Knowledge for Development

When all models are wrong


Date: 08/03/2015


More stringent quality criteria are needed for models used at the science–policy interface say Andrea Saltelli of the EC’s Joint Research Centre and Silvio Funtowicz at the University of Bergen, Norway. They argue that simple models could enable scientists and stakeholders to understand how assumptions and outputs are linked and that complex and often over-parameterized mechanistic models should be used only for more speculative investigations outside the policy realm. They present a seven-rule checklist to aid in the responsible development and use of models. These are: (i) use models to clarify, not to obscure; (ii) adopt an ‘assumption-hunting’ attitude; (iii) detect pseudoscience; (iv) find sensitive assumptions before they find you; (v) aim for transparency; (vi) don’t just ‘do the sums right,’ but ‘do the right sums’; and (vii) focus the analysis, don’t do perfunctory sensitivity analyses, merely changing one factor at a time. (Issues in S&T, 30/01/2015)

» Visit this site