Disagreement Among Many: Why Numbers Do Not Matter

The equal weight view on peer disagreement holds that one should give the beliefs of one’s epistemic peers – roughly, those who are as likely as oneself to be right about matters in the relevant domain – the same weight that one gives one’s own belief. This view is standardly interpreted to mean that one should, in the case of many disagreeing peers, adopt a credence towards the thesis under debate that equals the arithmetic mean of the respective credences of all of one’s peers, including oneself. I argue that considerations concerning the cardinalities of sets of experts and concerning sampling show that this interpretation is wrong; they rather suggest that we should simply suspend belief (in a non-standard sense of ‘suspension of belief’ that will be explained). An important consequence is the epistemic insignificance of numbers: how many peers actually favour each position does not matter. This result holds true even if we assume that the relevant peers are experts whose mean probability of being right is greater than 0.5. It thereby seems to contradict the well-known Condorcet jury theorem; rightly understood, however, it does not.