Reproduce “LOGOS: LLM-driven End-to-End Grounded Theory Development and Schema Induction for Qualitative Research” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Xinyu Pi, Qisen Yang, Chuong Nguyen
LOGOS, developed by researchers at UCSD, introduces an LLM-driven, end-to-end framework for automating grounded theory development and schema induction in qualitative research. The system transforms raw text into a structured, hierarchical theory with explicit semantic relationships, achieving up to 88.2% alignment with human-generated schemas and consistently outperforming baseline methods across diverse datasets.
Key insights
- Full automation of the grounded theory workflow, from raw text to a structured, hierarchical theory, is achievable by leveraging LLMs for nuanced interpretation and iterative refinement.
- The explicit construction of a 'relation-aware conceptual graph' over codes, combined with iterative refinement, is crucial for optimizing code abstraction, reusability, and semantic richness, moving beyond simple thematic analysis.
- A principled 5-dimensional statistical metric (reusability, fitness, coverage, parsimony, consistency) with a train-test split protocol provides a standardized and unbiased framework for evaluating automatically generated qualitative codebooks.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.