We introduce CollabCoder, a system leveraging Large Language Models (LLMs) to support three CQA stages: independent open coding, iterative discussions, and the development of a final codebook.
-
In the independent open coding phase, CollabCoder provides AI-generated code suggestions on demand and allows users to record coding decision-making information (e.g. keywords and certainty) as support for the process.
-
During the discussion phase, CollabCoder helps to build mutual understanding and productive discussion by sharing coding decision-making information within the team. It also helps to quickly identify agreements and disagreements through quantitative metrics, in order to build a final consensus.
-
During the code grouping phase, CollabCoder employs a top-down approach for primary code group recommendations, reducing the cognitive burden of generating the final codebook.
bash scripts/install.sh
Modify the .env according to server settings.
- run the backend Express server and frontend react server
bash scripts/start.sh
CollabCoder: A GPT-Powered Workflow for Collaborative Qualitative Analysis
@misc{gao2023collabcoder, title={CollabCoder: A GPT-Powered Workflow for Collaborative Qualitative Analysis}, author={Jie Gao and Yuchen Guo and Gionnieve Lim and Tianqin Zhang and Zheng Zhang and Toby Jia-Jun Li and Simon Tangi Perrault}, year={2023}, eprint={2304.07366}, archivePrefix={arXiv}, primaryClass={cs.HC} }
Please contact the author (https://gaojie058.github.io/) for any questions: gaojie056@gmail.com