Enhancing multi-agent cooperation: the implementation and impact of a communication module in AI systems

As artificial intelligence increasingly permeates everyday life, the development of Large Language Models (LLMs) capable of effectively simulating complex social interactions and fostering cooperative behavior has become paramount. This paper explores the enhancement of cooperative capabilities in LLM-augmented Autonomous Agents (LAAs) through the integration of a dedicated communication module. Initial findings indicate that this module significantly enhances collaborative outcomes by enabling agents to share observations and insights, thereby fostering a more comprehensive understanding of their environment. This enhanced understanding has resulted in a 45% increase in per capita reward. Moreover, improved communication has facilitated more sustainable resource management and refined behavioral predictions among agents, leading to more strategic planning in multi-agent systems. Furthermore, the establishment of consensus through communication has proven invaluable in coordinating actions and promoting increased cooperation. However, the research also uncovers that while inter-agent communication can augment cooperation, it introduces risks of bias and misinformation.

• Autores: Ortiz Almanza, David Santiago
• Tipo de recurso: Trabajo de grado de pregrado
• Fecha de publicación: 2024
• Institución: Universidad de los Andes
• Repositorio: Séneca: repositorio Uniandes
• Idioma: eng
• OAI Identifier: oai:repositorio.uniandes.edu.co:1992/74464
• Acceso en línea: https://hdl.handle.net/1992/74464
• Palabra clave: Communication; Cooperation; Artificial intelligence; Large language models; Melting Pot; Ingeniería
• Rights: embargoedAccess
• License: Attribution-NonCommercial 4.0 International