Hernan Felipe Bobadilla Rodriguez (University of Vienna)
Scientists often resort to computer simulations to explain and understand natural phenomena. Several philosophers of science claim that these epistemic goals are related: Explanations provide understanding. Controversially, while some philosophers say that explanations are the only way to gain understanding, others argue that there are alternative, non-explanatory ways to gain understanding.
The aim of this paper is to assess explanations and understanding gained by means of computer simulations. In particular, I focus on assessing mechanistic explanations and mechanistic understanding – in the “new mechanist” sense. Furthermore, I examine the relations between mechanistic explanations and mechanistic understanding.
In order to achieve these aims, I perform a case study based on an agent-based computer simulation, known as the Olami, Feder and Christensen model (OFC, 1992). The OFC model predicts and explains aspects of a robust behaviour of earthquakes, known as the Gutenberg-Richter law. This behaviour consists in the robust power-law distribution of earthquakes according to their magnitudes across seismic regions. Roughly speaking, the OFC model simulates the power-law distribution of earthquakes by modelling the reciprocal influence between frictional forces and elastic deformation at a generic geological fault. In this case, a geological fault is represented as a cellular automaton in which cells redistribute elastic potential energy to their neighbouring cells when local thresholds of static friction are exceeded.
I deliver the following results:
1) The OFC model is a mechanistic model. That is, the component elements of the OFC model can be interpreted as mechanistic elements, namely entities, activities and organization.
2) The OFC model is a mechanism, namely a computing mechanism á la Piccinini (2007), which produces phenomena, namely outputs in a computer program.
3) A description of the OFC model, qua computing mechanism, mechanistically explains the power-law distribution of model-earthquakes.
4) The mechanistic explanation of the power-law distribution of model-earthquakes in the OFC models does not hold for real earthquakes. This is due to the lack of mapping between the mechanistic elements of the OFC model and the putative mechanistic elements in a geological fault. In particular, a mapping of mechanistic entities is problematic. The mechanistic entities in the OFC model, namely cells of the cellular automaton, are arbitrary divisions of space. They are not working parts in a geological fault.
5) However, the OFC model provides mechanistic understanding of the power-law distribution of real earthquakes. The OFC model provides us with a mechanism that can produce power-law distribution of earthquakes, even though it is not the actual one. Information about a possible mechanism may give oblique information about the actual mechanism (Lipton, 2009). In this sense, surveying the space of possible mechanisms advance our mechanistic understanding of real earthquakes.
