Methodology and Statistics
International Conference, 14 - 17 September 2003
FDV, Ljubljana, Slovenia


Frans N. Stokman
University of Groningen
The Netherlands

My focus is on simulation as a tool in macro-micro-macro analyses of social systems. Most simulation models make too many ad hoc assumptions and are not or only very loosely connected with empirical data. In contrast, I argue for simple models, well grounded on theory, adding complexity stepwise. Following the paradigms of structural individualism, the emphasis lies on multi-agent modeling, taking the incentive structures of social actors as starting points. The model(s) should focus on the basic processes through which simultaneous and interdependent individual actions are transformed into collective outcomes. By feeding the models with parameters that specify the incentive structures of social actors and their social constraints (and are obtained in efficient data collection procedures), powerful models can be developed that have high predictive value and provide valuable instruments for social intervention.

I will illustrate the approach with models of collective decision making and social network evolution. Particularly the collective decision making models have been validated in extensive scientific and applied research, providing important and valuable extra insights even to 'experienced negotiators'.