EURO 2025 Leeds
Abstract Submission

2506. DMACoop: A Multi-Agent Hierarchical Coopetition Framework for Enhanced Investment Decision-Making

Invited abstract in session TB-28: Multi-Agent Systems and Reinforcement Learning for Decision Support, stream Decision Support Systems.

Tuesday, 10:30-12:00
Room: Maurice Keyworth 1.03

Authors (first author is the speaker)

1. Qing Yin
Alliance Manchester Business School, University of Manchester
2. Richard Allmendinger
Alliance Manchester Business School, The University of Manchester
3. Sam Veevers
ECI Partners
4. Xian Yang
Alliance Manchester Business School, University of Manchester

Abstract

In recent years, the integration of generative artificial intelligence (GenAI) into financial analysis and investment decision-making has attracted considerable interest. While existing methods have demonstrated promise through cross-functional agent collaboration aimed at achieving collective goals, they predominantly rely on a simplistic agent organizational structure (manager-analyst collaboration), often neglecting the complex dynamics of competition among multiple agents. This paper introduces the Director-Manager-Analyst Coopetition (DMACoop) framework, a novel hierarchical model that more accurately replicates the complex structures of real-world investment firms. The DMACoop framework facilitates optimal investment decisions by promoting cooperation and healthy competition among managers, guided by directors. Each manager leads their analyst teams to focus on specialized functions. Additionally, this framework strategically allocates specific risk-related responsibilities across various levels, enhancing risk-sharing across the hierarchy. We concentrate on the investment decision-making task within private equity firms by evaluating the investment potential of target companies. Our findings demonstrate that our multi-agent coopetition system significantly outperforms traditional collaborative models, delivering superior accuracy, efficiency, and adaptability in complex financial environments.

Keywords

Status: accepted

Back to the list of papers