Computational Social Science: Agent-based social simulation

versus Descriptive. Models can vary in the degree to which they attempt to incorporate the detail of particular targets. An example of a model which aims to be a detailed representation of a specific location and the developments there is the work by Dean et al. (1999) on the Long House Valley, in northern Arizona near Monument Valley. The model covers a time from about A.D. 400 to 1400 and consists of agent households that inhabit a digitized version of the Long House Valley landscape. Agents have rules for determining their agricultural practices and residential locations, as well as for reproduction and mortality. Each run of the model generates a unique history of population, agricultural output, and settlement patterns which can be compared with archaeological evidence from the Valley. In contrast, a series of papers (Conte and Castelfranchi 1995, Castelfranchi, Conte and Paolucci 1998, Saam and Harrer 1999, Staller and Petta 2001, Flenthe, Polani and Uthman 2001, Hales 2002, Younger 2004) has explored the relationship between norms and social inequality using a very simple ‘game’ in which agents controlled by ‘norms’ (i.e. behavioural rules) search a regular grid for ‘food’, which they consume to maintain their energy level or ‘strength’. These authors carried out experiments to study the mean and variance of the distribution of strength under various normative arrangements. Two examples are ‘blind aggression’, when agents attack other agents to grab their food, regardless of whether the attacked agent is stronger than they are, and ‘finders-keeper’, when agents respect ‘property rights’ and do not attack other