Data and Reproducibility Materials for "AI-Assisted Arts-Based Practice: Student Self-Reports and Artwork Interpretation from an Affective Science Perspective"

This repository contains the de-identified data and reproducibility materials supporting the manuscript “AI-Assisted Arts-Based Practice: Student Self-Reports and Artwork Interpretation from an Affective Science Perspective” The package includes: (1) aggregate classroom completion counts and summary statistics for fluid-painting, mandala, and collage classes; (2) de-identified call-level valence, arousal, and dominance scores from repeated multimodal AI analyses of 30 archived artworks; (3) technical repeatability, within-artwork variation, and processing-completeness tables; (4) the exact Chinese-language prompts supplied to the model, together with author-prepared English translations and model-configuration records; (5) Python scripts for reproducing and verifying the released results. Student-level responses, item-level questionnaire data, identifiable student-artwork links, artwork images and embedded text, raw narrative AI outputs, internal system identifiers, image hashes, and exact timestamps are not included to protect participant privacy. AI-generated scores represent model-based annotations of perceived expression in artworks. They should not be interpreted as measures of students’ psychological states, diagnoses, clinical outcomes, or treatment effects.