Munich Center for Mathematical Philosophy (MCMP)
print


Breadcrumb Navigation


Content

Talk: Cristian Larroulet Philippi (Cambridge)

Location: Ludwigstr. 31, ground floor, Room 021.

24.01.2024 at 18:00 

Title:

Inductive risk, values, and credences: Assessing the Bayesian response

Abstract:

In a classic response to the argument from inductive risk (AIR), Richard Jeffrey (1956) articulated a Bayesian strategy that promised to insulate scientific reports from non-epistemic values. At its core, the ‘Bayesian response’ contends that binary cognitive attitudes towards hypotheses – e.g., belief/disbelief, acceptance/rejection – have no place in the repertoire of scientific experts, holding that scientists ought only to communicate their degrees of belief in hypotheses, thus dividing the labour of practical action: scientists furnish the credences, policymakers the utilities. Despite many objections, the Bayesian response is alive and well (Henschen 2021; Cassini 2021; Hatchwell and Papineau forthcoming). Our aim is to offer a broader assessment of the Bayesian response than has been offered thus far. Instead of asking whether the Bayesian response succeeds in blocking the AIR – the current focus of the literature – our assessment considers the Bayesian picture of scientific advice as a whole, assessing its broader commitments regarding the science-policy interface. Our first two interventions challenge the two normative commitments of the Bayesian picture directly: the outsourcing of utilities to policymakers and restricting expert advice to the communication of credences. We argue that both commitments cannot be maintained in light of the realities of scientific assessment for policy. Our third intervention is a plea for clarification from proponents of the Bayesian response: namely, for a specification of which probabilities scientists are meant to be reporting to policymakers in the first place. Given that the Bayesian response is offered not as a thesis in formal epistemology, but as a normative proposal to guide real-world scientific advisors, abstract talk of ‘credences’ won’t do – the Bayesian’s burden is to specify which account of credence they are working with, and how that account might be operationalised to yield justifiable numbers (by the Bayesian’s own lights) in an advisory context. Running through various operationalizations—scientists’ degrees of beliefs as reported by them, scientists’ degrees of beliefs as measured by their actions, and the (probabilistic) outputs of scientific models—we find the Bayesian’s burden to be a rather heavy one. Taken together, the three interventions render a negative assessment on the Bayesian picture of proper scientific advice.

Note: This talk is based on joint work with Ahmad Elabbar (University of Cambridge)