Use large language models (LLMs) tools for supporting qualitative research: A phenomenological study of Chinese clinical researchers’ experiences and perceptions (Preprint)

Abstract

The integration of large language models (LLMs) into qualitative research has shown promise in enhancing analytical depth and efficiency. However, there remains a substantial gap in empirical evidence regarding how clinical researchers, particularly in China, utilize these technologies in real-world settings.

This study aimed to explore the firsthand experiences, perceptions, and challenges of Chinese clinical researchers using LLMs in qualitative research, identify significant ethical dilemmas and practical issues, and inform future research, policy-making, and educational initiatives.

We conducted a qualitative phenomenological study using semi-structured interviews with researchers from two major tertiary hospitals in Southwest China. Ten Participants were interviewed to solicit their experiences and perceptions regarding using various LLM tools in qualitative research. Colaizzi’s analytic method was employed for thematic analysis of the interview transcripts.

The study identified three key themes: (1) LLMs were extensively used across different stages of research, particularly in manuscript writing, data analysis, and interpretation of results; 2) Researchers experience an emotional transition from initial excitement to a more critical appreciation of the strengths and limitations of LLMs, leading to a more rational understanding of AI’s role and application in research; 3) Ethical concerns and the need for skill development in effectively using AI tools are prominent, calling for specialized training.

LLM tools are pivotal assets in supporting qualitative research among Chinese clinical researchers, offering significant advantages in handling complex data and improving research efficiency. However, the integration of these tools into clinical research practices also raises important ethical questions and necessitates a comprehensive approach to training. Guidelines and educational programs are needed to enhance the ethical and effective use of AI in research to ensure these technologies augment rather than compromise the integrity of scientific research.