On Wednesday 27th February, Dr Roman Frigg, London School of Economics, will give the CS4 talk “Laplace’s Demon and the Adventures of his Apprentices”
Building 53 Room 4025, Highfield Campus, 4-5pm. Refreshments served after the talk.
“The sensitive dependence on initial condition associated with nonlinear models imposes limitations on the models’ predictive power. These limitations have been widely recognised and extensively discussed. In this paper we argue that the severity of these limitations notwithstanding, we haven’t seen the worst yet: it is structural model error rather than sensitive dependence that truly limits our ability to make useful predictions with nonlinear models. If a nonlinear model has only the slightest structural error, then its ability to generate useful prediction is lost. This puts us in a worse epistemic situation than sensitive dependence because we can guard against the effects of sensitive dependence by replacing deterministic predictions by probabilistic predictions, a route that is foreclosed in the case of structural model error. We reach this conclusion by retelling the tale of Laplace’s demon, but with a twist. In our rendering the Demon has two apprentices, the Freshman Apprentice and the Senior Apprentice, who have abilities that fall short of the Demon’s in ways that turns them into explorers of sensitive dependence and structural error respectively. We discuss in what way the problems we describe affect actual modelling projects and end by making a tentative suggestion about how to guard against the worst effects.”