LUCSUS seminar: Inductive Risk Considerations in Detection and Attribution Studies
In recent years a debate has flared up within the detection and attribution community concerning the appropriate methodology for attributing climate change as a cause for extreme events. The conventional approach, that is based on the use of dynamic climate models and frequentist statistics, has been criticised by some that argue for an alternative, Beyesian approach, that downplays the importance of dynamical models. An important difference between the two approaches is how they balance different types of statistical error. In this paper we analyse this debate, and the differences between the two main approaches, in terms of what philosophers of science refer to as inductive risk and argue that methodological choices in this case should be informed by the wider social risks involved. This points towards adopting a Beyesian approach to detection and attribution.