Agent Based Modeling Simulation Of Social Adaption And Long Term Change In Inner Asia

Agent Based Modeling Simulation Of Social Adaption And Long Term Change In Inner Asia - Transcript

Prepared for David Sallach et al 2007 The First World Congress in Social Simulation Kyoto Japan Springer

Agent based Modeling Simulation of Social Adaptation and Long Term Change in Inner Asia
Claudio Cioffi Revilla1 Sean Luke1 2 Dawn C Parker1 J Daniel Rogers3 W illiam W Fitzhugh3 William Honeychurch1 3 Bruno Frohlich3 Paula DePriest3 and Chunag A martuvsh in4
1

Center for Social Comp lex ity Krasnow Institute for Adv anced Study George Mason Univ ersity Fairfax Virg inia 22030 USA
2

Department of Co mputer Science Vo lgenau School of Information Technology and Engin eer ing G eorge Mason Univ ersity Fairfax Virg inia 22030 USA
3

Department of Anthropology National Museum of Natural History Smithson ian Institution P O Box 37012 Wash ington DC 20013 7012 USA
4

Mongolian Institute of Archaeology Mongolian Academy of Scien ces Ulanbataar Mongolia

Summary W e presen t a new international project to develo p temporally and spatially calibrated ag ent based models of the rise and fall of polities in Inner Asia Cen tral Eur asia in the p ast 5 000 years Gaps in th eory data and computational models for exp laining long term so ciopolitical chang e both growth and decay motiv ate this pro ject We expect three contributions 1 new theor eticallygrounded simulation mod els validated and calibrated by th e best available data 2 a new long ter m cross cultural datab ase with sever al d ata sets and 3 new concep tual theor etical and methodological contr ibutions fo r understanding social complexity and long term ch ange and adap tation in real and artificial so cieties Our theoretical framework is based on explaining so ciopolitical evo lution by the process of canonical var iation Keywords compu tational social science agent b ased mo deling env ironmental adaptation po litical d evelop ment so cial co mplex ity Inner A sia MASO N toolkit



This project is funded by the US N ational Scien ce Foundation grant BCS 0527471 More inform ation is available at http cs gmu edu eclab projects asia Thanks to Nig el Gilb ert Scott Moss Akira N am atam e D avid S allach and two anonymous referees for comm ents

1 Motivation and Purpose
Inner Asia is th e heartland of th e O ld World a bridge and large scale social network link ing Asia and Europe across the stepp e and a laboratory for understand ing long term social and political adaptations in the face of great challenges domestic and for eign human and physical Dep ending on th e epoch Inner Asia has fluctuated from being a cor e with influen ce o n neighboring regions China Russia South Asia Eastern Europe and the Middle East to a passivereactiv e periphery of such regions The nomads long distance contacts and exchange rapid tr ansport technologies and comp lex po lities o f Inner Asia offer opportunities to dev elop and test n ew th eories on the emergence of horizon tal and vertical polities Ferguson Mansbach 1996 as dynamic adaptive responses to social and environmen tal changes W e use d iachronic data from texts and from three ar chaeo logical projects located on a north south transect in th e Mongolian steppe and d evelop agen t based simula tion models that b uild upon and ex tend extan t compu tational social science models to gener ate th e emergen ce of mu ltiscale n etworks over space and time Our project aims to con tribute tow ards a b etter understand ing of social dynamic r esponses and collective behavior
11

1 22
33

Figure 1 Inn er Asia and Mongolia showing active surv ey sites 1 D arkhat Muron 2 the Khanui and Terhiyn Valleys and 3 B aga G azaryn Chuluu Unnumbered area represents a p rior survey d ataset at Egiin Gol 1996 2000

1 1 R esearch Goals We bu ild on extant efforts in computational h istor ical dynamics or agen tbased cliodynamics p araphrasing Turch in 2003 2004 Parisi 1998 by pursuing three synergistic goals 1 to develop test and analyze a n ew interdiscip linary theo ry of long term societal change and adaptation to complex and evolv ing so cial and physical environments a gener ative theory formalized by a spatial mu lti agen t computational model 2 to contribute to th e shared understand ing of social co mplex ity in the so cial sciences by in tegr ating concepts and principles within the p roposed theoretical framework and resear ch methodology and

3

3 to produce and d issemin ate n ew interdiscip lin ary data created by this project such as a new long term d ataset and diachronic atlas of Inner Asian polities 1 2 Observed Facts The space time un iverse of human and social dynamics wo rld system history Cioffi Revilla 2006 is vast and h eterogeneous in ter ms of origins and longterm evolution of social co mplex ity and environmen tal div ersity Regions of social space time with great dynamism and originality e g Asia in recen t millennia mix with others wher e social complexity was less pronounced e g North America 5000 years ago Th e long ter m fabric of social transformation is not all woven of the same material Sign ifican t differ ences and reg ularities o ccur in human and social dynamics across space and time Per egrine Ember 2004 Flannery 1999 Mar cus 1998 so comp arative resear ch is essential Inner Asia consists of Mongolia Inner Mongolia Tibet Manchuria Eastern Turkestan Xinjiang and parts of eastern Cen tral Asia and southern Sib eria Lattimore 1940 and w e focus on th e Mongolian steppe f ig 1 above The importance of th is r egion for understanding socio political dev elopmen t and in tern ational dynamics is well known as Inner Asia gave r ise to some of the world s most expansive empires From the 3rd millennium BC onward the eastern steppe of Inner Asia showed signif icant var iab ility in environment productive resources subsistence practices and sociopolitical organization as well as diver se networks of inter cu ltural contact including the f amed Silk Road network from ca 200 BC Malkov 2004 2005 From the end of the 1st millenn ium BC several hierarchical centr ally integrated and militar ily powerfu l polities emerged in Inner Asia includ ing the Xiongnu Turk Uighur and Khitan states Rogers 2007 The 13th century AD CE witnessed the Mongolian empire from the Sea of Japan to the Mediterran ean Sea The immensity of G enghis Kh an s imperial project and it s cultur al political and economic eff ects have long challeng ed scholars to explain how a relatively obscure group of steppe nomads managed to conquer and rule such a substantial swathe of the Old World According to an thropological models steppe conditions productiv e instability low population d ensity h igh mobility lower the potential for comp lex socio political organization and favor more egalitar ian polities Barfield 2001 Salzman 1999 Johnson and Earle 2000 By contrast th ere is compelling eviden ce th at by the early 1st millennium BC patterns of political co mplexity and greater collective action capacity asabiya Turchin 2003 from Ibn Khaldun characterized small groups later polities in Inner A sia Shelach 1999 Tsybik tarov 1998 Askorov et al 1992 H ieb ert 1992 1 3 Theoret ical D ef icits Social co mplex ity is a measure of differ entiation and in teg ration in a society and is char acter ized by hereditary social hier archy occup ational sp ecialization centr alized decision mak ing and governance capable of providing social viability in th e face of emerging ch allenges Comp lex so cieties range from small scale

groups with simple hier arch ies to the highly specialized multi str atified in tegrated polities of the modern world So me of th e most viable theor ies of so ciopolitical developmen t today ar e in terdisciplinary Epstein Axtell 1996 Feinman Marcus 1998 Flannery 1999 Turchin 2003 and they aim at integr ating social and environmen tal dynamics using div erse scien tif ic concepts an d princip les appropriate to th e complex topic of long term political d evelopmen t Needed is an in terd isciplinary theory th at bu ilds on extant progress with diverse an thropological economic so cio logical political psychological and environmental dynamics Such a th eory should be constructed w ith concepts and principles appropriate for th e onto logy of social complexity and polity in ter actions unobstructed by d isciplinary boundaries Formally th e new methodolog ical p aradigm of object based modeling is id eal for modeling and understanding human and social dynamics based on w ell defin ed attributes and behaviors instantiated in computational models of evolutionary adap tiv e agents capable of self gen erating and sustain ing higher order social complexity on mu ltiple scales

Figure 2 How d ata theo ry and computational models are integ rated in our project

The empirical record observed facts or main explanandum on social complexity in Inner A sian is our poin t of d epartur e fig 2 bottom This is used to inform our theory build agent b ased simu lations con ceptu alization formalization imp lemen tation and for valid ating calibrating an d testing the simu lations f eedback from testing Results from the simulation and empir ical files are used to test and r efin e the computational mod el as detailed in the next sections

2 Theoretical framework
2 1 C onceptual framework Change response and the social dynamics of complexity The processes emphasized in r ecent anthropological models in clude prestig ebiased cultur al tr ansmission amb itious agents social lead ership dynamics and technological change man ipulation of ideo logical and material med ia and strate

5

gic managemen t of intra and ex tra group relations Cioff i Rev illa et al 2005 Collective action capacity asabyia is important in Tur chin s metaethn ic frontier model 2003 arguably the most advanced formal and empir ically tested theory tod ay clear ly the state of the art in formal modeling and compu ter simulation of long term h istor ical changes in territorial states Collins 2003 Korotayev et al 2006 present a r elated project on long ter m world macrodynamics Situ ational chang es produce transf ers of information between indiv iduals and their environments causing th e former to sometimes co alesce in to social relationships th at span new networks Such a system is comp lex in the sense of von Neumann 1951 because it is iterativ ely capab le of gener ating increasingly more complex successors Indiv idual agency occurs in th ese processes en abled by opportunity and w illingness su ch as provided by extant r elatio ns cu ltural preceden t and resource and env ironmen tal constr ain ts Cioffi Rev illa Starr 1995 As with other networked phenomena social n etworks grow opportunistically prefer entially and through a nonlinear punctuated process w ith connectivity governed by power laws When successfu l collective action yields new bonds between individuals forging new relationships b ased on shared knowledge and experience and contributing to the emergence of norms processes and institu tions th at constitute social complexity These in turn increase a group s or a polity s collective action cap acity asabyia as a mission cr itical sociopolitical resource 2 2 Canonical t heory The process f or emergent social complexity fig 2 left We use the canonical theory of social complexity dev eloped by CioffiRev illa 2002 2005 formally der ived from the gen eral th eory of political uncertainty Cioffi Revilla 1998 As a canonical theory the iterative and uncer tain process of institutional emergen ce and political dev elopment and occasional d ecay is explained as resu lting from a succession of non deter ministic phase tr ansitions that occur in space time based on path dependen t variations on a common theme called the fast branching process As each canonical var iation of the same f ast process o ccurs for a giv en society social co mplexity may accrue decrease or remain the same producing the slow accrual pro cess based on emerging experience statecr aft bonds of trust norms institutions and other collective goods Negative externalities may also be produced lead ing to decay in so ciopolitical comp lex ity as explained below The canonical theory explains sociopolitical evolution by th e process of canonical variation

2 2 1 The fast micro process at the agent level Figure 3 shows the main even ts in the fast canonical pro cess deno ted by G C N U and S and their failure modes C S Th e process as a whole generates a political samp le sp ace outco mes on the righ t So cial complexity emerg es ev ent S the top outcome in as a path dependen t phase transition produced by a process of several differ ent albeit specif ic outcomes The full model not shown here has

five oth er detailed sub tr ees tr iggering or production rules for generating each of the f ive main events in th e fast branch ing process Cioff i Rev illa 2005 The first stage beg ins when an existing group lacks a system of governmen t i e the commun ity is not y et a polity even t G in fig 3 left and ma y end in a different situation politically complex phase when such a community has formed a system of government the co mmunity is a polity after iterations of the even t P in fig 3 top right In Inner Asian th is o ccurred up to th e early N eolithic per iod during which time not even ch iefdoms are archaeolog ically observable Th e process h ad begun by 2500 BC term inus ante quem

Figure 3 C anonical theo ry of social complexity the fast iterative branch ing process In turn each m ain ev ent G C N U S and complim entary failure modes has an associated cau sal model in conditional logic such th at the composite cau sal structu re bran ching process with all component sub models constitutes a sequ ential con ditional model C ioffiRevilla 1998 239 41 2002 2005

A situation al ch ange makes a group metastable because a potential for increased decreased sociopolitical complexity is created but not immed iately realized The realization of a potential for complexity depends on how the rest of the fast process evolves and on how people and environmen ts interact Xiongnu and Mongol societies su cceed ed many o thers failed Given a situation al change C the group may or may no t understand the need for collective action caused by the change ev ent N in fig 3 after C Causally N is an information processing event involving sign al detection cognition and other causal events and is modeled accordingly If th e group does not understand the situation al ch ange it may be destroy ed or dispersed without further political development outcome X hence the cr itical role of intellig ence If the group grasps the situational change it may or not be willing and ab le to undertake collective action U depending on its cap acity asabiya Collective action o ccurs in sever al modes Lichbach 1996 d etailed in Cioff i Revilla 2005 Sometimes so ciety fails to under take collective action U even when it understands a situational change C by in capacity If co llectiv e action occurs w ith p ersistent situational chang e then it may su cceed S or fa il S depending on the situ ation and action If it fails society may be destroyed outcome Z Th e

7

Mongol f ederation succeeded oth ers f ailed and w ere absorb ed or destroyed as the Mongol imperial polity expanded Sever al outcomes in th e outcome space even ts denoted by X can all eventu ally lead to state failure for example by long term loss of collective action capacity Finally if the society succeeds at time t then the consequen ces or societal effects w ill aug ment its po litical complexity outco me A at time t 1 b ecause ev en if only on a small scale and temporarily mob ilization of resources lessons about who to trust hierar chies of leader s and followers specialized assignmen ts d ivision of labor information sharing coordin ation experience and other elemen ts of governance w ill hav e been realized through the exper ience Significantly collective action capacity CA C for d ealing w ith the n ext situational change thr eat or opportunity will in crease The phase tran sition in th e quantum increase in CA C is observab le by the formation of multiplex networks on several scales cognitive individual group and institutional Such a phase tr ansition has enduring organization al eff ects on the group and the nex t time their situ ation changes and demands collective action they w ill draw on more CA C and cope better th ey will have more governance exper ien ce than b ef ore Th e phase transition also means th e realization of the poten tia l th at h ad been created by the ear lier situ ational ch ange when the group had become metastable after the in itial phase 2 2 2 The slow macro process at the so cieta l level A single passage through the fast canonical process w as just descr ibed O ver time a group will experience many such processes each as a v ariation on the common th eme of challeng e response Failure paths lead to po litical d ecay or even destruction events ben eath P so gains in political comp lexity are not preordained asabyia is not produced automatically In Inner Asia the slow process at th e societal lev el ev entually generated statelevel polities Xiongnu T rk Uighur Khitan Mongol empire and others and some failures The Xiongnu polity formed ca 200 BC beca use the Xiongnu society at that time was able to overcome through collectiv e action a situ ational change giv en by Chinese attack s from the south which took place while the Xiongnu had a pre state system of government H ad Xion gnu society f ailed in their collective action it would hav e tr ansition ed into on e of the oth er forms of social complexity in the ou tcome space For ex ample it may have b een destroyed conquered or d ispersed ev ent X Th e latter wer e failure instan ces in dealing with invasions economic demograph ic or enviro nmental changes For examp le power struggles af ter the 15th century produced political d eclin e and Mongolia was even tually conquered by the Man chus in th e 1 7th and 18th centuries

3 Towards computational implementation
3 1 Model instant iation and analysis with MA SON The canonical th eory is b eing instantiated w ith an agent b ased model in the new MASON simulation environment h ttp cs gmu edu eclab projects mason

MASON is a fast easily extendab le multiagen t simu lation library w ith v isualization too ls and other modules and is a join t effort of Sean Lu ke original cr eator of MASON and Cioffi Rev illa MASON h as been co funded by George Mason University s Evo lution ary Computation Laboratory and G MU s Cen ter for Social Complexity MASON was explicitly designed to foster cross fertilization between computer science and the so cial sciences and so supports the interd isciplinary goals of NSF s Human and So cial Dynamics Pr iority Area Other discr ete even t mu lti ag ent simulators exist bu t MASON meets our design cr iteria better than o ther systems because it is faster portable comp letely separab le visualization modeling w ith ch eckpoin ting and guaranteed replicability Luke et al 2005 Th ese MASON features are essential for our goals in cluding th e special needs of evolu tionary computation 3 2 Inner Asia Agent Based Model ModelofInnerAsia MIA fig 4 will be a simulation that instan tiates the canonical theory fig 3 a model for understanding polity flu ctuations produced by societal responses to challenging ch anges in the tr adition of earlier terr itorial competition models Cioff i Revilla Go tts 2003 The objective is to gen erate pristine social comp lexity over a territory not evolv e pre existing polities The design prin cip le is to build the A BM using the fast and slow dynamics of the canonical theory wh ich endogenizes the process of collective responses or failures

The planned simu lation will have individual ag ent granular ity during the initial process wh en co llectiv e action cap acity asabyia is relatively low and limited to group hunting skills elements of leadersh ip intellig ence coordination so me logistics MIA will b e developed as a resear ch programme Lakatos 1970 through a sequence of models w ith progressive problemsh ifts The models will

9

form a hierarchy from local to global each model aggreg ating and abstracting results of the previous model Model I will consist of a sing le group of human agen ts in a simple environment to understand the fundamental human and so cial d ynamics of th e f ast branching process fig 3 ear lier This model will exp lore the complete outco me space p articu lar ly th e r elative fr equency for the top explan andum event accru al of social comp lex ity by successfu l collective action W e will draw on the W etlands model in MA SON Cioffi Rev illa et al 2004 see Fig 5 below giv en its focus on hunter gather er groups or households that ex change information in a changing environment wh ere min imal so ciality emerges Basic land use patterns are alr eady present in the Wetlands version of Model I sin ce groups mak e diff erential use of the landscap e feeding tr ansit shelter etc

a

b

c

d

Figure 5 The MASON Wetlands simulation Cioffi Revilla et al 2004 separates computation from visu alization to increase sp eed and computational p ower Th e visualization layers portray the sp atial distribution of a ethnic g roups b climate dark blue is very rainy white is dry and offers sh elter c food sources g reen and d the composite Wetlands world w ith nom adic groups searching food seeking sh elter and adapting to climate

Model II will expand the env ironments and groups to grow the first ch iefdoms or interaction network of simple au tonomous polities Mod el II will explore why some groups evo lved politically comp lex so cieties in some ecotopes and o thers did not An explanation based on the canonical th eory would use asabyia as a key causal var iab le group cultural attr ibute Model III will b e calibrated to run for a long histor ical per iod similar to the duration for the f irst state level Xiongnu system to emerg e ca 200 BC Each mod el w ill in crease the space time scale of the simulation from local to in creasingly global with several mesoscales The main simu lation loop for each time step will be formalized from the canonical theory starting from what triggers situ ational ch anges event C fig 3 to the production of one of th e political ou tcomes in th e outcome sp ace of the f ast branching process P or one of the X failure events in fig 3 The loop will include explicit situation dep endent information processing Devlin 1991 Simon 1996 multi mode d ecision making collective action problem solving processes Lichb ach 1996 includ ing CA capacity dynamics Turchin 2003 and opportu

nity willingness conditioning Cioff i Revilla and Starr 19 95 Starr 1978 The main loop always b egins from situational change C Sigmoid functions for modeling various tipp ing point trigg ers called driv en thresho ld systems in complexity theory Rundle et al 1996 will be develop ed including mo deling th e probability of new object formations as complexity evolves 3 3 Model Calibration and Verification We aim at level 2 or 3 models Ax tell Epstein scale capab le of replicating the evolution of political landscapes with sufficien t detail for historical r ecognition In addition to hand calibration of the model b ased on emp irical data the mod el w ill also auto calibrate using g lobal stochastic op timization tech niques su ch as Evolutionary Computation EC Fogel Michelew icz 2000 Mitch ell 1996 see Gilb ert and Troitzsch 2005 ch 10 for an overview of EC in social simu lations EC iteratively makes small modif ications to th e mod el p arameters and rules i e it evolves the code tests those modif ications ag ainst a set of known data th e train ing set and updates mod el parameters and ru les based on feedback from testing even tually producing a model that fits th e data as closely as possib le EC and related methods have been successful at discovering or calibrating models in ant co lony optimization competitiv e game playing and robot team simulation The techniqu es may also be used to in crease mod el robustness by selecting a subset of parameters over which w e w ish th e mod el to produce invariant resu lts The EC th en tr ies to optimize the model in th e f ace of changin g settings from these parameters such changes may also be co ad aptive to the op timization system The primary challenge for EC is model complexity hence speed If th e model takes a long time to run it will be ch allenging to use in an EC framework requiring larg e numbers of runs But while EC is not cen tral to o ur project we will be able to d emonstr ate the efficacy of these techniques on a mo del of this size Co PI Sean Luke has developed a popular stochastic optimizatio n system called ECJ EC in Jav a www cs gmu edu eclab Luke 2000 specially designed to deal with large scale models ECJ dovetails well w ith MA SON After calibrating the model through hand tuning and comp uter optimization we will then validate the fin al mod el ag ainst a separate validation set of data to claim gener ality by compar ing distribu tions and sto chastic functions log it models statistical momen ts life tab les of actors settlement and land use patterns Markov processes counting processes and other features in the databases Our virtual h istor ies must match the real histories ob tained fro m th e archaeological ethnographic and env ironmental data

5 Conclusions
Our project is mo tiv ated by gaps and deficits in theory data and computational models for exp laining long term sociopolitical chang e in terms of both developmen t and decay We anticipate thr ee contr ibutions to our understanding of human and social dynamics in response to change and long term adap tation 1 a

11

new theoretically grounded simulation model valid ated and calibrated by the best availab le data 2 a n ew long term cross cultural database with sever al datasets and 3 new con cep tual theoretical and methodolog ical co ntributions for understand ing social comp lex ity and long term chang e and adaptation in real and artificial societies Our theoretical framework is based on explaining sociopolitical evolution by a process called canonical variation We have provided an initial roadmap for these contr ibutions in th e sections of this paper Besid es th eir intr insic value each contr ibution can provide foundations for furth er scientific adv ances in theory d ata and methodology as well as suggest n ew lin es of scien tif ic investigation

References
Askarov A V Volkov N Ser Odjav 1992 Pastoral and nomadic tribes at th e beginning of the first m illennium B C In H istory of C ivilization of C entral Asia A H D ani and V M Masson eds pp 459 468 Paris U NESCO Barfield T 2001 The Sh adow Empires In Empires SE Alcock T D Altroy KD Morrison CM Sinopoli eds pp 10 41 Camb ridge University P ress Brem er SA M Mihalk a 1977 Machiavelli in m achin a In P roblems of Wo rld Modeling KW D eutsch ed Boston Balllinger Cioffi R evilla C 1998 Politics and Un certainty C ambridg e Univ ersity Press Cioffi R evilla C 2002a Inv arian ce and universality in social agent based simulation s Proceedings of the N ation al Acad emy of Sciences of the U S A 9 9 Suppl 3 no 14 May 14 2002 pp 7314 7316 http www pnas org cgi content full 99 suppl 3 7314 Cioffi R evilla C 2002 N ecessity and sufficiency in social phen omena theo retical and methodological progress In N ecessary Conditions G Goertz H Starr ed s L anham MD Rowm an Littlefield 2002 pp 295 303 Cioffi R evilla C 2005 2002 A canonical theo ry for the em ergence and developm ent of social comp lexity Journal of Mathematical Sociology 29 133 153 Cioffi R evilla C H Starr 1995 reprinted 2002 Opportunity willingness and political uncertainty Journal of Theoretical Politics 7 4 447 476 Cioffi R evilla C 2006 The B ig Collapse A Brief Cosmology of Globalization In G lobalization and Global History ed s B Gills WR Thompson London and New York Routledge Cioffi R evilla C NM Gotts 2003 Comparative analy sis of ag ent based social simu lations GeoSim and FEARLUS models Journ al o f A rtificial Societies and Social System s 6 4 http jasss soc su rrey ac uk JASSS htm l Cioffi R evilla C S Luke DC Parker JD Rogers WW Fitzhugh W Honeychurch B Frohlich P DePriest 2006 Agent B ased Dynamics of Social Complexity Modeling Adaptive Beh avior and Long Term Change in Inner Asia Proposal to the National S cience Foundation Human and Social Dynam ics P rogram Devlin K 1991 Logic and In form ation Cambridge University P ress Epstein J 2006 Gen erative So cial S cience Princeton University P ress Epstein J RL Axtell 1996 G rowing Artificial Societies C ambrid ge MA MIT P ress Feinman G J Marcu s eds 1998 Archaic S tates School of Am erican Research Press Ferguson YH RW Mansbach 1996 Polities Authority Identities and Change Colum bia SC Univ ersity of South C arolina Press Flannery KV 1999 Process and ag ency in early state form ation Cambridg e Archaeological Journal 9 3 21

Fogel DB Mich alewicz Z 2000 How to Solve It B erlin Springer V erlag Gilbert N K Troitzsch 2005 Simulation for the Social S cientist 2nd ed Open University Press Hieb ert F 1992 Pazyryk Chronology and E arly Horse Nomad s R econsidered Bulletin of the A sia Institu te 6 117 129 Johnson A T Earle 2000 The Evolution of Hum an Societies Stanford University P ress Korotay ev A A Malkov D Khaltourin a 2006 Introduction to Social Macrodynamics Secural Cycles and Millenial T rends Mo scow Russia URSS Scientific L iteratu re and Textbooks Lakato s I 1970 Falsification and th e Methodology of Scien tific Research Programm es In I Lakatos A Musgrav e ed s C riticism and the G rowth of Knowledge C ambridge University Press Lattimore O 1940 Inner A sian Frontiers o f Chin a N ew York Ox ford University Press Lichbach M 1996 Th e Cooperato r s D ilemm a Un iversity of Michigan Press Liverani M D P arisi 1998 Assiri v irtu ali un laborato rio did attico Iter 1 27 38 Luke S 2000 Issues in Scaling Genetic Programm ing Ph D Dissertation Dep artmen t of Computer S cience University o f Mary land Co llege P ark Mary lan d Luke S C Cio ffi R evilla L Pan ait K Sullivan 2005 MASON Simulation 81 7 517 527 Malkov AS 2004 Spatial Modeling of Historical Dynamics Pro ceedings of the International Conference on Mathem atical Modelling of Social and Eco nomical Dynam ics Moscow Russia June 23 25 2004 Malkov AS 2004 The silk road simu lation Paper presen ted at the 34th Annual Meeting of the Society for Cro ss Cultu ral Research Santa Fe New Mexico Febru ary 23 27 2005 Marcus J 1998 The Peaks and Valleys of Ancient States In Archaic States G Feinm an J Marcus eds pp 59 94 S anta F e S chool of American Research P ress Mitchell M 1996 An Introduction to G enetic A lgorithms Cambridge MA MIT P ress Parisi D 1998 A cellular automata model o f th e exp ansion of the Assyrian empire In Cellu lar Autom ata R esearch Towards Industry edited by S B andini R Serra and FS Liverani London Springer Peregrin e P M Ember eds 2004 Encyclopedia of Prehisto ry New York Kluwer Academic Planum Rogers JD 2007 The Contingencies o f State Form ation in Inner Asia A sia P erspectives in press Rundle JB W Kline S Gro ss 1996 Rupture Characteristics R ecu rrence and Predictability in a S lider B lock Model for E arthquakes In Redu ction and Predictability of Natural Disasters edited by JB Rundle W Klin e and D Turcotte R ead ing MA Addison Wesley Salzman P 1999 Is Inequality Universal Current Anthropology 40 1 31 61 Shelach G 1999 Lead ership Strategies E conomic A ctiv ity an d Interreg ional Interaction N ew Yo rk Plenum Press Simon HA 1996 The S ciences of the Artificial 3rd ed Th e MIT Press Starr H 1978 Opportunity and willingness as ord ering concepts in the study of w ar International Interactions 4 363 87 Tsybiktarov AD 1998 Kul tura plitochnykh mogil Mongolii i Zabaikal ia Culture of the Slab Burials of Mongolia and Z abaik al e Ulan Ude N auka Turchin P 2003 Histo rical Dynamics Princeton Univ ersity Press von Neumann J 1951 The G eneral and Logical Theo ry of Au tom ata In C erebral Mech anism s in B ehav ior edited by AL Jeffress N ew Yo rk John Wiley