An Artificial Intelligence Agent as a Research Co-Author: An Agentic Approach to Qualitative Sociology

Abstract

This article examines the transformation of research in the social sciences under the influence of agentic systems based on large language models. Most researchers use generative artificial intelligence (AI) in the mode of one-off interaction, treating it as an “intelligent search engine” without memory of context, access to tools, or the ability to perform multi-step tasks. However, the architecture of an AI agent—a system that combines a language model with tools such as web search, code execution, file reading and writing, long-term memory, and an orchestrator—turns a passive response generator into an active co-author. The article proposes a conceptual framework for an agentic approach and describes a six-stage research cycle: from hypothesis formulation and mapping of the research field through work with sources, automatic interview transcription, and initial coding to quantitative analytics and interpretation. Each stage is examined through specific practices and tools: Whisper and GigaAM for transcription; coding agents such as OpenCode and Claude Code for analytical programming; Obsidian as a knowledge management system with agentic access; and Semantic Scholar and Elicit for systematic literature search. Particular attention is paid to the Russian context: the legal requirements for the localization of personal data, the need for local deployment of models when working with sensitive materials, and the development of domestic language models. The article discusses the ethical issues of co-authorship, including the identification of hallucinations, methodological transparency, the problem of authorship, the risks of bias in Anglocentric models, and reproducibility. Three key competencies of the researcher of the future are formulated: systems thinking, critical verification, and ethical awareness. The article concludes that research is shifting from the paradigm of AI-assisted research to that of AI-augmented research, in which the researcher acts as an architect of intelligent systems while retaining full responsibility for the results.