Players in pricing games with complex structures are increasingly adopting artificial intelligence (AI) aided learning algorithms to make pricing decisions for maximizing profits. This is raising concern for the antitrust agencies as the practice of using AI may promote tacit algorithmic collusion among otherwise independent players. Recent studies of games in canonical forms have shown contrasting claims ranging from none to a high level of tacit collusion among AI-guided players. In this paper, we examine the concern for tacit collusion by considering a practical game where EV charging hubs compete by dynamically varying their prices. Such a game is likely to be commonplace in the near future as EV adoption grows in all sectors of transportation. The hubs source power from the day-ahead (DA) and real-time (RT) electricity markets as well as from in-house battery storage systems. Their goal is to maximize profits via pricing and efficiently managing the cost of power usage. To aid our examination, we develop a two-step data-driven methodology. The first step obtains the DA commitment by solving a stochastic model. The second step generates the pricing strategies by solving a competitive Markov decision process model using a multi-agent deep reinforcement learning (MADRL) framework. We evaluate the resulting pricing strategies using an index for the level of tacit algorithmic collusion. An index value of zero indicates no collusion (perfect competition) and one indicates full collusion (monopolistic behavior). Results from our numerical case study yield collusion index values between 0.14 and 0.45, suggesting a low to moderate level of collusion.

Parallel pricing refers to the practice of firms making independent pricing decisions to match each other’ prices. This paper provides plausible economic intuitions about the market characteristics that affect the emergence and persistence of airline parallel pricing behavior by reviewing and synthesizing the relevant factors identified in the industrial organization literature as influencing cooperative outcomes or tacit collusion. One of the key findings of this paper is that the combination of high price transparency, quick price response, and repeated interactions among competing airlines can make parallel pricing as a self-evident outcome, even in the presence of information asymmetries, cost discrepancies, variations in capacity and service quality, as well as fierce market competition. Another finding is that as personalized pricing becomes more prevalent, the likelihood of success of parallel pricing may decrease. It remains unclear what other forms of cooperative outcomes will emerge as the industry continues to adapt to changing technological and market landscapes. With these findings, this paper contributes to the groundwork for developing a comprehensive model of parallel pricing in the airline industry.

This study examines how the digital economy transforms business competition and proposes sustainable solutions for emerging challenges. Using an interdisciplinary approach combining economic theory, legal frameworks, and technological assessment, it identifies four key disruptions to traditional competition: algorithmic collusion enabling price-setting without explicit agreements; data power abuse creating market entry barriers through strategic data control; self-preferencing practices allowing platforms to favor their services over competitors; and “killer acquisitions” eliminating future threats by acquiring potential competitors. Results reveal critical gaps in current competition laws, including difficulties defining relevant markets for digital services with network effects, limitations of traditional price-based analysis in zero-price markets, challenges measuring consumer welfare in “free” services, and tensions between protecting competition and encouraging innovation. For Vietnam as an emerging digital economy, the study recommends a comprehensive regulatory strategy: establishing a Digital Economy Unit within the Vietnam Competition Authority; expanding market dominance criteria beyond market share to include data control and network effects; introducing specific anti-competitive behavior rules; implementing transaction-value thresholds for merger control to capture high-value, low-revenue acquisitions; and adopting flexible regulatory approaches for technological change. These recommendations balance fair competition with innovation encouragement. The analysis is particularly relevant to dual transformation toward digitalization and sustainability, as digital platform concentration can either promote or hinder green innovation and sustainable development. The study demonstrates competition law reform's crucial role in supporting both digitalization and environmental sustainability objectives, with significant implications for Vietnam’s broader sustainability agenda.

Autonomous pricing algorithms are increasingly influencing competition in digital markets; however, their behavior under realistic demand conditions remains largely unexamined. This paper offers a thorough analysis of four pricing algorithms—Q-Learning, PSO, Double DQN, and DDPG—across three classic duopoly models (Logit, Hotelling, Linear) and under various demand-shock regimes created by auto-regressive processes. By utilizing profit- and price-based collusion indices, we investigate how the interactions among algorithms, market structure, and stochastic demand collaboratively influence competitive outcomes. Our findings reveal that reinforcement-learning algorithms often sustain supra-competitive prices under stable demand, with DDPG demonstrating the most pronounced collusive tendencies. Demand shocks produce notably varied effects: Logit markets suffer significant performance declines, Hotelling markets remain stable, and Linear markets experience shock-induced profit inflation. Despite marked changes in absolute performance, the relative rankings of the algorithms are consistent across different environments. These results underscore the critical importance of market structure and demand uncertainty in shaping algorithmic competition, while also contributing to the evolving policy discussions surrounding autonomous pricing behavior.

<p>The increasing use of artificial intelligence in strategic market environments has raised concerns about the emergence of algorithmic collusion. Recent studies have shown that reinforcement learning (RL) agents can learn to sustain supra-competitive prices in repeated pricing games without explicit communication or coordination, posing significant challenges for competition policy and antitrust regulation. This paper investigates the mechanisms underlying collusive behavior in multi-agent reinforcement learning environments and proposes algorithmic modifications designed to promote convergence toward competitive Nash equilibria.</p> <p>We begin by analyzing a repeated duopoly pricing game formulated as a potential game. Despite the favorable equilibrium properties of this class of games, we demonstrate that standard tabular Q-learning agents frequently converge to collusive outcomes rather than competitive equilibria. To address this problem, we introduce Smooth Q-Learning, a novel extension of Q-learning inspired by the Smooth UCT Search framework. The algorithm incorporates stochasticity by partially conditioning action selection on observed competitor behavior, thereby improving exploration and reducing the tendency toward collusive strategies. We further enhance this approach by integrating a planning component, resulting in the Smooth Dyna-Q algorithm, which combines model-based simulation with reinforcement learning to accelerate convergence.</p> <p>Simulation results show that both algorithms reliably identify Nash equilibria and substantially reduce the incidence of collusive behavior. Smooth Dyna-Q achieves particularly strong performance, exhibiting faster convergence and greater stability than standard Q-learning. To evaluate robustness beyond potential games, we test the algorithms in the repeated Bertrand pricing environment of Calvano et al. (2020), where conventional Q-learning is known to converge to tacit collusion. The proposed algorithms continue to perform effectively, although stronger smoothing parameters are required to ensure equilibrium convergence in more complex environments.</p> <p>The findings suggest that algorithmic collusion is not an unavoidable consequence of reinforcement learning and that carefully designed learning mechanisms can promote competitive market outcomes while preserving adaptive decision-making capabilities.</p>