- This event has passed.
PhD Student Session: Joe Roussos & Bastian Steuwer
23 October 2019, 4:30 pm – 6:00 pm
Bastian Steuwer: “Aggregation, Balancing, and Respect for the Claims of Individuals”
Abstract: Most non-consequentialists share a strong resistance to the aggregation of harms across different individuals. Most of these skeptics regarding aggregation would, however, still like the numbers to count when one can save either a lesser or greater number from equal or similar harm. Limited Aggregation is an approach that reconciles these two commitments. It is motivated by a powerful idea: our decision whom to save should respect each person’s separate claim to our help; in particular it should respect those in need whose claims are the greatest. I develop a new view on aggregation which I call Hybrid Balance Relevant Claims . This view is well-grounded in the reasons we have to be skeptical of aggregation in the first place. It is more faithful to the motivations for limited aggregation than alternative views, like Aggregate Relevant Claims. My view can respond to recent challenges that limited aggregation, in particular Aggregate Relevant Claims, cannot be extended to cases where not all group members face the same plight. I show how all these problems can be resolved by paying greater attention to the underlying rationale for limited aggregation.
Joe Roussos: Expert deference as a belief revision schema
In Bayesian epistemology, there is an unresolved question of the proper way for agents to respond to expert testimony. The orthodox Bayesian insists on “supra-Bayesianism”–i.e., that the agent should update by conditioning on the expert’s report. A more idealised, but seemingly simpler, approach is to insist that the agent adopts a stance of “expert deference”. An agent defers to an expert when she changes her own credence to match the expert’s reported probability. This is made to fit with Bayesian conditioning by ensuring the agent has the right priors, i.e., that for any X, given that the expert reports probability p for X, the agent’s conditional probability for X is also p. I argue that each of these approaches is overly-demanding and doesn’t fit our intuitive picture of what happens in expert testimony cases.
I propose an alternative view of expert deference, on which it is seen not as a constraint on priors but as a belief revision schema. I argue that it allows for a much simpler model that better captures what is going on in cases of expert deference. The result is a more “user ready” model that still obeys widely endorsed Bayesian norms.