This paper studies the incentive of a social media platform (SMP) to increase polarization of its user network. I propose a two-group network model where the SMP earns revenue from user-data driven personalization. The objective of the SMP is to maximize the amount of valuable data generated. To this end, it relies on an algorithm that, at a cost, encourages users to form new links. Within a microfounded model, I show that two opposite forces impinge on the SMP. 1) The relative-size effect incentivizes the SMP to increase polarization since this increases amount of data it gathers. 2) The diversification effect incentivizes the SMP to decrease polarization since this increases value from data. Balancing these two forces, the platform decides the optimal level of polarization it induces. Overall, the result provides a rationalization for opposite empirical results concerning the effect of an SMP on polarization. Further, the SMP aggravates inefficiencies relative to polarization that maximizes user welfare. Finally, if users prefer interacting with same-group linked users, the SMP internalizes this heterogeneity and has a greater incentive to increase polarization.
The figure shows the effect of the SMP for α = 4. When initial polarization is lower than 2.125, the SMP increases polarization, so the final polarization (in blue) is higher than the initial polarization (in red). When the initial polarization is higher than 2.125, the SMP decreases polarization, so the final polarization is lower than the initial polarization.
This paper studies the regulation of a social media platform (SMP), focusing on the tension between platform incentives and user data control rights. The SMP earns revenue from data based personalized advertising, while users are connected on a network exhibiting data externalities. A microfounded model reveals a fundamental friction: users prefer limited data sharing, whereas unrestricted data collection maximizes platform revenue. Consequently, the SMP offers its service to collect user data and users simultaneously accept or reject the offer. Externalities dilute quality of service offered and excessive data is shared in equilibrium. Regulatory measures such as the data minimization principle restricts data access to the SMP and increases user welfare. In contrast, a strict consent policy akin to the GDPR burdens users with complete data control and decreases user welfare.
The figure shows that SMP payoff from personalized advertising, g(x), is increasing and concave in number of data points x analyzed about a user. User payoff, f(x), is non-monotonic and non-negative in number of data points x analyzed. This establishes the fundamental friction between an SMP and its users.
Segregation via Social Media Platforms
This paper studies the incentive of a firm to segregate users onto its different social media platforms (SMPs). I propose a network model with data externalities, where each SMP earns revenue from user-data driven personalization. A firm offers a price vector to each user to join either platform. Each user decides which platform to join and whether to share data. Users benefit from personalization but incur harm due to privacy concerns. Two factors affect the decision of the firm. 1) The firm has an incentive to segregate users as that reduces the price offered to a user. 2) The firm has an incentive not to segregate users as that increases the amount of data generated. Balancing these two factors, the firm determines a threshold privacy level and it segregates users when their privacy concerns are above this threshold. Finally, the results provide a policy intervention as a solution. When privacy concern of users is high, implementing a strict privacy policy reduces the likelihood of segregation by the firm.
Works in Progress
Optimal Disclosure to a Platform (joint with Nenad Kos)
Product Design under Regulation (joint with Olga Chiappinelli)
Assortative Mating (joint with Erick Stivens Padilla-Galviz and Montserrat Vilalta-Bufí)