DECIPHERING ALGORITHMIC COLLUSION: INSIGHTS FROM BANDIT ALGORITHMS AND IMPLICATIONS FOR

The paradox of the current situation lies in the fact that it is possible to reconcile a high degree of price differentiation and variability with the firms' ability to identify distinct signals that allow for the identification of a focal point for possible coordination. [...] The principle behind Q-learning is the estimation of the value of a state-action pair, or Q-value, which reflects the expected utility of taking a specific action in a given state and then adhering to the optimal policy thereafter (Watkins & Dayan, 1992). [...] In the context of price signaling, the 'promise' is the suggested future price, and the 'performance' is the adherence to this price by competitors, thereby creating a non-verbal collusive environment. [...] The challenge for lawmakers and regulators is to craft legislation and guidelines that can effectively address the nuanced and technical nature of such collusion without stifling innovation and the legitimate use of algorithms to improve market efficiency. [...] Guidelines on the applicability of Article 101 of the Treaty on the Functioning of the European Union to horizontal co-operation agreements.