Data and analysis code for "Beyond Empathy: Multidimensional and Context-Dependent Evaluation of AI Counseling Through LLM Simulation"

This repository contains the dataset and analysis code associated with the study “Beyond Empathy: Multidimensional and Context-Dependent Evaluation of AI Counseling Through LLM Simulation.” The dataset comprises 75 simulated counseling conversations generated across 25 combinations of five patient profiles and five counseling strategies, with three independently generated conversations per condition. Each conversation consists of five counselor turns, yielding 375 evaluated counselor responses. The repository includes turn-level dialogue data and evaluation scores for Emotional Support, Exploratory Depth, Usefulness, Autonomy Support, and Maladaptive Risk, together with analysis code used to reproduce the statistical analyses reported in the study.