Leveraging Rights of Data Subjects for Social Media Analysis: Studying TikTok via Data Donations

Abstract

TikTok is a relatively novel and widely popular media platform. In response to its expanding user base and cultural impact, researchers are turning to study the platform; however, TikTok, like many social media platforms, restricts external access to data. Prior works have acquired data from scraping the platform, user self-reports, and from accounts created by researchers for the study's purpose. Existing techniques, while yielding important insights, contain limitations for gathering large-scale quantitative insights on how real TikTok users behave on the platform. We bridge this research gap by implementing a data donation system to collect TikTok data. Our system leverages users' right to access their data enabled by the EU's GDPR regulation. We recruit 347 TikTok users, ask them to request their data from TikTok, and then use our system to customize, anonymize, and donate their data. We collect 4.9M videos viewed 9.2M times by our participants -and associated engagement metrics -to analyze how people consume content on TikTok, how prevalent liking behavior is on TikTok, and whether there are substantial differences across our participants' demographics. We conclude our work by discussing the lessons learned and future avenues for implementing data donation systems, which we believe offer a promising avenue for collecting user behavioral traces to understand social phenomena through the lens of the Web.

TikTok and the Art of Personalization: Investigating Exploration and Exploitation on Social Media Feeds

“…Particularly, given a set of social media feed attributes that include content, user, and engagement attributes, we design and implement a framework to assess which video recommendations are the result of personalization (i.e., exploit recommendations) and which are not (i.e., explore recommendations). We validate and demonstrate the applicability of our framework on a dataset of real traces from TikTok users collected by [24], as well as other baselines, including traces obtained from automated accounts on TikTok and a randomized baseline. Contributions.…”

“…Dataset from Real Users. Our dataset of traces from real TikTok users, is based on previous work by Zannettou et al [24]. The authors relied on the EU's General Data Protection Regulation (GDPR), particularly the right of access by data subjects.…”

“…Each trace provides a comprehensive view of the user's actions on TikTok, including the user's viewing history, like history, search history, follow history, etc. (see [24] for a comprehensive list of all the fields included in the datasets). Additionally, for each video referenced in the traces from real-world TikTok users, the authors collected additional metadata, as the traces from TikTok include only video identifiers.…”

“…We obtained approval from our institution's Ethical Board Review Committee before conducting any data collection/analysis. We emphasize that the real-world traces dataset, obtained by Zannettou et al [24], was collected after obtaining explicit consent from the participants before data donation. Also, the video metadata collection focuses on publicly accessible videos at the time of data collection (i.e., we do not have any data about deleted videos or videos from uploaders with private accounts).…”

Potholes and Power: A Multimodal Critical Discourse Analysis of “Look At This F*ckin’ Street” on Instagram

“…The size of the total data export is 15.9 GB. The use of donated data in social media research is relatively rare, but in the interest of treating this data ethically and protecting user privacy (Caddle et al, 2021;Zannettou et al, 2023), I have redacted usernames in screenshots and did not review or use any non-public follower data. Detailed information about this dataset and my sample can be found in Table 1.…”