Talk 1 on Human-AI Cooperation for Fairness Elicitation
Date:
The talk delves into the complex notion of fairness and the difficulties involved in mathematically quantifying it for decision-makers. We investigate strategies for eliciting individuals’ fairness concepts using straightforward inquiries and comparisons. To express fairness concepts mathematically, we adapt the power mean function family. The introduction of the supremum distance metric $\Delta$ allows us to evaluate the maximum disparity between concepts in diverse scenarios. Moreover, we outline two approaches to formulating questions and establishes an upper threshold for the supremum distance metric $\Delta$ between any pair of fairness concepts.