To Lead or to be Led: A Generalized Condorcet Jury Theorem under Dependence

Juliette-Michelle BurkhardtJuliette-Michelle Burkhardt, Jonas KargeJonas Karge, Sebastian RudolphSebastian Rudolph, Dominik RusovacDominik Rusovac

Juliette-Michelle Burkhardt, Jonas Karge, Sebastian Rudolph, Dominik Rusovac
To Lead or to be Led: A Generalized Condorcet Jury Theorem under Dependence
AAMAS '24: Proceedings of the 2024 International Conference on Autonomous Agents and Multiagent Systems, to appear

Details
Bibtex

KurzfassungAbstract
Aggregating pieces of information or beliefs held by (abstract) agents is central to a variety of belief merging applications. When the merging process aims at retrieving an underlying ground truth, the Condorcet Jury Theorem (CJT) allows identifying voting rules that almost surely track the true piece of information for large groups of agents, given that specific conditions are met. As essential assumptions, the CJT relies on all agents being equally competent as well as independent from one another. In the search for a generalization of the CJT applicable to real-world scenarios, both aforementioned assumptions were weakened separately. In this work, we provide a generalization of the CJT that allows, at the same time, for heterogeneous competence levels across agents as well as a degree of dependence modeled through an opinion leader exerting influence on the electorate. Additionally, we derive a concrete bound on the number of agents necessary to successfully track the underlying ground truth, and examine the bound’s tightness by means of statistical simulations.
Projekt:Project: SECAI
Forschungsgruppe:Research Group: Computational LogicComputational Logic

@inproceedings{BKRR2024,
  author    = {Juliette-Michelle Burkhardt and Jonas Karge and Sebastian Rudolph
               and Dominik Rusovac},
  title     = {To Lead or to be Led: A Generalized Condorcet Jury Theorem under
               Dependence},
  booktitle = {AAMAS '24: Proceedings of the 2024 International Conference on
               Autonomous Agents and Multiagent Systems},
  year      = {2024}
}