Popularity Signals in Optimizing Trial-Offer Markets

Online marketplaces, such as Itunes, Amazon, Orbitz, and many others, use popularity signals to optimize efficiency both from a customer and market standpoint. However, in the last decade or so, there has been a significant debate about the role of social influence on market efficiency, predictability, and fairness. In this talk, we study these issues for trial-offer markets under various models of customer behavior, popularity signals, and position bias. We show that popularity, when used properly, can bring significant benefits in market efficiency, without introducing unpredictability or inequalities. The talks covers both computational complexity results and asymptotic behavior of various policies for these markets, using techniques from computer science and stochastic approximation. Agent-based simulation results will also be presented.

Joint work with Andres Abeliuk, Manuel Cebrian, Franco and Gerardo Berbeglia, and Felipe Maldonado.