Peking University, Apr 17, 2026: How should platforms like Netflix or Amazon decide what to recommend? A new study shows that by carefully designing how information is revealed to users, platforms can encourage earlier participation, improve social learning, and ultimately enhance overall welfare. The research, titled “Information Design for Social Learning on a Recommendation Platform,” was conducted by Assistant Professor Lyu Chen from Peking University HSBC Business School and published recently in the Journal of Economic Theory.

Why it matters
In the digital age, platforms like Netflix or Amazon improve their recommendations by learning from users' likes, ratings, and reviews, also known as "social learning." However, a common problem is that users tend to wait for others to try new products first. This leads to a lack of early feedback, which makes it harder for platforms to generate accurate recommendations.

Key Findings
The study finds that the best recommendation strategy follows a "U-shaped" pattern over a product's lifecycle. At the beginning, when a product is new and information is limited, platforms need to keep high standards for recommendations, because users tend to be cautious and may not trust what they see. As more data is collected over time, users become more confident in the platform, so it can gradually lower its standards and encourage more users to try the product.

When negative feedback appears, the platform may temporarily stop recommending the product. However, this does not mean the product is ruled out; it may be reintroduced later to gather more feedback.

Future Implications
The study shows that when user feedback is more accurate, or when a product has a larger potential user base, platforms can adopt lower recommendation thresholds throughout the product’s lifecycle.

Moreover, while previous research suggests that optimal recommendations may cycle between recommending and withholding when systems have limited memory of past information, this study instead assumes no such memory constraints.

*This article is featured in PKU News "Why It Matters" series. More from this series.
Read more: https://www.sciencedirect.com/science/article/pii/S0022053126000141

Written by: Ma Xuan
Edited by: Wong Jun Heng
Source: PKU News (Chinese)