Abstract:
In this talk, I will provide a cognitive/computational unpacking of the Condorcet Jury Theorem. The core of the talk will focus on a paper co-authored with Lu Hong, in which we construct a formal framework to study group accuracy on classification tasks. In the paper, we derive necessary and sufficient conditions for perfect accuracy and then characterize a group’s ability to make accurate classifications as a function of its diversity, size, and the accuracy of its members. For groups relying on the majority rule, we show that increasing individual accuracy produces a setwise increase in collective accuracy, but that, contrary to intuition, increases in group size do not. We further show that increasing diversity, as measured by pairwise disagreement, only guarantees increased accuracy for highly diverse groups. Finally, we derive a general condition for increasing group size to raise or lower expected accuracy assuming individuals drawn from a population. In addition to the paper, I will reframe the model using disciplinary lenses to discuss the potential for making good decision in complex environments.

(To attend this online event, please complete the form to receive email instructions and announcements for this and future SBEE Seminars.)