cial Societies Social Science from the Bottom Up �

In the past tractability issues have generally forced economists to focus their attention on specialized aspects of economic behavior without detailed consideration of psychological cultural demographic and migratory pro cesses The recent development of agent based compu tational frameworks such as the Sugarscape model de veloped by Joshua Epstein and Robert Axtell at the Brookings Institution raises new questions concerning the appropriate bounds of economic analysis for these frameworks permit the study of social activity from a much more inclusive perspective Indeed such frame works may at last provide a common paradigm for social science as a whole Using Sugarscape animations as illustration this fas cinating monograph persuasively makes the case that an inclusive agent based computational approach to the study of dynamic social systems can be both productive and fun The monograph consists of six chapters with technical details largely relegated to three appendices Chapter I provides general motivation for the authors agent based computational methodology Noting that this methodology has a long lineage including the semi nal work of Thomas Schelling the authors argue that the recent advent of more powerful computational tools has nally made large scale agent based modelling practical In particular object oriented programming languages such as Java C and Object Pascal the language used for Sugarscape permit the construction of arti cial societies consisting of heterogeneous agents situated in external environments Agents and environmental fea tures e g resource sites are instantiated as objects with internal attributes states and behavioral rules Non genetic attributes may adapt in response to agent agent and agent environment interactions and genetic at tributes may be passed in modi ed form to descendants Such an arti cial society can grow itself over time with large scale social structures and group behavior emerg ing from the bottom up i e from local agent agent and agent environment interactions There is no need to im pose ctitious top down coordinating mechanisms such as representative agents market clearing and environ mental stationarity Sugarscape is introduced in Chapter II in simple form Agents inhabit a two dimensional grid of sugar bearing sites where sugar is a renewable resource which agents must constantly collect and consume in order to survive Each agent has a xed randomly determined genetic en dowment consisting of a sugar metabolism sugar burned per time step and a vision level that helps it search for sugar Any sugar collected by an agent in excess of its metabolic needs is added to its sugar holdings wealth An agent dies any time its wealth drops to zero or below Experiments focusing on wealth distribution include an additional feature agents die in nite time even if they maintain positive wealth levels and any agent that dies is replaced by another randomly generated agent De spite the simplicity of this setting the migration patterns and distributions of wealth and genetic endowments that arise and persist are seen to be related in interesting ways to the xed structural characteristics of the arti cial so ciety namely sugar capacity and renewability rates at each site and local agent search rules In Chapter III the authors introduce sexual repro duction inheritence by children of parental genetic en dowments cultural transmission modi cation of non genetic attributes via agent agent interactions and combat wealth plunder The authors are then able to grow crude caricatures of early social history character ized by the formation of combative spatially segregated tribes cultural groupings In Chapter IV the authors introduce a second resource spice Agents have randomly determined metabolic needs for spice and sugar and so have an incentive to engage in sugar spice trades Only bilateral Pareto im proving trades between neighboring pairs of agents are permitted at prices determined locally by a bargaining rule When trade is the only behavioral mode prices tend ultimately to bunch around an equilibrium price level even though potential gains from trade are never fully extracted When additional behavioral modes are introduced however such as sexual reproduction and cultural transmission the markets that emerge tend to display persistently higher price volatility with random drift Various other related topics e g credit arrange ments and the emergence of nancial hierarchies are also explored in this thought provoking chapter The construction of a uni ed immunology and epi demiology model is the concern of Chapter V The au thors examine the consequences of allowing intra agent and inter agent spread of infectious diseases to interact with social processes such as trade and migration The concluding Chapter VI reviews previous ndings and dis cusses future possible research directions In summary the authors provide preliminary but in triguing evidence that computational frameworks such as Sugarscape can provide laboratories for exploring which micromechanisms su ce to generate important social phenomena As they note it is not the generation per se that is surprising but rather that simple local rules often su ce for this generation The acid test of explanation may thus one day come to be can you grow it