All that glitters is not a deduction: Non-deductive methods in computational modelling (Extended abstract for IACAP 2023)

Computational modelling and simulations are often compared with experiments. It has been argued that these methods should be distinguished from experiments, that they are theoretical or that they require new epistemology. Many of these position are based on the intuition that as these methods rely on computation which can be reconstructed as series of deductive steps, the results they produce are conclusions of deductive arguments. Through a model-theoretical reconstruction of in silico experiments, I will demonstrate that the deductivist framework is not entirely adequate to capture their epistemology. Consequently, deduction is not an adequate criterion demarcating simulations from experiments.