Large Language Model for Qualitative Research: A Systematic Mapping Study

What is it about?

This paper gives an overview of how Large Language Models (LLMs) – such as ChatGPT – are being used to support researchers in analyzing text. Traditional qualitative analysis of interviews, comments, or survey answers usually takes a lot of time and manual work. LLMs can speed up steps like finding themes, grouping answers, and spotting patterns. The authors carried out a systematic review of the literature. They searched scientific databases, kept only studies that clearly described how LLMs were applied, and analyzed eight papers published mainly in 2023 and 2024. These studies explored the use of LLMs in healthcare, education, cultural studies, and technology, but not yet in Software Engineering. Overall, LLMs often achieved results similar to traditional methods and saved a great deal of time: tasks that once took weeks could be done in hours. Still, there are challenges: models can sometimes “hallucinate” (produce unfounded answers), they depend on well-designed prompts, and they may miss emotional or contextual subtleties in the text. The article concludes that LLMs have strong potential to transform qualitative research, especially when combined with researchers’ critical judgment. It also highlights future opportunities, such as improving prompt design, reducing bias and hallucinations, and developing easy-to-use tools for people without AI expertise.

Featured Image

Why is it important?

This study is important because it provides a clear and organized view of how Large Language Models (LLMs) are being applied in qualitative research, a field where tasks are usually time-consuming and heavily dependent on human judgment. Unlike other works, it not only gathers recent studies but also systematically examines how the models were configured, tested, and compared with traditional methods, highlighting metrics, limitations, and opportunities for improvement. By mapping gaps—such as the lack of applications in Software Engineering—and suggesting ways to make LLMs more reliable and accessible, the paper serves as a practical guide for researchers who want to use these tools critically and in a well-grounded way.