Wlodek Rabinowicz (Lund University), "Probability as Value"

Colloquium Series

"Probability as Value"

Wlodek Rabinowicz (Lund University)

Friday, September 22, 2023
3:30-5:30 PM EST
Leacock 927

Abstract: According to the Fitting-Attitude Analysis of value (FA-analysis, or short), to be valuable is to be a fitting object of a pro-attitude. A very similar approach can be taken regarding probability: For a proposition to be probable is for it to be a fitting object of (high) credence - fitting given the available evidence. To put it otherwise, to be probable is to be credible. Indeed, many probability theorists, from Poisson onwards, did adopt this ‘epistemic’ interpretation of the concept of probability. J. M. Keynes’s A Treatise on Probability (1921) may be seen as the culmination of this development, even although its influence can also be discerned in later work (such as Carnap’s on the logical interpretation of probability). In Rabinowicz (2008, 2012), I proposed an FA-modelling of value relations that makes room not just for standard value relations (better, worse, equally good), but also for different types of value incommensurability. In this talk, based on Rabinowicz (2020), I will present a structurally similar modelling of probability relations. The modelling provides a new account of Keynesian incommensurable probabilities, which goes beyond Keynes in distinguishing between incommensurabilities of different types.

As compared with my earlier work on this topic (Rabinowicz 2017), the main new element is an argument that credence is a kind of pro-attitude, and that probability therefore is a kind of value and not merely a concept that is formally similar to value. The same goes for probability relations: they are value relations of a certain kind. Another new element is the discussion of a distinction between two versions of the FA-analysis of relations in general and of probability relations in particular. On one version, what determines such relations are comparisons between the degrees of pro-attitudes that are fitting toward different items; for example, whether it is fitting to favor one item more than the other. On the other version, value relations instead are determined by comparisons of the degrees of fittingness; for example, whether it is more fitting to favor one item than the other. I argue that the former version of the FA-analysis is considerably more plausible than the latter.