Can AI Perform Qualitative Coding Like Human Researchers? Application of GPT-3 in Thematic Analysis of Qualitative Data

• In Event: The AI Dilemma in Qualitative Research

Abstract

This study explored the feasibility and comparability of utilizing Artificial Intelligence (AI) for qualitative data analysis in equity-focused research. To do this, we compared thematic analyses of interview transcripts conducted by human coders and GPT-3. The results suggest that AI can proficiently perform thematic analysis of qualitative interview data when supplied with suitable prompts, demonstrating considerable comparability with human coders. Despite the inherent biases originating from training data, it is possible to mitigate these biases, allowing for the application of AI in equity-focused endeavors. We encourage readers to engage in the evaluation process and form their own opinions. Furthermore, we discuss the potential implications of integrating AI into qualitative research, hoping to stimulate discussion in the field.

Authors

• Yan Jiang, University of California - San Diego

• Lillie Ko-Wong, University of California - San Diego

• Ivan Valdovinos, San Diego Community College District