Artificial Intelligence, Data and Competition

This paper examines how data inputs shape competition among artificial intelligences (AIs) in pricing games. The dataset assigns labels to consumers and divides them into different market segments, thereby inducing multimarket contact among AIs. We document that AIs can adapt to tacit collusion via market allocation. Under symmetric segmentation, each algorithm monopolizes a subset of market segments with supra-competitive prices while competing intensely in the remaining market segments. Market segments with higher WTP are more likely to be assigned for collusion. Under asymmetric segmentation, the algorithm with finer segmentation adopts a Bait-and-Restraint-Exploit strategy to "teach" the other algorithm to collude. However, the data advantage does not necessarily result in competitive advantage. Our analysis calls for a close monitoring of the data selection phase, as the worst-case outcome for consumers can emerge even without any coordination.