User Misconceptions of LLM-Based Conversational Programming Assistants

This repository consolidates the analyzed conversational data for the paper "User Misconceptions of LLM-Based Conversational Programming Assistants", submitted to CHASE 2026. The repository includes a .zip archive containing the datasets and materials used in the analysis, organized as follows: Filename Description conversation_id_map.csv Maps each WildChat conversation hash ID to its corresponding paper misconception ID (e.g., C1–C500). gold_dataset.csv Contains the gold-standard subset (50 conversations), which were double-verified for potential misconceptions and user intentions, including a flag indicating whether they were used for validation. misconception_codebook.csv List and describe the final potential misconception categories used in the coding process. intention_codebook.csv List and describe the final user intention categories. conversation_labels.tsv All individual coder labels cover open coding (A1, A2) and validation (A4, A5, A6). wildchat_snippets.parquet Contains the 500 random sample raw WildChat snippets used for analysis (identical structure to the source dataset). analysis_overview.ipynb Jupyter notebook used to load, describe, and analyze the dataset and labeling outcomes.