Conference Agenda

In economic theory it is typically assumed that agents choose their actions based on an equilibrium strategy. Determining this equilibrium behavior is crucial, but understanding the inverse problem, where observed actions are used to infer the underlying valuations leading to such behavior, is equally significant. The field of structural estimation employs statistical techniques to explore this issue. Initial methods were limited to simple settings due to the computational hardness of the equilibrium problem. Further developments in this field have bypassed the need for an analytical equilibrium by representing the latent parameters as a function of the observed actions. Although this task is not as hard as deriving the equilibrium strategy analytically, it is still challenging. Thus, these approaches are limited to rather simple auction settings. In this study, we draw on recent advances in equilibrium learning to address these challenges. Therefore, it is only necessary to specify the underlying game and not perform complex transformations. We introduce an estimation framework that utilizes Normalizing Flow models to represent the latent parameters, such as prior distributions in auction games. This framework iteratively optimizes these models by comparing the induced equilibrium strategy with observed data. Using laboratory data, we demonstrate that our framework can accurately recover latent parameters across various settings. This capability allows us to analyze scenarios previously unexplored, thus providing a novel tool to infer valuations from bid data.

Automated algorithms are increasingly used for decision-making, raising questions about potentially harmful outcomes such as reduced customer welfare. Cooperation by algorithms in the pricing context is known as tacit or algorithmic collusion.

In the light of algorithmic pricing, recent findings suggest that Reinforcement Learning algorithms such as Q-learning tend to set prices above the Nash equilibrium level.

We analyze the behavior of simple independent online learning algorithms with bandit feedback in Bertrand oligopolies. The agents compete repeatedly against each other by setting prices and update their beliefs about the market and their actions over time.

Our empirical observations suggest that most bandit algorithms reliably converge to prices at or close to the Nash equilibrium. In particular, signs of collusion vanish when pairing different algorithms. In line with these findings, we can provide convergence guarantees for the class of mean-based algorithms in some oligopoly pricing models. These findings indicate that multiple independent learners can approximate certain Nash equilibria of this class of games.

We propose a unifying framework for smoothed analysis of combinatorial local optimization problems, and show how a diverse selection of problems within the complexity class PLS can be cast within this model. This abstraction allows us to identify key structural properties, and corresponding parameters, that determine the smoothed running time of local search dynamics. We formalize this via a black-box tool that provides concrete bounds on the expected maximum number of steps needed until local search reaches an exact local optimum. This bound is particularly strong, in the sense that it holds for any starting feasible solution, any choice of pivoting rule, and does not rely on the choice of specific noise distributions that are applied on the input, but it is parameterized by just a global upper bound φ on the probability density. The power of this tool can be demonstrated by instantiating it for various PLS-hard problems of interest to derive efficient smoothed running times (as a function of φ and the input size).

Most notably, we focus on the important local optimization problem of finding pure Nash equilibria in Congestion Games, that has not been studied before from a smoothed analysis perspective. Specifically, we propose novel smoothed analysis models for general and Network Congestion Games, under various representations, including explicit, step-function, and polynomial resource latencies. We study PLS-hard instances of these problems and show that their standard local search algorithms run in polynomial smoothed time.

This study examines the strategic dynamics of Blockchain Platform Providers (BPPs) and Blockchain Service Providers (BSPs) through a simultaneous move game lens, emphasizing their interactions in the blockchain ecosystem. By analyzing the critical decision variables of pricing and quality of security, we analyze how the actors optimize end-user acquisition and overall profitability. Utilizing Nash equilibrium, the paper identifies optimal strategies that balance competitive pricing with investments in security, a trade-off that is crucial for enhancing platform attractiveness and operational efficiency.

The findings elucidate the pivotal role that strategic interdependencies play in shaping the blockchain environment, with implications for both theoretical exploration and practical application in technology management. By providing a nuanced understanding of the factors that drive decision-making among BPPs and BSPs, this research contributes to the strategic literature on blockchain technology management and offers a foundational perspective for economic actors in this rapidly evolving sector. The paper further discusses the broader implications of these strategies for market competition and technological innovation within the blockchain industry.