Algorithms versus Friends in Content Recommendations: a Large-scale Randomized Field Experiment

Algorithms and friends are the two dominant mechanisms in online content recommendations and have significantly shaped the information that people are exposed to and consume daily. Theoretically, the content suggested by friends can be different from that recommended by the algorithms in both content quality and the presence of social influence, and thus perform differently in users’ engagement and retention. However, limited empirical evidence has causally identified such differences. We, therefore, conduct a massive-scale field experiment on WeChat involving more than 2.8 million users. We randomly assign the users into three groups: algorithmic recommendations, friends’ recommendations without social cues, and friends' recommendations with social cues.  Our results show that content engagement (i.e., click rate and stay time) is significantly larger under algorithmic recommendations, while social engagement (i.e., like rate) and retention are significantly higher under friends’ recommendations. We also observe that the content recommended by friends is more socially contagious. Social influence, the effects of social cues, significantly lift content engagement, social engagement, and retention. Furthermore, we delved into the mechanisms and find that data richness and the degree of homophily positively affect the content engagement in algorithmic and friends’ recommendations respectively. The effects of displaying social cues are affected by the degree of social engagement in WeChat Moment. Our findings have rich implications for the mechanisms and management of content recommendations by algorithms (AI) and friends (humans). 

Social Advertising Effectiveness Using a Large-Scale Randomized Field Experiment

Study 1: Social Advertising Effectiveness Across Products
Study 2: Public versus Private Responses to Social Advertising 

Study 1:

Most of the empirical evidence on social advertising effectiveness focuses on a single product at a time. As a result, little is known about how the effectiveness of social advertising varies across product categories or product characteristics. We, therefore, collaborated with a large online social network to conduct a randomized field experiment measuring social ads effectiveness across 71 products in 25 categories among more than 37 million users. We found some product categories, like clothing, cars and food exhibited significantly stronger social advertising effectiveness than other categories like financial services, electrical appliances, and mobile games. More generally, we found that status goods, which rely on status-driven consumption, displayed strong social advertising effectiveness. Meanwhile, social ads for experience goods, which rely on informational social influence, did not perform any better or worse than their theoretical counterpart search goods. Social advertising effectiveness also significantly varied across the relative characteristics of ad viewers and their friends shown in ads. Understanding the heterogeneous effects of social advertising across products can help marketers differentiate their social advertising strategies and lead researchers to more nuanced theories of social influence in product evaluation.

​

Study 2: 

​

One unique feature of social advertising is the coexistence of public and private consumer responses. Social media allows users’ certain responses (e.g., likes) to be publicly revealed to one's social network. More importantly, public responses can also influence private responses (e.g., clicks).  We, therefore, use data from a large-scale field experiment with a major social media platform (WeChat Moments) to investigate how the display of social cues (friends' likes) affects users' public (likes) and private responses (clicks) to social ads. We find that, on average, displaying the first social cue significantly enhances the liking rate and the clickthrough rate. Nevertheless, although showing additional social cues can further increase users' tendency to like an ad, it does not further increase the clickthrough rate. This empirical pattern is consistent with the interplay between informational and normative social influence in social advertising. Overall, we find that the coexistence of the two forces can enhance the conformity effect on the public liking response. However, when normative social influence dominates, a crowding-out effect on the private clicking response may occur. Our results have rich implications for advertisers and social media platforms in regard to the design of social advertising policies and social networks.

Emotions in Online Content Diffusion

To investigate the impact of emotions on the spread of online content, we analyzed a random sample of 387,486 online articles and their diffusion cascades, in which more than 6 million unique individuals shared the articles on a massive-scale online social network. We detected the degree of eight discrete emotions (i.e., surprise, joy, anticipation, love, anxiety, sadness, anger, and disgust) embedded in the content of each article with a newly generated domain-specific and up-to-date emotion lexicon. Our results suggest that articles with a higher degree of emotion generally reached a significantly larger number of individuals and diffused significantly more deeply, broadly, and virally but more slowly. Anxiety and love significantly increased cascade size, depth, breadth, and structural virality, whereas sadness significantly decreased these factors but accelerated article diffusion. We also find that the average age and network degree of the individuals, and especially the proportion of the weak ties involved in a cascade, significantly mediated the effects of emotions that lead to a differential diffusion process. Consistent emotions were detected between articles and the associated comments, confirming that readers well received the emotions expressed in the articles.

Do Online Social Referrals Lead to Better Customer-Product Matches: A Deep Learning Approach

New social media has been enlarging the impact of social referrals on product diffusion. However, there is no direct evidence on how effectively referrers match products with referred customers leading to high returns of online social referrals. To answer this question, we analyzed 158,421 referrals of 25,709 instant apps from 12,051 referrers to 91,995 recipients (those referred), 12,051,000 contacts randomly selected from referrers' local social networks, and random users of the apps. We leveraged a state-of-the-art graph embedding framework to estimate users' latent preferences for the apps based on a massive amount of user-app historical usage sequences. We find that referrers recommend products to contacts with substantially stronger preferences for the referred products than the non-referred contacts. Referrers' contacts have significantly greater preferences for the referred products than referrers' non-contacts. These results confirm the effectiveness of active (actively screen and select contacts to match products) and passive (homophily-based) matching in online social referrals. We also find that referrers exhibit significantly stronger preferences for the referred products than the recipients. The matching outperforms among narrow-appeal products and more engaged referrers with a smaller local social network. Our findings shed light on the mechanisms and management of social referrals.