Navigation menu

Algorithms versus Friends in Online Content Recommendations: A Large-scale Field Experiment 2022-11-23

Subject:Algorithms versus Friends in Online Content Recommendations: A Large-scale Field Experiment

GuestYifan Jiao, Ph.D. candidate, HKU Business School

Host:Liu Jialu, Assistant Professor, ACEM-SJTU

Time:Friday, Nov 25, 2022 14:00-15:30

Venue:Tencent Meeting

(Please send email to for meeting number and password.)


Information recommendation is fundamental to online economic activities. Content platforms intensively rely on algorithms or social networks to recommend and personalize online content for their users. Nevertheless, little is known about the different impacts of these two dominating recommendation mechanisms on user engagement with the content or the platform. The differences in content recommended and whether social cues are present constitute two essential drivers of their different impacts. We, therefore, conducted a large-scale field experiment involving over 2.1 million users and over 6.8 million items (online content) recommended by algorithms or friends on WeChat. We randomly assign users into three groups: users in Group I are exposed to the content recommended by algorithms only; those in Group II are primarily exposed to the content recommended by friends (i.e., contacts on WeChat); whereas Group III uses the same design as Group II without displaying social cues (the names of friends who shared the content) in friend-recommended content. We find that algorithmic recommendation leads to a higher content click-through rate and dwell time, while friend recommendations exhibit a larger content share rate and platform retention (measured as the number of platform login days). The content difference is the main driver for the higher content click-through rate and dwell time of algorithmic recommendations and for the larger content share rate of friend recommendations. The effects of social cues (influence) contribute to the larger platform retention associated with friend recommendations. We further demonstrate that the performance differences between algorithms and friends recommendations also depend on the characteristics of users and content. Our findings have rich implications for the design and management of online content recommendation mechanisms.


Yifan Jiao is a Ph.D. candidate in the Department of Innovation and Information Management at HKU Business School. Her research primarily focuses on the digital platforms, and she is interested in investigating practical questions that help businesses engage and retain customers and design monetization strategies on social media and online game platforms. Her research has been published in Production and Operations Management (POM) and Service Science. She also has several papers under revision at Manufacturing & Service Operations Management (M&SOM) or under review at Management Science.