Kai Li
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Kai Li has a total of 43 co-authors. Here are the top co-authors by number of shared publications:
- Ching-I Teng
Graduate Institute of Management, Chang Gung University
3 shared publications - Tai Chiu Edwin Cheng
Department of Logistics and Maritime Studies, The Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong SAR, China
2 shared publications - Xiaofei Zhang
Harbin Institute of Technology
2 shared publications - Boxin Zhang
School of Management Science and Engineering, Tianjin University of Finance and Economics, Tianjin, China
1 shared publication - Cheng Zhou
Business and Tourism School, Sichuan Agricultural University
1 shared publication - Dan J. Kim
University of North Texas, Denton, TX, USA
1 shared publication
Latest Publications
- 2 October 2026
- Published by Taylor & Francis in International Journal of Electronic Commerce
- Vol. 30 (4) , 567-598
Mechanical, Thinking, or Feeling? The Differential Effects of AI-Generated Reviews on Consumer Trust
- 2 October 2026
- Published by Taylor & Francis in International Journal of Electronic Commerce
- Vol. 30 (4) , 567-598
Mechanical, Thinking, or Feeling? The Differential Effects of AI-Generated Reviews on Consumer Trust
- Open Access
- 22 March 2026
- Published by MDPI AG in Journal of Theoretical and Applied Electronic Commerce Research
- Vol. 21 (3) , 94
AI Review Bots vs. Humans in Handling Negative Reviews: Who Builds More Trust?
As artificial intelligence (AI) increasingly engages in managing consumer interactions on e-commerce platforms, an important question arises: how do public AI-generated replies to negative reviews compare with human replies in fostering consumer trust? This study investigates consumer trust in reply source (AI review bots vs. human agents) across three reply strategies, namely, default, thinking, and feeling, in a public review context. AI review bots are defined as automated systems that publicly respond to consumer reviews. Using a controlled laboratory experiment, we find that when using the default strategy, human replies elicit greater trust than AI replies, mediated by higher perceived authenticity and persuasiveness. Conversely, when using the thinking strategy, AI replies outperform human replies in building consumer trust, as they are perceived as more authentic and persuasive. In the feeling strategy, there is a convergence between AI and human replies. These findings demonstrate that the effectiveness of AI versus human replies depends on the strategy adopted. Theoretically, this study extends research on human–AI interaction by introducing a three-strategy framework to systematically compare AI and human communicators in public review contexts. Practically, the results guide e-commerce sellers and platforms on when and how to deploy AI review bots to effectively manage consumer trust in response to negative reviews.
- 22 March 2026
- Published by MDPI AG in Journal of Theoretical and Applied Electronic Commerce Research
- Vol. 21 (3) , 94
Open AccessAI Review Bots vs. Humans in Handling Negative Reviews: Who Builds More Trust?
As artificial intelligence (AI) increasingly engages in managing consumer interactions on e-commerce platforms, an important question arises: how do public AI-generated replies to negative reviews compare with human replies in fostering consumer trust? This study investigates consumer trust in reply source (AI review bots vs. human agents) across three reply strategies, namely, default, thinking, and feeling, in a public review context. AI review bots are defined as automated systems that publicly respond to consumer reviews. Using a controlled laboratory experiment, we find that when using the default strategy, human replies elicit greater trust than AI replies, mediated by higher perceived authenticity and persuasiveness. Conversely, when using the thinking strategy, AI replies outperform human replies in building consumer trust, as they are perceived as more authentic and persuasive. In the feeling strategy, there is a convergence between AI and human replies. These findings demonstrate that the effectiveness of AI versus human replies depends on the strategy adopted. Theoretically, this study extends research on human–AI interaction by introducing a three-strategy framework to systematically compare AI and human communicators in public review contexts. Practically, the results guide e-commerce sellers and platforms on when and how to deploy AI review bots to effectively manage consumer trust in response to negative reviews.
- Top Cited PapersOpen Access
- 8 January 2026
- Published by MDPI AG in Journal of Theoretical and Applied Electronic Commerce Research
- Vol. 21 (1) , 29
Transforming E-Commerce with AI: Navigating Innovation, Personalization, and Ethical Challenges
Artificial Intelligence (AI) has become a primary agent of change in the contemporary e-commerce landscape
- 8 January 2026
- Published by MDPI AG in Journal of Theoretical and Applied Electronic Commerce Research
- Vol. 21 (1) , 29
Top Cited PapersOpen AccessTransforming E-Commerce with AI: Navigating Innovation, Personalization, and Ethical Challenges
Artificial Intelligence (AI) has become a primary agent of change in the contemporary e-commerce landscape
- 26 March 2025
- Published by Emerald Publishing in Internet Research
- Vol. 36 (4) , 1329-1357
Unlocking the power of multidimensional ratings: empirical analyses and designs from mediation models
Purpose: Prior papers have not fully examined the relationship between multidimensional (MD) ratings and review helpfulness, which may limit the potential of MD ratings in mitigating information asymmetry. Thus, leveraging signaling theory and cued recall, we analyzed the direct and indirect impacts of MD ratings on review helpfulness and subsequently derived corresponding design guidelines. Design/methodology/approach: We adopted a multiple mediation model to examine the direct and indirect effects of MD ratings on review helpfulness. We conducted a series of econometric analyses to ensure the robustness of our conclusions. Additionally, Fisher’s permutation test was adopted to assess the potential heterogeneity in these impacts across various hotel types. Findings: Leveraging the review data from TripAdvisor.com, we found that, beyond the positive direct impacts of MD ratings, there are significant indirect impacts positively mediated by review writing effort but negatively mediated by content polarity, accounting for 55% of the total impact. Moreover, we revealed that these impacts exhibit significant heterogeneity across different hotel types, online rank and rating valence. Finally, we incorporated both these identified impacts and their heterogeneity into the design of an MD ratings system, yielding guidelines regarding the quantity and category of dimensional ratings. Originality/value: We extended the determinants of review helpfulness to the areas of MD ratings and highlighted the impact of the interaction of review functions on review helpfulness. Besides, we emphasized the importance of the signaling environment and illustrated how the effects of signals were shaped by cognitive processes, i.e. cued recall. Finally, we offered practical design rules regarding the quantity and category of dimensional ratings, aiming to maximize the power of MD ratings.
- 26 March 2025
- Published by Emerald Publishing in Internet Research
- Vol. 36 (4) , 1329-1357
Unlocking the power of multidimensional ratings: empirical analyses and designs from mediation models
Purpose: Prior papers have not fully examined the relationship between multidimensional (MD) ratings and review helpfulness, which may limit the potential of MD ratings in mitigating information asymmetry. Thus, leveraging signaling theory and cued recall, we analyzed the direct and indirect impacts of MD ratings on review helpfulness and subsequently derived corresponding design guidelines. Design/methodology/approach: We adopted a multiple mediation model to examine the direct and indirect effects of MD ratings on review helpfulness. We conducted a series of econometric analyses to ensure the robustness of our conclusions. Additionally, Fisher’s permutation test was adopted to assess the potential heterogeneity in these impacts across various hotel types. Findings: Leveraging the review data from TripAdvisor.com, we found that, beyond the positive direct impacts of MD ratings, there are significant indirect impacts positively mediated by review writing effort but negatively mediated by content polarity, accounting for 55% of the total impact. Moreover, we revealed that these impacts exhibit significant heterogeneity across different hotel types, online rank and rating valence. Finally, we incorporated both these identified impacts and their heterogeneity into the design of an MD ratings system, yielding guidelines regarding the quantity and category of dimensional ratings. Originality/value: We extended the determinants of review helpfulness to the areas of MD ratings and highlighted the impact of the interaction of review functions on review helpfulness. Besides, we emphasized the importance of the signaling environment and illustrated how the effects of signals were shaped by cognitive processes, i.e. cued recall. Finally, we offered practical design rules regarding the quantity and category of dimensional ratings, aiming to maximize the power of MD ratings.
- Open Access
- 1 January 2025
- Published by HICSS Conference Office
The Impact of AI-Generated Product Summaries on the Speed and Efficiency of E-commerce Consumer Decision-Making
This research investigates how Artificial Intelligence (AI)-generated summaries influence decision-making among consumers on e-commerce platforms. By analyzing data from two Chinese digital deal platforms, Zdm and Sqkb, the study applies a Difference-in-Difference-in-Differences (DDD) methodology to evaluate how these summaries affect consumer behavior. The results demonstrate that AI-generated summaries significantly speed up consumer decision-making and reduce hesitation, with effects varying by product type and price. This research underscores the potential role of AI in enhancing consumer experiences and decision-making processes on online digital deal platforms.
- 1 January 2025
- Published by HICSS Conference Office
Open AccessThe Impact of AI-Generated Product Summaries on the Speed and Efficiency of E-commerce Consumer Decision-Making
This research investigates how Artificial Intelligence (AI)-generated summaries influence decision-making among consumers on e-commerce platforms. By analyzing data from two Chinese digital deal platforms, Zdm and Sqkb, the study applies a Difference-in-Difference-in-Differences (DDD) methodology to evaluate how these summaries affect consumer behavior. The results demonstrate that AI-generated summaries significantly speed up consumer decision-making and reduce hesitation, with effects varying by product type and price. This research underscores the potential role of AI in enhancing consumer experiences and decision-making processes on online digital deal platforms.
- 18 February 2022
- Published by Elsevier in Decision Support Systems
- Vol. 157, 113755
Exploring the spoiler effect in the digital age: Evidence from the movie industry
- 18 February 2022
- Published by Elsevier in Decision Support Systems
- Vol. 157, 113755
Exploring the spoiler effect in the digital age: Evidence from the movie industry
- 2 January 2022
- Published by Taylor & Francis in Journal of Global Information Technology Management
- Vol. 25 (1) , 54-82
Supply–Demand Docking Joint Decision of Cross-border E-commerce under the Influence of Internet Celebrity
While cross-border e-commerce (CBEC) is booming in the world today, there are many obstacles hindering the development of the market, such as fake commodities and inferior goods, unreliable and lengthy transit times, and so on. The emergence of Internet celebrity brings new vitality to CBEC, as they build a bridge of trust between brand merchants and consumers. At the same time, the more complex CBEC supply chain with Internet celebrity involved needs to be integrated. This study propose a business-to-business-to-consumer (B2B2C) supply–demand docking business model and accordingly four incentive mechanisms (namely high-quality production incentive, Internet celebrity marketing incentive, equity sharing incentive and cooperative voting mechanism) to achieve the CBEC supply chain vertical integration. Our analysis shows that a higher commission proportion of CBEC platform results in lower quality improvement level and demand, and higher sale price and consumer experience level. We also find that joint decision-making actually realizes Pareto improvement of decentralized decision-making. Moreover, the introduction of cooperative voting mechanism ensures the fair expression of the interest demands of participants in the joint decision, providing a win-win situation for all parties. This study provides a sufficient basis for the vital role of Internet celebrity in CBEC.
- 2 January 2022
- Published by Taylor & Francis in Journal of Global Information Technology Management
- Vol. 25 (1) , 54-82
Supply–Demand Docking Joint Decision of Cross-border E-commerce under the Influence of Internet Celebrity
While cross-border e-commerce (CBEC) is booming in the world today, there are many obstacles hindering the development of the market, such as fake commodities and inferior goods, unreliable and lengthy transit times, and so on. The emergence of Internet celebrity brings new vitality to CBEC, as they build a bridge of trust between brand merchants and consumers. At the same time, the more complex CBEC supply chain with Internet celebrity involved needs to be integrated. This study propose a business-to-business-to-consumer (B2B2C) supply–demand docking business model and accordingly four incentive mechanisms (namely high-quality production incentive, Internet celebrity marketing incentive, equity sharing incentive and cooperative voting mechanism) to achieve the CBEC supply chain vertical integration. Our analysis shows that a higher commission proportion of CBEC platform results in lower quality improvement level and demand, and higher sale price and consumer experience level. We also find that joint decision-making actually realizes Pareto improvement of decentralized decision-making. Moreover, the introduction of cooperative voting mechanism ensures the fair expression of the interest demands of participants in the joint decision, providing a win-win situation for all parties. This study provides a sufficient basis for the vital role of Internet celebrity in CBEC.
- 1 January 2022
- Published by Inderscience Publishers in European J. of Industrial Engineering
- Vol. 16 (5) , 556
Remanufacturing with material restrictions in monopoly and duopoly
- 1 January 2022
- Published by Inderscience Publishers in European J. of Industrial Engineering
- Vol. 16 (5) , 556
Remanufacturing with material restrictions in monopoly and duopoly
- 16 November 2021
- Published by Emerald Publishing in Aslib Journal of Information Management
- Vol. 74 (2) , 202-224
Understanding the influence of privacy protection functions on continuance usage of push notification service
Push notification service (PNS) is an important approach to distribute personalized information to users timely and is getting more and more popular. However, users' privacy concerns are a major inhibiting factor in their continuance usage of PNS. This study investigates the effect of privacy protection functions provided by PNS sites in enhancing users' perceived fairness on the basis of justice theory to mitigate users' concerns of information privacy. The mechanism underlying such influence on users' continuance usage of PNS comprises privacy concern and privacy-control self-efficacy. Four scenario-based surveys are conducted to test the proposed hypotheses. The authors test the research model with a sample of 360 participants by ANOVA and PLS. Results show that the proposed privacy protection functions have direct positive effects on users' privacy-control self-efficacy, negative effects on privacy concern and indirectly affect their continuance usage of PNS. Furthermore, the interaction effects between two approaches have different impacts on users' privacy concern and privacy-control self-efficacy. This study provides some suggestions and guidance for PNS providers to design effective privacy protection technologies.
- 16 November 2021
- Published by Emerald Publishing in Aslib Journal of Information Management
- Vol. 74 (2) , 202-224
Understanding the influence of privacy protection functions on continuance usage of push notification service
Push notification service (PNS) is an important approach to distribute personalized information to users timely and is getting more and more popular. However, users' privacy concerns are a major inhibiting factor in their continuance usage of PNS. This study investigates the effect of privacy protection functions provided by PNS sites in enhancing users' perceived fairness on the basis of justice theory to mitigate users' concerns of information privacy. The mechanism underlying such influence on users' continuance usage of PNS comprises privacy concern and privacy-control self-efficacy. Four scenario-based surveys are conducted to test the proposed hypotheses. The authors test the research model with a sample of 360 participants by ANOVA and PLS. Results show that the proposed privacy protection functions have direct positive effects on users' privacy-control self-efficacy, negative effects on privacy concern and indirectly affect their continuance usage of PNS. Furthermore, the interaction effects between two approaches have different impacts on users' privacy concern and privacy-control self-efficacy. This study provides some suggestions and guidance for PNS providers to design effective privacy protection technologies.