Yan Zhu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Yan Zhu has a total of 8 co-authors. Here are the top co-authors by number of shared publications:
Jianqing Chen
2 shared publicationsSrinivasan Raghunathan
2 shared publications- Ammu George
National University of Singapore (NUS) - Asia Competitiveness Institute
1 shared publication - Jianqing Chen
Jindal School of Management, The University of Texas at Dallas, Richardson, Texas 75080
1 shared publication - Paul Cheung
Asia Competitiveness Institute, Lee Kuan Yew School of Public Policy, National University of Singapore, Singapore
1 shared publication - Srinivasan Raghunathan
Jindal School of Management, The University of Texas at Dallas, Richardson, Texas 75080
1 shared publication
Latest Publications
- 8 October 2026
- Published by Institute for Operations Research and the Management Sciences (INFORMS) in Information Systems Research
Helping Oneself by Helping a Rival: Social-Media Platforms’ Policies on Enabling Features for Influencers
Social-media platforms and influencers are often seen as collaborators. However, both rely primarily on advertising services, introducing a competitive dimension to their relationship. As the dominant players in this relationship, platforms have used design features, which they control, to manage this dynamic. Some platforms have adopted an antagonistic approach by disabling certain features (e.g., clickable shopping-related hyperlinks) for influencers, while others take a more cooperative stance, allowing influencers to use these features for free or for a fee. We develop a game-theoretic model to explore the competition between the two parties and study the platform’s decision to enable or disable a feature that enhances the effectiveness of an influencer’s service. We show that even when the platform cannot charge the influencers for using the feature, it can still benefit if the advertisers’ valuation of the platform’s service is not significantly lower than that of the influencer’s service. While enabling the feature may result in a loss of market share, the platform can potentially offset this loss by raising its prices. A win–win situation for the platform and the influencer may come at the expense of some advertisers, suggesting the platform’s helping the influencer can lead to a “collusive” outcome, despite the competition for advertisers. If the platform can charge a fee for using the feature, it should always enable the feature at an appropriate fee, as this will benefit the platform. The additional fee revenue may come with no change in the platform’s advertising revenue, with an increase in the advertising revenue, or at the expense of some advertising revenue. The platform strategically uses the fee to influence the influencer’s service offering, which affects the platform’s revenue mix under the new business model.
- 8 October 2026
- Published by Institute for Operations Research and the Management Sciences (INFORMS) in Information Systems Research
Helping Oneself by Helping a Rival: Social-Media Platforms’ Policies on Enabling Features for Influencers
Social-media platforms and influencers are often seen as collaborators. However, both rely primarily on advertising services, introducing a competitive dimension to their relationship. As the dominant players in this relationship, platforms have used design features, which they control, to manage this dynamic. Some platforms have adopted an antagonistic approach by disabling certain features (e.g., clickable shopping-related hyperlinks) for influencers, while others take a more cooperative stance, allowing influencers to use these features for free or for a fee. We develop a game-theoretic model to explore the competition between the two parties and study the platform’s decision to enable or disable a feature that enhances the effectiveness of an influencer’s service. We show that even when the platform cannot charge the influencers for using the feature, it can still benefit if the advertisers’ valuation of the platform’s service is not significantly lower than that of the influencer’s service. While enabling the feature may result in a loss of market share, the platform can potentially offset this loss by raising its prices. A win–win situation for the platform and the influencer may come at the expense of some advertisers, suggesting the platform’s helping the influencer can lead to a “collusive” outcome, despite the competition for advertisers. If the platform can charge a fee for using the feature, it should always enable the feature at an appropriate fee, as this will benefit the platform. The additional fee revenue may come with no change in the platform’s advertising revenue, with an increase in the advertising revenue, or at the expense of some advertising revenue. The platform strategically uses the fee to influence the influencer’s service offering, which affects the platform’s revenue mix under the new business model.
- Open Access
- 2 August 2026
- Published by Elsevier in SSRN Electronic Journal
Economics of AI Labeling on Social Media Platforms
Generative AI has blurred the distinction between human-created and AI-generated content in social media, creating uncertainty for consumers who have heterogeneous preferences for the two content types. While some consumers have a strict preference for the content type with which they engage, others have a soft preference such that the content quality shapes the content type they choose to engage with. Simultaneously, human-created content has become an additional revenue source, besides advertisements, for social media platforms because humancreated content has emerged as key training data for AI models. Some platforms have started to label AI-generated content in response to developments in generative AI, an effort that regulators increasingly mandate or promote, even though the identification of a content's true type is imperfect. We study platforms' incentives to label AI-generated content. We find that while labeling can be profitable, it may actually hurt an advertising-driven platform by diminishing consumer engagement. A platform that monetizes human-created content as well may have even less incentive to adopt labeling. Labeling can also harm content creators, consumers, and overall welfare. The driving force behind these results is creators' strategic response to labeling, which, in turn, depends on how labeling shifts consumers' engagement decisions. While labeling enhances the utility of consumers with strict preferences for content type, incentivizing creators to improve quality, it relaxes competition between creators to attract soft-preference consumers, disincentivizing creators from improving quality. The latter occurs because labeling, by enabling consumers to better distinguish content types, elevates the salience of content type relative to content quality in consumers' engagement decisions. The platform prefers more engagement from human-created content when it monetizes human-created content. However, a high (low) accuracy in detecting AI-generated (human-created) content suppresses the human creator's incentive to improve content quality, diminishing the platform's incentive to label. The results have strong implications for social planners and other stakeholders in social media.
- 2 August 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessEconomics of AI Labeling on Social Media Platforms
Generative AI has blurred the distinction between human-created and AI-generated content in social media, creating uncertainty for consumers who have heterogeneous preferences for the two content types. While some consumers have a strict preference for the content type with which they engage, others have a soft preference such that the content quality shapes the content type they choose to engage with. Simultaneously, human-created content has become an additional revenue source, besides advertisements, for social media platforms because humancreated content has emerged as key training data for AI models. Some platforms have started to label AI-generated content in response to developments in generative AI, an effort that regulators increasingly mandate or promote, even though the identification of a content's true type is imperfect. We study platforms' incentives to label AI-generated content. We find that while labeling can be profitable, it may actually hurt an advertising-driven platform by diminishing consumer engagement. A platform that monetizes human-created content as well may have even less incentive to adopt labeling. Labeling can also harm content creators, consumers, and overall welfare. The driving force behind these results is creators' strategic response to labeling, which, in turn, depends on how labeling shifts consumers' engagement decisions. While labeling enhances the utility of consumers with strict preferences for content type, incentivizing creators to improve quality, it relaxes competition between creators to attract soft-preference consumers, disincentivizing creators from improving quality. The latter occurs because labeling, by enabling consumers to better distinguish content types, elevates the salience of content type relative to content quality in consumers' engagement decisions. The platform prefers more engagement from human-created content when it monetizes human-created content. However, a high (low) accuracy in detecting AI-generated (human-created) content suppresses the human creator's incentive to improve content quality, diminishing the platform's incentive to label. The results have strong implications for social planners and other stakeholders in social media.
- Open Access
- 29 July 2026
- Published by Elsevier in SSRN Electronic Journal
Is an Educated Consumer the Best Customer? The Value of Consumer AI
Consumer AI—AI tools that assist consumers in their purchase decisions—is becoming ubiquitous in online retailing. Unlike traditional product-centric technologies that focus on product-related information, consumer AI centers on consumers by providing expert information about them. In this paper, we develop a game-theoretic model to examine the impact of consumer AI on both manufacturers and consumers. We consider a manufacturer that sells two products to consumers whose utility depends on both the severity of their needs and how well the products satisfy their needs. With AI, consumers are better informed in both dimensions. We find that the value of AI to the manufacturer critically depends on the type of products and the quality of AI. When the products serve a niche market, AI always benefits the manufacturer. In contrast, when the products serve an excessive mass market, AI always hurts the manufacturer. When the products serve a mass market but market penetration is not very high, whether AI hurts the manufacturer depends on AI quality—it hurts the manufacturer if AI's fit precision is low and its valuation precision is moderate. Moreover, although consumer AI is designed to help consumers, we find that it does not always benefit them, as AI adoption leads to higher prices. AI can even hurt both the manufacturer and consumers, resulting in a “lose-lose” outcome. Jointly considering consumer AI and product-centric technologies, we show that AI that improves consumer knowledge about their idiosyncratic needs and product-centric technologies that improve consumer knowledge about product quality can complement each other from the manufacturers' perspective. Therefore, to fully realize the value of consumer AI, manufacturers should improve the quality of product-centric technologies in addition to enhancing AI quality.
- 29 July 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessIs an Educated Consumer the Best Customer? The Value of Consumer AI
Consumer AI—AI tools that assist consumers in their purchase decisions—is becoming ubiquitous in online retailing. Unlike traditional product-centric technologies that focus on product-related information, consumer AI centers on consumers by providing expert information about them. In this paper, we develop a game-theoretic model to examine the impact of consumer AI on both manufacturers and consumers. We consider a manufacturer that sells two products to consumers whose utility depends on both the severity of their needs and how well the products satisfy their needs. With AI, consumers are better informed in both dimensions. We find that the value of AI to the manufacturer critically depends on the type of products and the quality of AI. When the products serve a niche market, AI always benefits the manufacturer. In contrast, when the products serve an excessive mass market, AI always hurts the manufacturer. When the products serve a mass market but market penetration is not very high, whether AI hurts the manufacturer depends on AI quality—it hurts the manufacturer if AI's fit precision is low and its valuation precision is moderate. Moreover, although consumer AI is designed to help consumers, we find that it does not always benefit them, as AI adoption leads to higher prices. AI can even hurt both the manufacturer and consumers, resulting in a “lose-lose” outcome. Jointly considering consumer AI and product-centric technologies, we show that AI that improves consumer knowledge about their idiosyncratic needs and product-centric technologies that improve consumer knowledge about product quality can complement each other from the manufacturers' perspective. Therefore, to fully realize the value of consumer AI, manufacturers should improve the quality of product-centric technologies in addition to enhancing AI quality.
- 13 December 2022
- Published by Taylor & Francis in Applied Economics Letters
- Vol. 31 (10) , 901-906
Pandemic arbitrage in foreign direct investments: A perspective on the modes of entry
Foreign direct investments (FDI) are considered long-term and less sensitive to global shocks as they involve large amounts of capital investment that are costly to reverse. This study examines whether there was a reallocation of FDI flows from destination markets more affected by the pandemic, resulting in a pandemic arbitrage. Using bilateral FDI inflows data from January 2019 to December 2020, we show that FDI flows declined to destination markets that performed worse than source markets in COVID-19 infection rates, with the effect more evident in greenfield FDI. Our results also show that bilateral colonial ties and destination market COVID-19 policy stringency impact the pandemic arbitrage in FDI flows, especially for M&As.
- 13 December 2022
- Published by Taylor & Francis in Applied Economics Letters
- Vol. 31 (10) , 901-906
Pandemic arbitrage in foreign direct investments: A perspective on the modes of entry
Foreign direct investments (FDI) are considered long-term and less sensitive to global shocks as they involve large amounts of capital investment that are costly to reverse. This study examines whether there was a reallocation of FDI flows from destination markets more affected by the pandemic, resulting in a pandemic arbitrage. Using bilateral FDI inflows data from January 2019 to December 2020, we show that FDI flows declined to destination markets that performed worse than source markets in COVID-19 infection rates, with the effect more evident in greenfield FDI. Our results also show that bilateral colonial ties and destination market COVID-19 policy stringency impact the pandemic arbitrage in FDI flows, especially for M&As.