InfluenceNet: AI Models for Banzhaf and Shapley Value Prediction

2025-03-11

AI Summary

Quantifying individual influence in multi-agent systems remains challenging, as classical power indices—Banzhaf and Shapley values—suffer from exponential computational complexity for large coalitions (n ≥ 10), hindering real-time analysis. Method: This paper introduces the first neural-network-based approach for power index estimation, proposing a supervised learning framework that integrates game-theoretic feature encoding, efficient coalition sampling, and a normalized value-regression loss—ensuring theoretical interpretability while enhancing scalability. Contribution/Results: Evaluated on standard voting game benchmarks, the method achieves a mean relative accuracy of 98.2% and accelerates inference by over 200× compared to Monte Carlo approximation. It enables sub-second influence analysis for coalitions of up to 100 agents, effectively breaking the traditional trade-off between accuracy and computational efficiency.

Technology Category

• Multiagent Systems: Adversarial Agents
• Game Theory and Economic Paradigms: Social Choice / Voting
• Cognitive Modeling & Cognitive Systems: Agent Architectures

Application Category

• Economics, Online Markets and Human Computation: Incentives in network design for Web infrastructures and ecosystems
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Social Networks and Social Media: Influence propagation, information diffusion, and the prediction on networks

Abstract

Power indices are essential in assessing the contribution and influence of individual agents in multi-agent systems, providing crucial insights into collaborative dynamics and decision-making processes. While invaluable, traditional computational methods for exact or estimated power indices values require significant time and computational constraints, especially for large $(nge10)$ coalitions. These constraints have historically limited researchers' ability to analyse complex multi-agent interactions comprehensively. To address this limitation, we introduce a novel Neural Networks-based approach that efficiently estimates power indices for voting games, demonstrating comparable and often superiour performance to existing tools in terms of both speed and accuracy. This method not only addresses existing computational bottlenecks, but also enables rapid analysis of large coalitions, opening new avenues for multi-agent system research by overcoming previous computational limitations and providing researchers with a more accessible, scalable analytical tool.This increased efficiency will allow for the analysis of more complex and realistic multi-agent scenarios.

Problem

• Efficiently estimates power indices for voting games using Neural Networks.
• Overcomes computational bottlenecks in analyzing large multi-agent coalitions.
• Provides scalable tools for complex multi-agent system research.

Innovation

• Neural Networks for power indices estimation
• Efficient analysis of large coalitions
• Superior speed and accuracy in predictions