The Editor and the Algorithm: Returns to Data and Externalities in Online News

Abstract

We run a field experiment to quantify the economic returns to data and informational externalities associated with algorithmic recommendation relative to human curation in the context of online news. Our baseline results show that algorithmic recommendation can outperform human curation in terms of user engagement. We find significant heterogeneity with regards to the type of data (personal data is much more effective than aggregate data, but diminishing returns set in quickly), user types (users closer to average consumption patterns engage more), local data network effects (the size of the user base increases engagement only for users close to average consumption patterns), and data freshness (missing updating of personal data quickly deteriorates performance). We characterize circumstances where human curation leads to higher engagement than algorithmic recommendation to highlight complementarities between the human editor and the algorithm. Investigating informational externalities highlights that personalized recommendation reduces consumption diversity and that these are reinforced over time. Moreover, users associated with lower levels of digital literacy and more extreme political views engage more with algorithmic recommendations.

Keywords: Field experiment, Economics of AI, Returns to data, Filter Bubbles