Background: With the increasing use of AI for decision-making, more and more situations arise in which multiple algorithms make decision in the same environment and therefore influence the outcome of each other’s decisions. This is especially true for algorithms that determine prices in non-monopolistic markets. Think, for example, of the competition between webshops, supermarkets and gas stations. In such situations, the algorithms may learn to work together to push up prices and increase their profits. To prevent this from occurring, governments have passed legislation which makes this collusion illegal when executed by managers. Unfortunately, the legislation may not apply to collusion by algorithms. Academic literature on the topic has focussed predominantly on showing that such collusion is possible using simulations and theory.
Topic: In this project, you will use insights gained from the recent literature on algorithmic collusion to design and test techniques for detecting the existence of algorithmic collusion from pricing data gathered by the CBS. This is a topic that simultaneously has high societal relevance and uses theory at the forefront of the field.
CBS: The Centraal Bureau voor de Statistiek or in English “Statistics Netherlands” describes itself in the following text:
“As a society, we want rapid information based on reliable figures so we know what is going on in the world around us. Statistics Netherlands (CBS) provides relevant and independent figures on a wide range of societal issues. This requires a high degree of flexibility from CBS, something our staff work hard to achieve on a daily basis. Whether the issue is climate change, sustainability, the housing challenge or poverty, we respond to the need for transparent and accessible information.”
This project will be done in collaboration with the consumer prices index team at the CBS and Frank Pijpers from the methodology section of the CBS.
Contact: For more information please get in touch with Janusz Meylahn at j.m.meylahn@utwente.nl