On Algorithmic Collusion and Reward-Punishment Schemes

A booming literature describes how artificial intelligence algorithms (AIAs) may autonomously learn to generate supra-competitive profits. Key to the widespread interpretation of the phenomena as ``collusion'', is the observation that the unilateral price cuts of an agent are followed by several periods of low prices and profits for both agents, which is construed as the signature of a reward-punishment scheme. We observe, however, that simple invitations to collude such as price increments towards the monopoly price are also followed by aggressive price wars. This raises doubt that what has been observed so far may not be collusion, but a failure to learn to compete