ABSTRACT The expanding role of artificial intelligence (AI) in social life compels a foundational reassessment of sociology's concepts, methods, and theoretical commitments. From generative language models to predictive policing, AI systems are no longer mere tools but increasingly agentic, opaque, and normative sociotechnical actors. This article identifies two central tensions that currently define the field: (1) diverse approaches of “sociology on AI” (critical analyses of AI's social impacts) versus “AI for sociology” (instrumental uses of AI in research), and (2) an analytical divide between externalist perspectives focusing on AI's societal consequences and internalist perspectives examining AI's inner workings. Rather than resolving these differences under a single framework, the article argues that recognizing and bridging them can help clarify the field's distinctive contributions. It proposes some potential and integrative agendas that link empirical investigation with critical reflection and connect technical engagement to sociological theory. By mapping its boundaries and identifying its gaps, the sociology of AI can move from fragmentation toward a more comprehensive dialog to shape the social worlds that AI continues to transform.
Canhui Liu (Fri,) studied this question.