<![CDATA[This study aims to develop and evaluate a deep learning model for the comprehensive assessment of divergent and convergent creativity dimensions. The dataset comprised 102 digital drawings obtained from Indonesian children aged 4 to 6 years using the Test for Creative Thinking - Drawing Production (TCT-DP). This study employed a quantitative model development approach, where ground-truth labels were derived from the 14 TCT-DP scoring criteria aggregated into divergent and convergent scores through label engineering. Using a Multi-Task Convolutional Neural Network (MT-CNN) based on MobileNetV2 architecture, the study analyzed extracted visual features to predict expert-rated scores. The results revealed a strong positive correlation (r = +0.51) between divergent and convergent thinking scores, challenging the traditional view of these processes as antagonistic and supporting an integrated model of creative cognition. From a technical perspective, the model demonstrated satisfactory predictive capability as a proof-of-concept, achieving a lower error rate for convergent scores (RMSE = 1.52) compared to divergent scores (RMSE = 1.97). It indicates that while structured convergent features are more machine-learnable, the abstract nature of divergent thinking remains a complex challenge. In conclusion, this study validates the feasibility of automated creativity assessment while offering empirical evidence for the interplay between generative and evaluative thinking in early childhood.]]>
Copyrights © 2025
