Multi-metric analysis of demographic fairness in ML reveals different fairness metrics produce conflicting assessments on the same system due to capturing distinct statistical properties. Using face recognition experiments, demonstrates that fairness evaluation reliability depends critically on metric choice, challenging assumptions of consistency.
Demonstrates that fairness can emerge as a property of multi-agent collaboration, potentially circumventing Arrow's impossibility theorem limitations in collective decision-making. This theoretical contribution suggests that distributed AI systems might achieve fair outcomes through collaboration mechanisms that single-agent or voting-based systems cannot.
LLM multi-agent systems spontaneously develop power-law distributions in cognitive influence, forming "intellectual elites" where a small fraction of agents disproportionately shape collective decisions without explicit design. This emergent stratification mirrors human social dynamics and challenges assumptions about egalitarian multi-agent collaboration. Critical implications for fairness and reliability in decision-making systems.