2026/02/08 by Matt Artz
Neuroscience · Social Sciences · #Anthropology: Ethics, History, Culture #Embodied and Extended Cognition #Language and cultural evolution
paper · doi:10.1080/00664677.2026.2614501
crossref issued 2026/02/08 · crossref published 2026/02/08 · crossref published-online 2026/02/08 · openalex publication_date 2026/02/08 · crossref created 2026/02/08 · crossref deposited 2026/02/08 · openalex created_date 2026/02/10 · crossref indexed 2026/07/31 · openalex updated_date 2026/08/01
This paper introduces multi-agent ethnography (MAE), an approach that positions LLM-based AI agents as configurable collaborators within distributed human−AI research networks. MAE extends anthropology's tradition of methodological innovation—from multi-sited to multi-species ethnography—by incorporating AI agents as research partners across the entire research lifecycle. Drawing on empirical evidence demonstrating AI's capacity to function as a ‘cybernetic teammate’ (Dell'Acqua et al. 2025), I argue that purpose-built agents designed with anthropological considerations can extend research capabilities beyond what either humans or AI achieve independently. To demonstrate this approach, I present the AI Anthropology Toolkit, an MCP server implementation that coordinates specialised agents through conversational interaction, enabling researchers to direct complex analytical workflows using natural language. The current implementation comprises three agents for codebook generation, transcript segmentation, and coding with thematic analysis. This conversational approach enables systematic comparative analysis across multiple epistemological perspectives simultaneously, allowing individual researchers to access the analytical breadth typically achieved through collaborative team configurations. MAE's architecture supports agent coordination across research design, fieldwork, analysis, and dissemination. While data privacy, environmental costs, and algorithmic bias remain important considerations, the Toolkit demonstrates that anthropologically informed AI tools are feasible to build, accessible to use, and capable of augmenting ethnographic practice across research phases, expanding what individual researchers can achieve.