Taming Uncertainty: The Limits to Quantification

Actuarial research has aimed to tame uncertainty, typically by building on two premises: (a) the malleability of uncertainty to quantification and (b) the separability of quantitative modelling from decision principles. We argue that neither of the two premises holds true. The first ignores deeper – ‘ontological’ and ‘framing’– uncertainties, which do not lend themselves to quantification. Drawing on Plural Rationality Theory, we observe that concerns about (and responses) to different types of uncertainty are at the same time causes and consequences of differing professional identities and ways of organising. The second premise does not hold either: disagreements about the validity of statistical models and about the appropriateness of decision criteria are closely linked. Hence, the governance surrounding quantitative models needs to take into account a wide range of legitimate, if contradictory, stakeholder perspectives.

Keywords: Model uncertainty, model governance, model risk, plural rationalities, epistemic uncertainty, ontological uncertainty