2026.Feb.10

REPORTS

Report on the 4th BAIRAL Research Meeting for Fiscal Year 2025
“Working with AI or AI-assisted? An experience of Frame analysis”

The 4th BAIRAL research meeting for 2025 was held on December 4th, 2025. This time, we invited Dr. Hui-Wen Liu, Professor in the Department of Communication at National Chengchi University (NCCU) in Taiwan, to speak on the theme of “Working with AI or AI-assisted? An experience of Frame analysis.”

The session focused on how generative AI can (and cannot) function within social science research workflows, especially for framing analysis of social media discourse in Taiwan. Prof. Liu traced a trajectory from earlier big-data and “ground truth” efforts, where human coders painstakingly labeled large datasets for machine learning, to recent attempts to “teach” AI established theories and methods so that it can assist in classifying complex, emotionally charged online content. This led to broader questions about what happens when news-derived analytical frameworks are imposed on social media texts, how categories like “emotion” emerge, and where AI currently sits on the spectrum from tool to assistant to potential research collaborator.

Across the talk, Prof. Liu emphasized methodological and epistemological issues, including the importance of clarifying units of analysis on social media, the difficulty of interpreting context-dependent expressions (emojis, “+1,” brief replies), and the challenge of evaluating AI’s “understanding” of a theory or codebook. AI’s behavior in the project suggested both its usefulness, for scaling up coding and surfacing mismatches between frameworks and data, and its limitations, as it failed to perform key reflexive moves that human researchers would normally make, prompting a critical reconsideration of what role AI should play in research design and interpretation.

During the discussion, audience members engaged actively with both conceptual and practical dimensions of the project, asking how the team evaluated AI’s comprehension of theoretical texts and what this implies for reliability and reproducibility. Participants also raised questions about appropriate tools, the balance between quantitative and qualitative uses of AI, and the kinds of skills future researchers will need, turning the Q&A into a broader exchange about academic labor, responsibility, and collaboration in an AI-mediated research environment.