Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

The AI Co-Ethnographer: How Far Can Automation Take Qualitative Research?

Fabian Retkowski, Andreas Sudmann, Alexander Waibel

TL;DR

The AI Co-Ethnographer (AICoE) presents an end-to-end, LLM-powered pipeline that expands qualitative data analysis from automated code mapping to open coding, consolidation, application, and pattern discovery, enabling inductive code development while preserving ethnographic depth. Through evaluations on CVDQuoding and HiAICS datasets, the approach demonstrates robust codebook development and meaningful pattern findings, though interpretive depth remains partly contingent on human oversight and data quality. The framework treats AI as an epistemic medium that augments human researchers in a collaborative, iterative process, rather than replacing ethnography, and highlights the need for multimodal extensions to handle diverse data forms. Overall, AICoE offers a scalable, transparent, and flexible blueprint for AI-assisted qualitative research with explicit human-in-the-loop safeguards and methodological rigor.

Abstract

Qualitative research often involves labor-intensive processes that are difficult to scale while preserving analytical depth. This paper introduces The AI Co-Ethnographer (AICoE), a novel end-to-end pipeline developed for qualitative research and designed to move beyond the limitations of simply automating code assignments, offering a more integrated approach. AICoE organizes the entire process, encompassing open coding, code consolidation, code application, and even pattern discovery, leading to a comprehensive analysis of qualitative data.

The AI Co-Ethnographer: How Far Can Automation Take Qualitative Research?

TL;DR

The AI Co-Ethnographer (AICoE) presents an end-to-end, LLM-powered pipeline that expands qualitative data analysis from automated code mapping to open coding, consolidation, application, and pattern discovery, enabling inductive code development while preserving ethnographic depth. Through evaluations on CVDQuoding and HiAICS datasets, the approach demonstrates robust codebook development and meaningful pattern findings, though interpretive depth remains partly contingent on human oversight and data quality. The framework treats AI as an epistemic medium that augments human researchers in a collaborative, iterative process, rather than replacing ethnography, and highlights the need for multimodal extensions to handle diverse data forms. Overall, AICoE offers a scalable, transparent, and flexible blueprint for AI-assisted qualitative research with explicit human-in-the-loop safeguards and methodological rigor.

Abstract

Qualitative research often involves labor-intensive processes that are difficult to scale while preserving analytical depth. This paper introduces The AI Co-Ethnographer (AICoE), a novel end-to-end pipeline developed for qualitative research and designed to move beyond the limitations of simply automating code assignments, offering a more integrated approach. AICoE organizes the entire process, encompassing open coding, code consolidation, code application, and even pattern discovery, leading to a comprehensive analysis of qualitative data.
Paper Structure (32 sections, 2 equations, 7 figures, 7 tables)

This paper contains 32 sections, 2 equations, 7 figures, 7 tables.

Table of Contents

  1. Introduction
  2. Related Research
  3. Methodology
  4. Open Coding
  5. Code Consolidation
  6. Code Application
  7. Pattern Finding
  8. Prompt Engineering
  9. Experiments and Results
  10. Semantic Relatedness of Codebooks
  11. Relevance of Code Assignments
  12. Quality of Theoretical Findings
  13. Discussion
  14. Alignments, Gaps, and New Perspectives in Codebooks.
  15. Coding Performance Disparities.
  16. ...and 17 more sections

Figures (7)

  • Figure 1: Conceptual Illustration of the AI Co-Ethnographer Pipeline
  • Figure 2: Side-by-side comparison of the codebooks developed by two human coders and the AICoE system for analyzing the HiAICS data. The comparison highlights overlapping themes, distinct coding approaches, and varying emphases in categories such as technical concepts, historical perspectives, ethical considerations, and individual interviewee experiences.
  • Figure 3: Examplary visualization of select relationships between codes between a human-developed codebook and the codebook of AICoE, as annotated by one of our expert annotators
  • Figure 4: High-quality findings generated by AI Co-Ethnographer from the HiAICS dataset, as rated by three evaluators. The Quality Score (1.00–5.00) represents the average across all evaluators and criteria.
  • Figure 5: Evaluation interface that allows human annotators to specify the relationships between different codebooks
  • ...and 2 more figures