Reproduce “LLM-Assisted Content Analysis: Using Large Language Models to Support Deductive Coding” 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
Robert Chew, John Bollenbacher, Michael Wenger, Jessica Speer, Annice Kim
RTI International developed the LACA framework, which integrates large language models like GPT-3.5 into deductive content analysis workflows, demonstrating that automated coding can achieve inter-rater reliability comparable to human coders while significantly reducing the time and effort required for large datasets.
Key insights
- GPT-3.5 achieved inter-rater reliability scores comparable to human coders across diverse datasets, particularly for categories with clear, objective criteria.
- The LLM's ability to provide reasoning for its coding decisions enhances transparency, allowing researchers to assess validity and refine codebooks.
- Disagreements between human and LLM coding often occurred in subjective or ambiguous categories, or those requiring cultural/contextual knowledge.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.