Speaker: Martino Banchio
Title: Learning-Robust Mechanisms and the Coordination Bias
We develop a theoretical model to study strategic interactions between adaptive learning algorithms. Applying continuous-time techniques, we uncover the mechanisms responsible for recently documented experimental evidence of such algorithms learning to charge supracompetitive prices. We show that inadvertent coupling between the algorithms’ estimates leads to periodic coordination on actions that are more profitable than static Nash equilibria. One parameter, the relative learning rate across different actions, determines the presence and extent of coordination. Equalizing learning rates across actions restores competition: counterfactual evaluations, facilitated by ex-post feedback, sidestep inadvertent coordination. We formalize feedback provision with the notion of learning-robust mechanisms: mechanisms that guarantee implementation even when algorithms play. We prove existence and uniqueness of the optimal learning-robust mechanism, which communicates personalized menu prices in a class of canonical market design problems.
As a reminder, this talk is virtual only and it starts sharply at 1:05pm. The permanent Zoom link for this academic year is below.