Spatiotemporal Scales in Modeling: Identifying Target Systems

[摘要] My dissertation addresses neglected roles of idealization and abstraction in scientific modeling. Current debates about epistemic issues in modeling presuppose that a model in question uncontroversially represents a particular target system. A standard line of argument is that we can gain knowledge of a target system simply by noting what aspects of the target are veridically represented in the model. But this misses epistemically important aspects of modeling. I examine how scientists identify certain phenomena as target systems in their models. Building on the distinction between data and phenomena introduced by Bogen and Woodward, I analyze how scientists target systems from data and from basic theoretical principles. I show that there are two crucial empirical assumptions that are involved in identifying phenomena. These assumptions concern the conditions under which phenomena can be indexed to a particular length or time scale and the conditions under which one can treat phenomena occurring at different length or time scales as distinct. The role of these assumptions in modeling provides the basis for a new argument that shows how, in many cases, idealizations and abstractions in models are essential for providing knowledge about the world in so far as they isolate relevant components of a phenomenon from irrelevant ones. My analysis of the identification of phenomena also shows that structural uncertainty arises in models when the scale of a phenomenon of interest is not properly identified. This clarification promises to improve the communication of the limitation of current climate models to policy makers.

[发布机构] the University of Pittsburgh