Your Preference or Mine? A Randomized Field Experiment on Recommender Systems in Two-sided Matching Markets
38 Pages Posted: 26 Apr 2021
Date Written: April 25, 2021
Research on recommender systems mainly focuses on product recommendations where the recommendations are primarily based on buyers’ preferences. However, user recommendation in two-sided matching markets where potential matches’ preferences may also play a role in the focal user’s decision-making, has not been systematically studied. Our study seeks to fill this gap by examining the impact of providing potential candidates’ preferences to focal users in user recommendations and exploring potential user heterogeneity. In collaboration with an online dating platform, we design and conduct a randomized field experiment where we present users with recommendations based on i) their own preferences, ii) potential matches’ preferences, or iii) mutual preferences. We find that while all three recommendations lead to an increase in message initiation, users choose candidates differently when recommendations include candidates’ preferences. Specifically, users tend to select people who prefer them despite the fact those candidates may not be the most desirable otherwise. This in turn leads to greater improvement in responses and matching. Interestingly, such differences in candidate selection within recommender systems do not spill over to selection in searching where preference information is not provided. We further observe heterogeneous effects wherein users who exhibit more diverse preferences and users who are more proactive are more sensitive to the matching side’s preferences. Our findings provide valuable insights on how to design user recommendation systems beyond the current practice, in which recommendations are primarily based on focal users’ preferences.
Keywords: Recommender systems, user recommendation, two-sided matching platforms, preference information, user heterogeneity, randomized field experiments
Suggested Citation: Suggested Citation