Title: On the epistemology of computer simulation: modeling as asymptotic approximation
Session Lead: Santiago Nunez-Corrales
Time: 11 am – noon, Wednesday, 2021-03-03
Related Materials: link
The character of computer simulation remains a topic of strong debate in philosophy of science. While clarity has been gained about what it means to simulate for specific problems and knowledge domains, a definition that is simultaneously general, actionable and structural remains elusive. For example, determining what it means for a simulation to be realistic depends on multiple layers of interpretation, some of which provide either not enough intellectual freedom to capture surprising aspects of nature straightforwardly (i.e. generative effects), or provide too much freedom to the point of allowing models that bear little correspondence to fundamental aspects of phenomena they intend (and fail) to capture (i.e. final state sensitivity).
Being in possession of such a theory of simulation would greatly facilitate answering a plethora of questions related the validity of simulations in general, and even about the requirements computer experiments must satisfy from construction to reproduction. We describe here preliminary work around computer simulations departing from a modified version of the 20-question game, called the Omega-questions game. By precisely defining the sequence of operators, the type and amount of information required to capture phenomena in various situations, we obtain an operational semantics around usual notions found across computer simulations. Using these results, we propose a refinement of the scientific method where computer experiments embody the notion of bi-simulation, becoming entities for which both object and meta-object languages capable of substantiating falsation become admissible.