The method of synthetic focus groups in the context of digital transformation of sociological research

The subject of the research is the use of synthetic focus groups (SFG), created using large language models (LLM), within the framework of the digital transformation of sociological analysis. The object of the research is innovative methods for collecting and interpreting qualitative data, aimed at simulating group discussions without the participation of real respondents. The article discusses the heuristic potential of SFG, the possibilities of applying the method in conditions of limited access to respondents, and its relevance for studying attitudes toward artificial intelligence in the field of higher education. Special attention is paid to the ways of organizing interaction with language models through a system of prompts, the formation of participant roles, and the analysis of the obtained discursive positions. The article also addresses the ethical and methodological challenges arising from the use of synthetic participants for research purposes. The methodology combines theoretical analysis of scientific publications with practical modeling of SFG, implemented through sequential prompts to language models and subsequent interpretation of the generated data. The scientific novelty of the work lies in the testing of the synthetic focus group (SFG) method as an innovative tool for collecting qualitative information using generative language models (LLM) such as Gemini, Qwen, Llama, Deepseek, and Mistral. The study pays particular attention to modeling SFG based on user prompts aimed at exploring the attitudes of students and faculty toward artificial intelligence in the higher education system. All models used emphasized the methodological limitations of SFG: the possibility of data distortion, the necessity of clearly indicating the synthetic nature of the participants, and the need to complement such studies with traditional methods. Such methods can be useful in situations requiring rapid hypothesis generation, preliminary testing of research scenarios, as well as in educational and expert practice. At the same time, the necessity of a critical approach to the use of SFG is emphasized, especially in the context of the validity and representativeness of the obtained data.

artificial intelligence, synthetic focus groups, large language models, digital sociology, user personas, bias, qualitative methods, sociological research, Virtual Focus Group, Multi Agent Simulation