Reproduce “Fair Agents: Balancing Multistakeholder Alignment in Multi-Agent Personalization Systems” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Andrea Forster, Peter Müllner, Denis Helic, Elisabeth Lex, Dominik Kowald

This paper introduces a conceptual framework for fair multi-agent personalization systems, which integrates LLM agents with social choice mechanisms to transparently balance competing stakeholder objectives. It provides a structured approach to address challenges in aligning abstract values, aggregating diverse fairness concerns, and evaluating system performance in complex multi-stakeholder environments.

Key insights

  • Fairness in multi-agent personalization requires a structured approach to align abstract stakeholder values with concrete, operational agent objectives.
  • Social choice mechanisms offer a theoretically sound and transparent alternative to opaque LLM-based negotiation for aggregating diverse preferences in multi-agent systems.
  • Effective fairness evaluation in multi-agent systems necessitates both quantitative metrics for agent and system performance, and qualitative assessment of real-world stakeholder perceptions.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.