Algorithmic Decision-Making in Private Law

Algorithmic decision-making is becoming widely used in private law. Today, algorithms are routinely employed to screen job applicants, execute transactions, and freeze bank accounts. However, existing rules and doctrinal approaches do not always provide adequate responses to the legal challenges posed by algorithmic errors and the lack of transparency in algorithmic decision-making systems, particularly in cases where decisions are generated by artificial intelligence operating as a “black box.” In this article, the author sets out to examine how the delegation of legally significant decisions to increasingly autonomous algorithms affects private law relationships. The author analyzes whether the established doctrinal principles and legislative rules are capable of resolving the emerging problems, or whether they require conceptual reconsideration. The study draws on the provisions of Russian civil law and personal data legislation, the provisions and practice of the EU General Data Protection Regulation (GDPR), as well as domestic and international academic literature on algorithmic decision-making, algorithmic transparency and accountability, and legal safeguards designed to mitigate the risk of unlawful algorithmic decisions. It was found that algorithmic errors and the opacity of algorithmic decision-making can lead to a discrepancy between the data subject’s will and the will expressed through the algorithmic decision. The concept of the algorithm as a representative (agent) proves unsuitable for addressing the resulting legal issues. It is argued that algorithmic decisions should be classified as legal facts constituting either lawful or unlawful acts of the individual who delegated the process of decision-making to the algorithm. The lawfulness of automated decisions should be assessed on a par with other actions of legal subjects, while taking into account the subjective element of the subject’s conduct, including algorithmic errors and the measures taken by the subject during the implementation and use of algorithms. On this basis, the author proposes a reinterpretation of the concepts of intent, knowledge, and good faith in the context of algorithmic decision-making. It is argued that algorithmic transparency and accountability should be regarded as components of a risk management system and as prerequisites for monitoring the lawfulness of automated decisions. The right to object to an algorithmic decision is effective only provided that human review includes an analysis of the decision-making logic, as well as the identification and correction of errors that influenced the decision. At the same time, judicial review is not a universal remedy against automated decisions, since the proper subject of litigation can be the unlawful conduct of the party responsible for the algorithmic decision and specific remedies applied in individual situations.