Generative Social Science
The page on Models in Social Science explains a number of serious methodological difficulties for analysis of social interactions, as described by Joshua Epstein and Robert Axtell. To overcome these, they advocate instead a methodology of generative social science, using simple agent-based models to simulate the development of artificial societies. "In this approach fundamental social structures and group behaviors emerge from the interaction of individuals operating in artificial environments under rules that place only bounded demands on each agent's information and computational capacity. We view artificial societies as laboratories, where we attempt to 'grow' certain social structures in the computer – or in silico – the aim being to discover fundamental local or micro mechanisms that are sufficient to generate the macroscopic social structures and collective behaviors of interest." (p. 4). "We consider a given macrostructure to be 'explained' by a given microspecification when the latter's generative sufficiency has been established. ... we interpret the question, 'can you explain it?' as asking 'can you grow it?' In effect, we are proposing a generative program for the social sciences and see the artificial society as its principal scientific instrument." (p. 177).
The book is based around a powerful model called the Sugarscape, geographically a large grid of locations each of which produces – over time and dependent on its own fertility and seasonal conditions – a food resource called "sugar". Individual agents within the model all have their own fixed characteristics (e.g. sex, metabolic rate, range of vision for sugar) and others that are variable (e.g. health, marital status, wealth). Their behaviour is determined by very simple rules (e.g. look around for sugar nearby; if you find some, go there, eat as much as you need, and save the remainder). Even this is sufficient to generate interesting behaviour patterns, which can vary in illuminating ways with the introduction of such things as seasonal weather (leading to migration), breeding (subject to resources etc.), inherited wealth, a second resource called "spice" (leading to trading patterns), cultural identities (which can be changed voluntarily in response to social conditions). Some Sugarscape resources are available from the links below:
- Sugarscape software on sourceforge.net
- Model of Wealth Distribution (by Uri Wilenksky, CCL)
- HIV and AIDS in the Sugarscape (by Andrea Wiggins, Syracuse University)
In a later collection, Generative Social Science (Princeton, 2006), Epstein collects together a number of papers written in the wake of the earlier book. Three of these concern the "Artificial Anasazi Project", in which the Sugarscape model was developed using real archaeological data in an attempt to model the development and eventual collapse of the Anasazi society which inhabited the Long House Valley in northeastern Arizona between roughly 1800 BC and 1300 AD. The final paper can be found here: The Evolution of Social Behavior in the Prehistoric American Southwest. The project was described in Nature by Jared Diamond (10 October 2002, pp. 567-9), who suggested that it "set new standards in archaeological research", and concluded in these words:
One overall message of this project lies in the saying, 'God is in the details'. Without all those data on rainfall, groundwater, soil types, crop yields and household behaviour, verbal claims such as 'The Anasazi abandoned their homeland because of drought' are as likely to be wrong as right. Another message (even to a computer-phobe like me) is that the trajectories of human societies are so complex that computer modelling is essential for evaluating the consequences of the data. Finally, models involving many input parameters provoke much scepticism. But here the input parameters are measured ones, and the model is really nothing more than a calculation of the consequences of those input parameters, given certain assumptions. Experience with the 'artificial Anasazi' shows that confronting a computer model with reality yields conclusions about the questions that interest any archaeologist: why populations increased, why people moved, why their settlements varied in size, and why their society finally disappeared. This study should inspire archaeologists to gather the masses of data required to test their own verbal claims about human societies in other times and places.