Pricing algorithms on digital platforms: between collusion and innovation

Abstract

Objective: this article examines the impact of self-learning pricing algorithms on Brazilian competition law. It further explores how the agents behind anticompetitive conducts resulting from tacit collusion, caused without human intervention, should be analyzed and held accountable by the antitrust authority.

Method: the methodology was based on the narrative analysis of international literature on collusive
behaviors mediated by autonomous learning systems. Guidelines issued by antitrust authorities
and institutions from various jurisdictions, case law from Cade, Brazil’s competition authority, and
the European Union’s AI Act were examined with the aim of proposing normative parameters for
preventive action in Brazil.

Conclusions: the study finds that only autonomous learning algorithms—specifically machine learning and deep learning, that lead to "predictable agent" and "digital eye" forms of tacit collusion require a revamp of the Brazilian competition authority's traditional investigative mechanisms. Any investigation involving autonomous algorithmic collusion should be conducted under the rule of reason, considering (i) the absence of precedents involving algorithmic collusion without human intervention, both in Brazil and other jurisdictions, (ii) the pro-competitive potential and efficiency gains provided by pricing algorithms, and (iii) the low likelihood, so far, of algorithmic collusion without human intervention occurring in practice. In addition, Cade should adopt a preventive posture by issuing guidelines or recommendations on the topic, given the current normative, institutional, and technological immaturity in detecting algorithmic collusion in the Brazilian jurisdiction.