ResearchTrend.AI
  • Papers
  • Communities
  • Events
  • Blog
  • Pricing
Papers
Communities
Social Events
Terms and Conditions
Pricing
Parameter LabParameter LabTwitterGitHubLinkedInBlueskyYoutube

© 2025 ResearchTrend.AI, All rights reserved.

  1. Home
  2. Papers
  3. 2210.12841
8
1

A Cooperative Reinforcement Learning Environment for Detecting and Penalizing Betrayal

23 October 2022
Nikiforos Pittaras
ArXivPDFHTML
Abstract

In this paper we present a Reinforcement Learning environment that leverages agent cooperation and communication, aimed at detection, learning and ultimately penalizing betrayal patterns that emerge in the behavior of self-interested agents. We provide a description of game rules, along with interesting cases of betrayal and trade-offs that arise. Preliminary experimental investigations illustrate a) betrayal emergence, b) deceptive agents outperforming honest baselines and b) betrayal detection based on classification of behavioral features, which surpasses probabilistic detection baselines. Finally, we propose approaches for penalizing betrayal, list directions for future work and suggest interesting extensions of the environment towards capturing and exploring increasingly complex patterns of social interactions.

View on arXiv
Comments on this paper