Crowd-based Platforms
Permanent URI for this collection
Browse
Recent Submissions
Item A Blessing or a Curse? The Impact of Platform-initiated Comment Moderation on Subsequent Answer Generation on Social Media Platform(2025-01-07) Zhang, Ran Alan; Ma, Yuhong; Zhang, Luna; Tan, YongSocial media moderation encompasses actions undertaken by platforms to uphold community norms. While existing literature predominantly examines the direct impact of moderation on reducing harmful behavior, limited attention has been given to its spillover effect on other content, particularly on unmoderated content. This study exploits a temporary shutdown of the commenting function on a large Q&A platform to investigate the spillover effect of comment moderation on the subsequent answer generation. Our analysis reveals that comment moderation induce a decrease in the volume of the subsequent answers, with an improved quality during and after the shutdown period. Our mechanism analyses show that, after comment moderation, contributors may become more conservative in providing new answers by writing more similar, specific, and longer content. In addition, comment moderation exerts a negative impact on the volume of the subsequent regular answers but has no significant effect on harmful answers, which provides supporting evidence for the plausibility of a chilling effect.Item Snack and Snap: A Novel Recipe for Yelp Reviews with Explainable AI(2025-01-07) Byarugaba, Yonah; Wang, Bingyang; Garg, Rajiv“Pictures are worth a thousand words," yet most platforms like Yelp, Google Maps, Instagram, Walmart, and Amazon require users to provide text, ratings, and images. Images often capture a user's intent, and the features within the images typically correlate with that intent. In this paper, we extract various features from images (such as edge distribution, color distribution, text within the image, focus, etc.) and compare simple vs. complex models to predict the ratings associated with these images. We find that features such as brightness and contrast significantly explain the rating at image-level, and models such as random forest and logistic regression provide a 0.84 F-1 score when predicting the rating. In the era of generative AI, we anticipate that sharing an image will allow platforms to auto-generate user intent and image ratings, thereby simplifying the dissemination of information.Item Navigating Product Diversification in Live Streaming E-commerce: Evidence from Douyin Platform(2025-01-07) Hu, Jingyun; Gao, Yi; Zhao, KeranThe past decade has witnessed the growing prevalence of live streaming selling (LSS). As streamers serve as a proxy between manufacturers and consumers, product assortment management has been a critical strategic factor for their success. Relying on data collected from Douyin, a prominent Chinese LSS platform, this study aims to investigate the impact of diversification on LSS performance through the lens of adding new categories in live streaming sessions. The results in this study show that product category diversification, in general, has a significant positive effect on sales during live streaming events. However, this effect is alleviated by the semantic similarity between the newly introduced and existing categories. In addition, we find that product diversification has a negative impact on the sales of incumbent categories. This study contributes to the literature on product assortment by providing insights into the role of product diversification in the context of real-time selling. We also discuss the managerial implications for the platform and streamers.Item Does More Informative Job Title Lead to More Successful Hirings? A Randomized Field Experiment on an Online Labor Market Platform(2025-01-07) Cheng, Yihang; Hu, Xiao; Zhu, Chen; Zhu, HengshuOnline labor market platforms have rapidly developed over the past few years. The job title, as a crucial piece of observable information, research about its impact on job-related behaviors is lack. In this study, we selected one of the largest blue-collar online labor market platforms in China, constructing a large-scale field experiment on job titles with different degrees of information disclosure depth and breadth generated by the proposed model based on Large language model (LLM), involving more than 50,000 job listings and more than 800,000 users in the platform, to investigate whether the information disclosed in the job title affects the likelihood of the job being viewed and successfully matched. Findings claim the compared with the central route (processing depth information), peripheral route information processing (processing breadth information) is the dominant method on online labor market platforms. Additionally, sufficient processing of breadth information significantly impacts subsequent behaviors.Item Domestic Policy, Global Shifts: The High-Tech Industry’s Response to Restrictive Immigration Policies(2025-01-07) Li, Kun; Hu, Lin; Gu, BinThe paper examines the unintended consequences on high-tech industry labor market of the 2017 "Buy American Hire American" executive order on the high-tech industry in the United States. It explores how the policy, aimed at promoting American-made products and reducing skilled foreign workers' entry, inadvertently shifted job opportunities to multinational corporations' overseas branches, potentially undermining the goal of increasing domestic employment and harming the sector's innovation capacity. Utilizing job posting data from LinkUp and a difference-in-difference approach, the study provides evidence that high-tech industries saw an increase in foreign job postings and a lack of significant increase in domestic job postings post-policy implementation. This suggests a strategic shift in job creation from the U.S. to international locations, influenced by the executive order's restrictive immigration stance.Item A Dynamic and Multilayered Examination of Comment Networks in a Crowdsourcing Challenge Community(2025-01-07) Li, Yiqi; Mao, Shufan; Lu, SelinaThis study examines how participants of crowdsourcing challenges (ideators) provide comments to one another under the dual community forces of collaboration and competition. Content analysis reveals comment types with various degrees of cooperativeness and self-interestedness. Based on comment-sending patterns, clustering analysis unveils ideators’ different roles in the communities: endorsers, self-promoters, and contributors. Results of longitudinal network analysis on four layers of comment networks present nuanced interaction patterns such as reciprocity, inertia, and homophily. Results suggest that active contribution tends to receive fair returns from the community. Pairs of ideators tend to share reciprocated comments, regardless of the comment types. Therefore, to gain substantial information, ideators should take the initiative and contribute substantially to peer competitors. Moreover, ideators tend to maintain existing habits of comment-giving. Ideators with similar ideas share coopetitive relationships through both cooperative and self-interested comments.Item The Effect of Review Summary: Evidence from a Natural Experiment(2025-01-07) Cheng, Yanan; Gao, Baojun; Zhang, Ran Alan; Li, XitongBy leveraging a unique policy of an online review platform that introduces generative AI reviews summary (GAIRS), this study examines how GAIRS affect the characteristics of user-generated content in the online reviews context. Constructing a unique dataset of online reviews for a matched set of hotels across TripAdvisor and Expedia, we apply a cross-platform difference-in-differences approach to assess the impact of GAIRS. Our findings elucidate the adverse effects of GAIRS on users' subsequent contributions, manifested in diminished quantity and length of subsequent reviews. Nonetheless, GAIRS also correlates with an increase in the average ratings of reviews. We identify the substitution and learning effects as two plausible explanations for these effects. Further analyses reveal that the substitution effect mainly reduces strongly negative reviews, particularly in lower-status hotels, while the learning effect primarily causes inexperienced users to emulate GAIRS. This research contributes to the expanding discourse on how GAI impacts user-generated content.Item Introduction to the Minitrack on Crowd-based Platforms(2025-01-07) Huang, Nina; Gu, Bin; Chen, Pei-Yu