Minimal Dataset and Code for: Social friction vs. Cognitive efficiency: A comparative analysis of help-seeking behaviors in human communities and generative AI

This repository contains the minimal dataset and author-generated Python code required to replicate the findings of the study: "Social friction vs. Cognitive efficiency: A comparative analysis of help-seeking behaviors in human communities and generative AI," submitted to PLOS ONE.The replication package includes:matched_corpus_N30000.csv: The complete, propensity-score-matched dataset consisting of 30,000 dialogue turns from LMSYS-Chat-1M (Human-AI) and Stack Exchange (Human-Human).analysis_and_figures_code.py: A comprehensive Python script used for text preprocessing, linguistic feature extraction (Hedges, Politeness markers, Vulnerability metrics, etc.), statistical analysis, and high-resolution figure generation.All data processing follows standard computational sociolinguistic protocols to ensure full transparency and reproducibility of the research.本仓库包含复现该研究成果所需的极简数据集与作者原创Python代码，该研究论文为"Social friction vs. Cognitive efficiency: A comparative analysis of help-seeking behaviors in human communities and generative AI"，已提交至PLOS ONE。本复现包包含以下内容： 1. `matched_corpus_N30000.csv`：完整的倾向得分匹配（propensity-score-matched）数据集，涵盖来自LMSYS-Chat-1M（人机对话）与Stack Exchange（人人对话）的30000轮对话回合。 2. `analysis_and_figures_code.py`：一款功能完备的Python脚本，可用于文本预处理、语言特征提取（含模糊限制语（Hedges）、礼貌标记（Politeness markers）、脆弱性指标（Vulnerability metrics）等）、统计分析以及高分辨率图表生成。 所有数据处理流程均遵循标准计算社会语言学（computational sociolinguistic）规范，以保障研究的完全透明性与可复现性。