Critical Subjectivity in Qualitative Social Work Research in the Age of Artificial Intelligence

As artificial intelligence (AI), particularly large language models (LLMs), becomes increasingly integrated into qualitative research, critical questions arise regarding interpretive authority and the status of subjectivity in knowledge production. In social work, qualitative inquiry has long enacted critical subjectivity, an epistemological stance that understands knowledge as situated, relational, reflexive, and ethically accountable. We argue that incorporating LLMs into qualitative data analysis represents not simply methodological innovation but a redistribution of interpretive authority. To clarify this shift, we introduce the concept of “patterned subjectivity” to describe the probabilistic form of interpretation generated through LLM output. Drawing on a cross-account experiment using synthetic interview data, we demonstrate that LLM analyses reproduce broad thematic patterns that vary in organization and emphasis. This patterned variability differs from the reflexive, embodied divergence characteristic of human critical subjectivity. Although reliability metrics may signal consistency, they do not resolve the epistemological question of who, or what, holds interpretive authority. LLMs may function as analytic tools or pattern generators, but they cannot substitute for the reflexive, relational, and ethically accountable labor central to qualitative social work research. The task, therefore, is not whether to adopt or reject AI but how to engage it without ceding epistemological authority.