Keywords
Model verification, model validation, NUSAP, sensitivity auditing, post-normal science
Start Date
15-9-2020 8:40 AM
End Date
15-9-2020 9:00 AM
Abstract
According to Eker et al. (2018) both a positivistic and relativistic approach are possible in dealing with validation and verification of mathematical models. While the terms are potentially misleading, we explore here what those so-called relativistic methods can offer. While the past science wars have left a sinister cloud on the incursion of the second culture into the first (Snow, 1959), and the term relativism - often accompanied by the adjective 'absolute' is synonymous with the desecrated post-modern constructivism (Otto, 2016; Pinker, 2018), we show how approaches based on contextualising knowledge are - even and especially in the case of mathematical models, plainly useful and realistic. Indeed, omitting this crucial step in the procedure of quality assessment of a model, or of model-based inference, may result in ambiguity as to the purpose, context, assumptions and limitations of a study (Edmonds et al., 2019). Our point of departure is the epistemological stance of post-normal science, which has to its merit several decades of reflection on the issue of quantification (Ravetz, 1971; Funtowicz and Ravetz, 1990). We offer some of the methods available from the toolbox of PNS, such as NUSAP (van der Sluijs et al., 2005) and sensitivity auditing (Saltelli et al., 2013). We conclude with some considerations on the need for a more general ethics of quantification (Saltelli, 2020) where mathematical modelling can take inspiration from parallel developments in field such as statistical modelling, metrics, and algorithms.
Post normal model validation? Surely you must be joking
According to Eker et al. (2018) both a positivistic and relativistic approach are possible in dealing with validation and verification of mathematical models. While the terms are potentially misleading, we explore here what those so-called relativistic methods can offer. While the past science wars have left a sinister cloud on the incursion of the second culture into the first (Snow, 1959), and the term relativism - often accompanied by the adjective 'absolute' is synonymous with the desecrated post-modern constructivism (Otto, 2016; Pinker, 2018), we show how approaches based on contextualising knowledge are - even and especially in the case of mathematical models, plainly useful and realistic. Indeed, omitting this crucial step in the procedure of quality assessment of a model, or of model-based inference, may result in ambiguity as to the purpose, context, assumptions and limitations of a study (Edmonds et al., 2019). Our point of departure is the epistemological stance of post-normal science, which has to its merit several decades of reflection on the issue of quantification (Ravetz, 1971; Funtowicz and Ravetz, 1990). We offer some of the methods available from the toolbox of PNS, such as NUSAP (van der Sluijs et al., 2005) and sensitivity auditing (Saltelli et al., 2013). We conclude with some considerations on the need for a more general ethics of quantification (Saltelli, 2020) where mathematical modelling can take inspiration from parallel developments in field such as statistical modelling, metrics, and algorithms.
Stream and Session
false