Computational Methods

Computational methods are research techniques that use algorithms, large-scale data processing, and machine learning to analyze cultural and social phenomena at scales that would be impossible through close reading or individual observation alone. They include network analysis, natural language processing, topic modeling, agent-based simulation, and increasingly, the use of AI systems to support interpretation. For anthropologists and adjacent researchers, computational methods have become a practical extension of the qualitative toolkit rather than a replacement for it.

Drawing on a long tradition of mixed methods research, computational work in anthropology resists the common framing that pits numbers against narratives. The most productive projects treat computational and interpretive methods as complementary, each answering questions the other cannot. Algorithms can surface patterns across millions of posts, transcripts, or network connections that no single researcher could read in a lifetime. Ethnographic interpretation gives those patterns meaning by grounding them in the lived contexts from which they emerged. Neither approach is complete on its own.

As machine learning has matured, the range of what counts as a computational method has expanded considerably. Older techniques like content analysis and social network analysis now coexist with newer approaches built on large language models, vector embeddings, and retrieval-augmented generation. Each new capability creates its own interpretive challenges. Bias in training data, the opacity of model decisions, and the drift in what systems can do from one release to the next all complicate the promise that computational tools will make research faster or more objective. Yet these same systems, used with care, open research questions that were previously impractical to pursue.

Within my own practice, computational methods sit alongside ethnography as equal partners in the research process. A project might begin with large-scale analysis of a corpus—interview transcripts, platform data, or archival material—and then narrow into close ethnographic work with the people whose practices produced that data. In other projects, the sequence runs the other way: ethnographic fieldwork generates questions that computational analysis helps answer at scale. The choice is practical rather than principled, and it depends on what the research actually needs.

For anyone interested in how computational and ethnographic methods come together, the edited volume Anthropology and AI offers several chapters on the topic, and the Anthropology in Business podcast features conversations with researchers working at this intersection. My consulting practice, Azimuth Labs, uses computational methods alongside ethnography and qualitative research to produce insights that neither approach could generate alone.