The construction of the role of AI in qualitative data analysis in the social sciences

Abstract

Generative AI (AI) is being promoted for and adopted by social science researchers at a rapid pace. This shift has enormous implications for research outcomes. A better understanding of the language that authors are using to construct a version of social science research processes that incorporate AI can inform reporting guidelines and best practices. Qualitative data analysis is a particularly important area of focus as it has historically positioned humans as the research instrument. Human researchers are coming from positions in the world that always impact their understanding of the data. In this paper, we used discourse analysis methods to interrogate how researchers publishing in peer-reviewed social science journals are proposing and/or explaining their use of generative AI to conduct qualitative data analysis. We gathered a corpus of 29 articles through extensive database and direct tables of content searches and documented the journal discipline, paper purpose, and AI platforms. We noted five discursive stances: (1) qualitative data analysis is inherently problematic; (2) qualitative data analysis is easily, and will inevitably be, automated; (3) AI will disrupt methods without (immediately) replacing humans; (4) AI-human hybrid methods are the future; and (5) de-centering ethical concerns and model limitations. We conclude by suggesting how these “AI as qualitative data analyst” discourses will complicate the relationship between qualitative methods and humans as researchers.