Published as: Terán, O., Edmonds, B., Wallis, S. (2000) Mapping the Envelope of Social Simulation Trajectories. In Moss, S. and Davisson, P. (eds.). Multi Agent Based Simulation 2000 (MABS2000) Boston MA. Lecture Notes in Artificial Intelligence 1979. Berlin et al.: Springer 229-243
Discovering and studying emergent phenomena are among the most important activities in social research. Replicating this phenomenon in “the lab” using simulation is an important tool for understanding it. Multi-Agent Systems (MAS) provide a suitable framework for such simulation. When such simulations are used to represent social processes there are necessarily indeterministic and arbitrary aspects, which are typically represented as either random choices (or numbers) or constants chosen by the programmer. Each such ‘choice’ means that the simulation takes one of the possible ‘trajectories’. The implicit theory that a simulation represents is precisely not in the individual choices but rather in the ‘envelope’ of possible trajectories – what is important is the shape of the whole envelope. Typically a huge amount of computation is required when experimenting with factors bearing on the dynamics of a simulation to tease out what affects the shape of this envelope. In this paper we present a methodology aimed at systematically exploring this envelope. Thus it complements methods like Monte Carlo analysis, the inspection of single scenarios and syntactical proof. We propose a method for searching for tendencies and proving their necessity relative to a range of parameterisations of the model and agents’ choices, and to the logic of the simulation language. The exploration consists in a forward chaining generation of the trajectories associated to such a range of parameterisations and agents’ choices. Additionally, we propose a computational procedure that helps implement this exploration by translating the MAS simulation into a constraint-based search over possible trajectories by ‘compiling’ the simulation rules into a more specific form, namely by partitioning the simulation rules using appropriate modularity in the simulation. An example of this procedure is exhibited.