figure below)
In our computer simulations, a 6km 6km city-center is repre-sented by alattice-like road network reminding of traffic flow in Manhattan or
Barcelona.
of social and economic systems can suddenly breakdown, and the balance between altruism and sel-fishness is therefore critical for the ultimate survival ofsocieties. Let us consider, for example, the problem ofclimate change. Will society remain cohesive and makea joint effort to reduce Co2 emissions? Or on thecontrary, will solidarity fall apart, resulting inuncoordinated individual behaviors? To address thisissue, we have developed a model which allows us topose the problem quantitatively. It basically consists ofa population where individuals have the option tocontribute or not to contribute to public goods.Moreover, they can also migrate, thus deciding howmany people and whom they relate with.The figure below illustrates the kind of results weobtain with this model. Each panel displays differentmeasures showing the degree of cooperation andaggregation in the emerging society. Our mainconclusion is that the success of society in stayingtogether and cooperatively working to a common enddepends very much on the level of greed, of individuals.Only intermediate levels result in high levels of socialcooperation and cohesiveness, which has strong impli-cations in terms of education and public policy.Too highgreediness can drive a social system to instability, as ithas been observed for the financial system.
In another project, we investigate the crucial questionof social cohesion. While social cohesion in humansocieties is largely dependent on high levels ofcooperation and social ties, both features are steadilychallenged by individual self-interest in social dilemmasituations. As outbreaks of civil wars and the recentcrisis of financial markets illustrate, the stability
Evolution of amodel society fortwo levels ofgreed:(a) 0.3 and (b) 0.8.The outcomeclearly dependson thisparameter.(a) For moderategreediness,society is stable,with high levelsof aggregationand cooperation.(b) For highgreediness,instability resultswith randombehavior andgeneral dissatis-faction.
a)
b)
www.ccss.ethz.ch
Phone: +41 44 632 83 50 e-mail: [email protected] www.sg.ethz.ch● ●
ETH Competence CenterCoping with Crises in Complex Socio-Economic Systems (CCSS)
Prof. Dr. Dr. Frank SchweitzerChair of Systems Design
Diversification degree k
Fin
ancia
l accele
ration a
10 20 30 40 50 60
0.02
0.04
0.06
0.08
0.1
3
6
11
19
32
Our research focuses on interdependences in finan-cial systems and their impact on systemic risk.To thisextent, the financial system is modeled as a networkof interdependent units which are connected byfinancial contracts. We find that a more connectedfinancial system is not necessarily more robust, as itis claimed in the literature (especially before thecurrent crisis). On the contrary, our analytical resultsshow that systemic risk is not a monotonic decreas-ing function of diversification. Instead, there is anoptimal level of diversification beyond which crisesbecome more frequent.Recently, our efforts have been directed in twodirections: One the one hand, we have investigatedto what extent our findings for the abstract modelremain valid if model parameters are varied. Wecould show analytically that the results are indeedrobust for various parameter modifications, forexample the intensity of financial acceleration (seeleft figure). On the other hand, we have developedmicro-founded models of financial systems in orderto investigate the non-monotonic behavior of sys-temic risk. For this, we have proposed a dynamicmodel of a bond market where financial institutionsissue bonds and purchase others' bonds. While basicassumptions about the model are consistent withthe finance literature,our network approach leads to
Systemic risk in a financial network. Expected first time for a systemicdefault to occur, as a function of diversification and financial
acceleration.
new and promising results regarding the relationbetween institutional couplings and systemic risk.This becomes non-trivial because the dynamics offinancial fragility displays super-exponential growthwith finite time singularities. In this context we planto strengthen the collaboration with the Chair ofEntrepreneurial Risk.We are working on the relation between systemicrisk and economic integration. Building on existingmodels of business fluctuations with heterogeneousagents, we investigate a system of firms and banksorganized in geographical regions. At one extreme ofthe integration range, firms borrow only from thebank of their region and sell their product only intheir region. At the other extreme,firms borrow fromand sell in all of the regions. Integration allows fordiversification of risk and reduction of profit fluctu-ations, but it also leads to potential financial conta-gion. We find that at beyond a certain level of inte-gration, systemic risk increases and gives rise tointermittent cascades of failures (right figure). Thisresult, qualitative similar to what found in ourabstract model,confirms the role of financial acceler-ation in the onset of systemic risk within a morerealistic context and in presence of heterogeneousfirms.
Systemic risk and geographic integration. As an example of lowintegration level, firms (circles) borrow from banks (squares) of two
different regions.
www.ccss.ethz.ch
Phone: +41 44 632 89 17 ● ●e-mail: [email protected] www.er.ethz.ch
ETH Competence CenterCoping with Crises in Complex Socio-Economic Systems (CCSS)
Excess volatility is one of the major unsolved problems in finance. It refers to theobservation that the volatility in financial markets is by far larger than the volatilityof the fundamentals, the data which are used to estimate the fair value of an asset.We approach this problem by modeling traders in an artificial stock market, whobase their trading decisions not only on the latest news of the fundamentals butalso on their own private information and the trading activity of their colleagues.We can show that by increasing the imitating behavior of the traders, the impactfrom the news onto price dynamics is amplified, resulting into a much highervolatility in the returns as in the news.By allowing the imitation strength,which canbe seen as a proxy for the uncertainty in the market, to change in time, our model isalso able to explain clustered volatility and crashes as they are seen in real markets.
Population has been growing increasingly fast over the last two centuries and was themost important driver of economic production measured as GDP. GDP itself can bedirectly related to CO2 emission which is believed to be the most important humanemitted greenhouse gas and harmful contributor to global warming. In the Kyotoprotocol, nations agreed not only to stabilize current emissions, but to bring back CO2share in the atmosphere to the 1990's level.This goal seems very ambitious to us. Considering population growth and economicdevelopment of emerging markets, such as China and India, which have just started tocatch up with "western consumer" lifestyle. China and India together represent 35% ofworld's population, but account only for 14% of gross world product and already emit aquarter of the current CO2 emissions.Analyzing this problem requires careful consider-ation of the interplay between population growth, industrial production and techno-logy.
Prof. Dr. Didier SornetteChair of Entrepreneurial Risks
A solution of the excess volatility puzzle based on stochastic resonance (G. Harras)
Economic model with endogeneous growth and evidence from globalpopulation and CO2 emmission trends (Andreas D. Hüssler)
We use the LPPL model in one of its incarnations to analyze two bubbles andsubsequent market crashes in two important indexes in the Chinese stockmarkets between May 2005 and July 2009. Both the Shanghai Stock ExchangeComposite and Shenzhen Stock Exchange Component indexes exhibited suchbehavior in two distinct time periods: 1) from mid. 2005, bursting in Oct. 2007 and2) from Nov. 2008, bursting in the beginning of Aug. 2009. We successfullypredicted time windows for both crashes in advance with the same methods usedto successfully predict the peak in mid. 2006 of the US housing bubble and thepeak in July 2008 of the global oil bubble. The more recent bubble in the Chineseindexes was detected and its end or change of regime was predicted inde-pendently by two groups with similar results, showing that the model has beenwell-documented and can be replicated by industrial practitioners.
Prediction of the 2005-2007 and 2008-2009 Chinese stock market bubbles (Financial Crisis Observatory)
CO2
CO
2in
ppm
Popu
latio
n in
mio
.
Population
year
year
time
www.ccss.ethz.ch
Phone: +41 44 633 39 43 ● ●e-mail: [email protected] www.ivt.ethz.ch
ETH Competence CenterCoping with Crises in Complex Socio-Economic Systems (CCSS)
Current integrated agent-based transport planningmodels use individual activity plans as a represen-tation of the transport demand. Each agent (repre-senting a virtual person in a microscopic simulation)has such a plan describing precisely his/her intendedactions during the analysis period under investiga-tion. The plan is called complete if it contains allinformation including activity types,times,locations,and routes.Currently, daily activity plans are often generated byusing an optimization approach: Various possibleactivity plans are created and examined using autility function. Finally, the assumed best plan is se-lected for execution.While suitable for modeling the average work day,this approach poses troubles when we want todescribe travel behavior in situations relatively faraway from every day situations, or when the state ofthe infrastructure shows strong fluctuations everyday.
The problem with such scenarios is the impossibilityof reliably predicting what is going to happen andhence – from the point of view of the agent – to planin advance how to perform. In real life, we deal withthat problem by simply reacting to whatever situ-ation we might encounter. This is usually done rela-tively quickly.We investigate how the described behavior can bereproduced by modeling for each agent a set ofendogenous needs (e.g. sleep, food, work, shop, lei-sure). These needs build up constantly and can besatisfied by performing a corresponding activity at asuitable location.The behavioral rule is now to monitor all of theseneeds and make sure that none of them “starves”.This is achieved by executing the correspondingactivities early enough to avoid overly high needlevels (right figure:two conflicting needs).The new modeling framework contains several newdata structures and algorithms to provide onlineinformation about the state of the infrastructure,locations, and the best routes to travel between anytwo locations (left figure:hierarchical routing table).
Prof. Dr. Kay AxhausenChair of Transport Planning and Systems
Need-based Short Term Travel Behavior
www.ccss.ethz.ch
Phone: +41 44 633 27 01 ● ●[email protected] www.ifb.ethz.ch/comphys
ETH Competence CenterCoping with Crises in Complex Socio-Economic Systems (CCSS)
The security of real networks, like power supply,communication, computer or spy networks, is an impor-tant issue. Networks are considered to be secure if theirfunction is not affected by the removal of nodes, whichcan be either random or targeted. Many networks areremarkably resistant against random failure, but amalicious, targeted attack can disrupt them removingonly a small fraction of nodes. We develop a method togenerate robust networks against malicious attacks, aswell as to substantially improve the robustness of a givennetwork by swapping edges and keeping the degreedistribution fixed. A novel measure for network robust-ness based on persistence of the size of the largest clusterduring attacks is introduced. The method was applied toseveral types of model networks and to the AS Internet.We find that the method improves robustness signifi-cantly. Our results show that robust networks have anovel "onion-like" topology consisting of a core of highlyconnected nodes hierarchically surrounded by rings ofnodes with decreasing degree. Networks of compromisedcomputers, botnets, are nowadays the main source ofmalicious internet traffic. Since botnets are used forsending unsolicited commercial emails,spams,character-istics of temporal patterns in receiving such emails can beused for revealing botnet properties.We model sending ofthe botmaster orders inside a botnet with a complexordering hierarchy,or sending of the spam emails throughinternet, as a dynamical process on directed complexnetworks. Our goal is to understand the source of correla-tions in interarrival times of spam massages. In the firstapproximation we model messages by random walkers,
sent using different dynamical rules from the chosensource nodes. We use the first arrival times of randomwalkers to the chosen receiver node to calculate the inter-arrival time distribution of messages.We find that even inthis simple model, for specific network topologies thedistribution of interarrival times deviates from exponen-tial distribution, meaning that the dynamical processesstudied are not random. The obtained distributions allowfor bursts in receiving spam,which is a property of spamm-ing process observed in the analysis of the real spam data.Traffic networks are essential elements of infrastructureand an important example of networks in which theinterplay between the network's topology and thedynamics taking place on it has a crucial role in network'sfunctionality. In order to understand influence of net-work's topology on its functioning, we introduce a trafficmodel on at first different model networks, and thenfinally on the road network of Switzerland. In modelingtraffic we use the ''coarse-grained" fluid-dynamicaldescription in which the traffic is viewed as a fluid formedby the vehicles. For given maximal velocity and maximaldensity of cars in each street, we study traffic flow ondifferent network topologies. Our aim is to gain the under-standing of mechanisms responsible for emergence oftraffic jams. Of our special interest is the study of thephenomenon of traffic gridlock, which is a special trafficjam affecting a whole network of intersecting streets. Thetraffic gridlock is an important case of infrastructurefailure. Understanding interaction between traffic dyna-mics and the topology of traffic networks, which is givingrise to this phenomenon,is an important step in increasingsecurity and robustness of infrastructure.
Prof. Dr. Hans J. HerrmannChair of Building Materials
www.ccss.ethz.ch
Phone: +41 44 632 67 59 ● ●e-mail: [email protected] www.icr.ethz.ch
ETH Competence CenterCoping with Crises in Complex Socio-Economic Systems (CCSS)
Prof. Dr. Lars-Erik CedermanChair of International Conflict Research
The ICR group studies crises generated by humanconflict, such as civil wars and ethnic conflict,focusing particularly on the consequences of macro-historical processes such as state formation andnationalism. While much of our recent research hascentered on data collection, including geographicinformation systems, the group also has a long-standing interest in computational modeling. Underthe auspices of CCSS, our computational framework“Geosim” was recently extended to a study of emer-gent state hiearchies. Lars-Erik Cederman & LucGirardin's (forthcoming in International StudiesQuarterly) agent-based model focuses on the shiftfrom indirect to direct rule, by explicitly representingthe causal mechanisms of conquest and internalstate building as organizational bypass processes.The computational findings confirm our hypothesisthat technological change is sufficient to trigger theemergence of modern,direct state hierarchies.In order to study another historical phase shift, Lars-Erik Cederman also initiated a project on war-sizedistributions in collaboration with CCSS colleagueDidier Sornette. By applying distributional analysisto conflict data from the last five centuries, we showthat the the size of interstate wars, measured as thenumber of combat fatalities, is power law distributed
and has increased significantly after the emergenceof nationalism following the French Revolution.Inthe study “Regions at Risk”, Sebastian Schutteestimates high-intensity conflict zones in civil wars.The project identifies guerrilla warfare and proposesa stylized model to analyze its spatial dynamics. Dataon infrastructure, terrain, population, and locationsof the primary political actors allow for the identi-fication of high risk zones along time-cost optimalcenterperiphery routes. These zones are good statis-tical predictors for conflict locations in civil wars.Thecombination of spatial statistics, political theory, andGeographic Information Systems enables this pro-ject to go beyond the traditional questions of conflictresearch.
Estimated risk zones for Central Africa. The red dots are conflictevents from the ACLED dataset.
The network of interstate alliance commitments in 1999.
Medieval states with deep hierarchies before the onsetof direct rule.
