Strategic Risk, Artificial Intelligence and Collusive Crime

Collusive crime involves multiple, organized wrongdoers cooperating for extended periods, and ranges from market cartels to fraud, corruption, money laundering, and drugs and human trafficking.

Concerns about its increase and effects are mounting, as are policy-makers’ worries about potentially collusive effects of Artificial Intelligence (AI) as incorporated in pricing algorithms, increasingly widespread in modern markets.

This project aims at deepening our understanding of the determinants of cooperation/collusion for humans and algorithms, and of their implications for public policy. The proposal is organized into two related parts.

The first analyzes ‘strategic risk’, the risk of miscoordination, as both a determinant of human cooperation in repeated strategic interactions and as a more efficient deterrence channel than Becker’s (1968) expected fines or Stigler’s (1964) secret undercutting.

It includes theoretical and experimental analyses under perfect and imperfect monitoring, with and without communication, aimed at clarifying our understanding of this force as a potential driver of effective public policies against collusive crimes. The second focuses on the collusive effects of AI-based pricing algorithms.

Using simulation techniques and laboratory games, it explores whether these algorithms lead to market collusion in realistic environments where human subjects are also present, and studies subjects’ decision whether or not to delegate pricing to them.