Deciphering Algorithmic Collusion: Insights from Bandit Algorithms and Implications for Antitrust Enforcement

This paper explores algorithmic collusion from both legal and economic perspectives, underscoring the increasing influence of algorithms in firms' market decisions and their potential to facilitate anti-competitive behavior. By employing bandit algorithms as a model—typically used in uncertain decision-making scenarios—we shed light on the mechanisms of implicit collusion that occur without explicit communication. Legally, the primary challenge lies in detecting and categorizing possible algorithmic signals, particularly when they function as unilateral communications. Economically, the task of distinguishing between rational pricing strategies and collusive patterns becomes increasingly complex in the context of algorithm-driven decisions. The paper stresses the need for competition authorities to identify atypical market behaviors. Striking a balance between algorithmic transparency and the prevention of collusion is critical. While regulatory measures could mitigate collusive risks, they might also impede the development of algorithmic technologies. As this form of collusion gains prominence in competition law and economics discussions, understanding it through models like bandit algorithms becomes essential, especially since these algorithms have the potential to converge more rapidly toward supra-competitive prices equilibria