AI Agents As Research Tools In Sociology: Automating Data Collection, Analysis and Interpretation In Social Research

This regular session examines the methodological, epistemological and practical implications of using AI agents as research tools in sociology. It invites 4–5 papers addressing how large language models, Retrieval-Augmented Generation (RAG) systems and agent-based architectures are reshaping data collection, corpus building, qualitative coding, text classification, summarization, interpretation and validation in social research. The session is organized around three axes. First, it considers LLMs and RAG as research infrastructure, focusing on tokenization, embeddings, retrieval quality, evidence-anchored generation and the risks of hallucination. Second, it examines AI agents as systems that combine language models with tools, memory and semi-autonomous action, enabling new forms of distributed research agency. This includes applications in large-scale studies of disinformation ecosystems, as well as the emerging use of AI agents as simulated focus groups capable of generating deliberative discourse, disagreement and perspective-taking. Third, the session addresses critical limits: where human judgment remains non-delegable, how tool design shapes methodological outcomes, and how human-in-the-loop, human-on-the-loop and human-out-of-the-loop models affect validity, accountability and interpretation. The format will be a regular RC33 session with 4–5 paper presentations, followed by collective discussion. The session welcomes empirical applications, methodological frameworks and critical reflections, especially contributions from early-career researchers and scholars from the Global South.