Beyond Mandatory Federations: Balancing Egoism, Utilitarianism and Egalitarianism in Mixed-Motive Games

Authors: Shaokang Dong, Chao Li, Shangdong Yang, Hongye Cao, Wanqi Yang, Yang Gao

AAAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments show that agents opting out of federation participation experience a reduction in egoism, and our approach outperforms multiple baselines in terms of both utilitarianism and egalitarianism. ... Experiments In this section, we demonstrate the superiority of FPF and LMF across various environments by addressing the following questions: (1) Can FPF and LMF outperform multiple baselines in balancing egoism, utilitarianism and egalitarianism? [in Figures 1, 4, 5, 6]. (2) Can FPF and LMF achieve empirical performance comparable to the mandatory participation D&F framework? [in Figure 2].
Researcher Affiliation Academia 1 School of Computer and Electronic Information, Nanjing Normal University, Nanjing, China 2 State Key Laboratory for Novel Software Technology, Nanjing University, Nanjing, China 3 School of Computer Science, Nanjing University of Posts and Telecommunications, Nanjing, China EMAIL, EMAIL, EMAIL.
Pseudocode Yes Algorithm 1: Flexible-Participation Federation (FPF) Algorithm 2: Local Multi-Federation (LMF)
Open Source Code Yes Code https://github.com/Shaokang-Agent/FPF-LMF.
Open Datasets Yes Classic Discrete Scenarios (Jiang and Lu 2019), including Job Scheduling, Matthew Effect, and Manufacturing Plant, are defined by resource constraints that drive agents to engage in dynamic competition or cooperation within mixed-motive games. ... Subsequently, we conduct experiments in Sequential Social Dilemmas (SSD), a mixed-motive game, as introduced in (Leibo et al. 2017; Hughes et al. 2018; Jaques et al. 2019). ... Finally, we evaluate the robust performance of LMF in fully cooperative MPE.
Dataset Splits No The paper describes various environments like Job Scheduling, Matthew Effect, Manufacturing Plant, Cleanup, Harvest, and MPE, but does not provide specific details on how the data within these environments were split for training, validation, or testing, nor does it reference standard splits or splitting methodologies.
Hardware Specification No The paper does not provide specific hardware details such as GPU/CPU models, processors, or memory used for running the experiments.
Software Dependencies No The paper does not list specific software dependencies, libraries, or solvers with their version numbers.
Experiment Setup Yes Input: The finite set N of N agents, participating set Np of Np agents, learning rate α, constrained coefficients λu for utilitarianism and λe for egalitarianism, entropy coefficient λh, update rate η, gradient error ν, replay buffer D = {Di}i N . Figures 1, 2, 4, and 6 show plots for FPF( u = 1, e = 0.1) and D&F( u = 1, e = 0.1), indicating specific hyperparameter values used in experiments.