Abstract
<title>Abstract</title> <p>Digital drawing provides a process-based record of how emotional states are expressed through color, stroke, spatial organization, and drawing behavior. This study examined whether AI-assisted prompting influences emotional expression in digital drawing and whether drawing-derived features offer preliminary evidence for emotion prediction. Thirty-one participants completed twelve emotion-based drawing tasks on an online platform and were randomly assigned to either an AI-assisted prompting or control condition. After data screening, 363 valid drawings were retained. Color, stroke, spatial, and behavioral features were extracted from each drawing. Linear mixed-effects models were used to examine the effects of AI-assisted prompting on drawing complexity and behavioral indicators while accounting for repeated measures across participants and emotion conditions. Exploratory machine learning models were also trained, and an interactive prototype was developed to demonstrate real-time prediction based on drawing features. The results showed that participants in the AI-assisted condition produced drawings with higher levels of visual complexity and behavioral engagement than those in the control condition, including longer drawing duration, higher stroke counts, more trajectory points, larger drawing areas, and greater color diversity. Exploratory machine-learning results indicated limited performance in classifying twelve fine-grained emotion categories, with comparatively better performance when emotions were grouped into broader affective categories. Overall, these findings suggest that digital drawing features may reflect general affective tendencies, and the prediction prototype should be regarded as an exploratory research tool rather than a diagnostic system.</p>