Thu Jan 28, 2021
Deference and Disagreement
How should we revise our beliefs given one or multiple experts’ opinions? One intuitive answer is that: we should defer to the experts individually (Individual Deference), and update on their opinions by linear averaging (Linear Averaging). However, it is well-known in the literature that Individual Deference and Linear Averaging are jointly incompatible if the agent is uncertain whether the experts will agree. In this talk I consider a principle weaker than Linear Averaging, and prove that it is also incompatible with Individual Deference, given prior uncertainty about expert agreement. I argue that, in light of this inconsistent triad, we should question the assumption that there are formal rules for how rational agents should revise their beliefs given one or multiple experts’ opinions. Updating is hard, and there may be no formal shortcuts.