Alex Liu, Min Sun, Lief Esbenshade, Michael Xiao, Victor Tian, Zachary Zhang, Kevin He — arXiv preprint (2026). 📄 Full text (arXiv)
Synthesis
A multi-phase human-LLM collaborative pipeline adapted open, axial, and selective coding to build a hierarchical codebook from 45,000 messages exchanged between K-12 educators and a generative AI platform — an instance of LLMs as analytic assistants at a scale manual coding cannot match.
LLMs generated candidate labels and structured annotations at scale across three phases, while human researchers retained conceptual authority over category definitions, merging decisions, and interpretive frameworks.
The resulting 72-item codebook was validated through systematic human coding of an independent 2,560-message sample, with reliability established via set-valued agreement measures; human coders extended the instrument with five codes the LLM-assisted phases had not surfaced.
The article provides a procedural account (what LLMs can do, in which phases, under what safeguards) directly relevant to AIED research methodology and to conceptualizing how educators actually use generative AI platforms.
Related Pages
Citation
APA: Liu, A., Sun, M., Esbenshade, L., Xiao, M., Tian, V., Zhang, Z., & He, K. (2026). Human-LLM collaborative inductive coding for conceptualizing K-12 educator AI use. arXiv:2607.28889.