AI-augmented social scientist: Labeling Methods for TikTok News content
The project critically investigated the application potential of generative AI as a method for the social sciences.
Background
TikTok is developing into a key platform for news, advertising, politics, online shopping, and entertainment in Germany with over 20 million monthly users in Germany. Especially among young people, TikTok plays an increasing role in their information environment. Large-scale content classification is one key necessity in empowering social science research to analyze the mechanisms behind information dissemination on TikTok. Due to the multi-modal (video, audio, text) nature of TikTok content, the classification of content remains an immense challenge. In this project, we addressed this challenge, exploring novel methodological possibilities to facilitate such classification at scale.
Objective
First of all, we will provided an expert-coded dataset of over 8,000 TikTok-News videos from 2023. The coding included descriptive variables on the formal character of the videos and theory-derived concepts from the communications sciences. Based on this dataset, we explored the ability of generative AI solutions to label multi-modal social media content.
Publications
-
Dataset Paper: News on TikTok: An Annotated Dataset of TikTok Videos from German-Speaking News Outlets in 2023. AAAI ICWSM Dataset Paper.
-
Dataset: News on TikTok: An Annotated Dataset of TikTok Videos from German-Speaking News Outlets in 2023 [Datensatz], GESIS datorium.
-
Blog Post: Tutorial: When and how to use the official TikTok API. Weizenbaum Methods Blog.
