The threshold of belief and the value of punishment
Julien Dutant (King’s College London)

05 March 2021, 16:00 | Online, via Zoom

Abstract: This paper explores a tension between two putative norms for rational reactions (i.e. reactive attitudes like blame or anger and retributive actions like punishing). The first (Reactive Risk Management) says that how it is rational to react to an (apparent) deed depends on how confident it is rational to be that the deed was done. The second (Reasoned Reactions) says that it is rational to react to an (apparent) deed if, and only if, it is rational to believe that it was done. The two conflict in the context of the Fallibilist idea that rational belief does not require certainty. A well-known source of conflict between them is the problem of ‘naked statistical evidence’ (Buchak 2014). However, one can make the norms compatible even in naked statistical evidence cases by rejecting ordinary intuitions (Laudan 2012, Papineau 2019), or by claiming that reactions have epistemically-sensitive values: namely, that they have no positive value if not done on the basis of knowledge (Littlejohn 2018). This paper considers a separate source of conflict between the two norms: the problem of single-case threshold variance. When one is facing a choice over a range of potential reactions, the level of confidence that rationalizes one reaction appropriate to a deed may differ from that of another reaction appropriate to that deed. This entails that one of the two norms fails. The problem affects even the views that reconcile the two norms with naked statistical evidence. The problem would be avoided if a certain hypothesis, which I call the “Blackstone invariance hypothesi”, was true. Unfortunately, I don’t see much prospect for the hypothesis to hold. I conclude with some challenges to meet if we instead give up one of the three ideas that generate the problem.

Free Attendance, but preregistration required: https://www.lancog.com/lancog/registration/