Making Visible 'The Invisible Hand': The Mission of Social Simulation

Abstract

Sometimes in the literature of Social Simulation it seems that the problem is just to provide new (kind of) data to the social sciences and policies, and that, in order to do so, one has to refer to classical cognitive or social theories, implement them, and run some experiment. In my view, this is a reductive and subordinated attitude, which in the end will offer only a limited contribution to the social sciences. Let me be a bit provocative on this. Do we realize that social theories have – in general - been built without any true experimental method? Do we realize that in many cases social theories have not been grounded on really operational or formal concepts? It is true that Social Simulation is “the most promising approach to the social sciences”, provided that we conceive it in a more radical way. As Axelrod claimed, computer-based simulation is a third scientific approach, which should be added to the traditional ‘inductive’ and ‘deductive’ ones (Axelrod 1997). Indeed, simulation is so important and crucial because it finally provides the social sciences with a truly “experimental” method for the validation and adjustment of the models and, in particular, for the specification of working architectures and not mere formal descriptions. As we will see later in this contribution, such experimental support is even more crucial for the development of new social policies and strategies than for social theory per se. After all, we cannot experiment in real social contexts with real people and their everyday life! In particular we should stress the role of Agent-based social simulation (ABSS). Science in fact cannot be satisfied with just “laws” of the phenomenon under study; detecting regularities and describing their time course or expected outcomes is not enough, and neither is finding mere correlations. What about the causal mechanism producing that regularity, that outcome? An explicit modeling of proximate causes is a fundamental level of explanation and understanding, like the diachronic, historical, developmental, or evolutionary explanations. However, ABSS is not simply a generative approach (that might just be useful for developing theories) but also a synthetic, constructive approach: a radically operational approach. When creating an ABSS model, the phenomenon is engineered and even re-produced. As a consequence, ABSS is fundamental in my view for studying two crucial issues: Combined and complex effects; emergence; “Proximate” mechanisms, behind the phenomenon; hidden, underlying “causes” of the observable behavior. In a sense, computer simulation makes ‘visible’ and ‘tangible’ the hidden, background mechanisms that are assumed behind an observable (not necessarily social) phenomenon. And in such a way it provides us with explicit and clear (operational) models.