ARTIFICIAL INTELLIGENCE AND ALGORITHMIC COLLUSION (AIAC) – AIAC

The rise of AI and automated pricing algorithms prompts questions about their market impact. It's hypothesized that these algorithms could facilitate collusion by optimizing utility, much like rational economic agents. This potential has been explored in contexts such as the European Digital Markets Act, and investigated by the UK’s CMA and the US FTC. This proposal aims to illuminate the determinants of algorithmic collusion through detailed simulations and empirical analysis, focusing on two research strands: "When Do Algorithms Collude?" and "Algorithms in the Wild."

The "when do algorithms collude" strand combines Economics and Computer Science to explore algorithmic behavior. While algorithms are often assumed to act as rational agents, their self-optimizing nature and external influences challenge this notion. The research focuses on analyzing algorithms’ potential for collusion, particularly in the context of the prisoner's dilemma, exploring the balance between cooperation and non-cooperation. Collaboration with applied mathematician Janucz Meylahn from the University of Twente provides rigorous theoretical guidance to design and implement simulations. Compared to the previous proposal, the focus is sharpened solely on algorithmic behavior, without comparing it to human actions, leveraging new collaborations and optimizing resources for postdoctoral research.

The strand "algorithms in the wild" examines the real-world implications of pricing software in e-commerce. By creating seller accounts and utilizing repricing software, it tests the intelligence of pricing software and its potential for collusive behavior. Building on substantial previous work, the project is poised to launch its first set of experiments. The requested funding will support the implementation of more sophisticated experiments.

Both research strands complement each other, with simulations enhancing theoretical understanding and real-world studies providing empirical data. This interplay ensures a comprehensive approach grounded in both theory and practice. Since last year’s proposal, advancements allow for synchronized progression toward more complex algorithms in both projects. Funding for a postdoctoral researcher will support the integration of advanced algorithms, boosting the project’s impact. This synergy positions the proposal to make significant contributions to developing informed regulatory frameworks and effective management practices for digital market challenges.