Strategic Interactions in Multi-Level Stackelberg Games with Non-Follower Agents and Heterogeneous Leaders

A team of researchers including Niloofar Aminikalibar, Farzaneh Farhadi, and Maria Chli has published a significant paper on arXiv introducing a new multi-level Stackelberg game framework that addresses a critical blind spot in traditional economic and AI models. The key innovation is the formal inclusion of 'non-follower' agents—entities that do not participate in market competition (generating no revenue) but whose behavior, like using shared infrastructure, directly creates and adapts to congestion. The authors argue that ignoring these agents, as most existing models do, leads to systematically distorted equilibrium predictions in any congestion-coupled system, from traffic networks to cloud computing resources.

The researchers instantiate their three-level framework with heterogeneous leaders (differing in decision horizons), strategic followers, and non-followers in the context of electric vehicle (EV) charging infrastructure. Here, charging station providers are the competing leaders, EV drivers are followers, and regular non-EV traffic constitutes the non-followers. The model demonstrates how accounting for the congestion caused by general traffic qualitatively changes the strategic incentives for where to build charging stations and what prices to set. This framework provides a more accurate tool for simulating and planning complex multi-agent systems, with direct applications for urban mobility planning, energy grid management, and distributed computing resource allocation, leading to better real-world infrastructure decisions.