The importance of being systematically surprisable: Comparative social simulation as experimental technique

We argue that computer simulation can serve as a functional equivalent for the experimental method in sociology, with respect to theory development. To this end we present accounts of experimentation and simulation by experimenting/simulating scientists and sociologists of science. From these analyses we conclude desirable features of a simulation method: generality, surprisability and power to separate. That means that it should be widely applicable, capable of surprising the researcher, and capable of separating surprising results that originate from sociological features of the model from those that stem from technical features. We demonstrate three methods that may provide these features: emergence, fixing points of reference, and comparative response testing. We develop the latter method in greater depth by discussing an exemplary simulation study.