Journal of Networking Technology Lanzhi Cheng 17 1 2026 https://doi.org/10.6025/jnt/2026/17/1/28-41 https://www.dline.info/jnt/fulltext/v17n1/jntv17n1_3.pdf This paper explores the integration of entropy-driven methods and AI-assisted neural networks in art, design, and education. It highlights how convolutional neural networks (CNNs), enhanced by deeper architectures and adaptive algorithms, enable precise image analysis and generation in creative fields. Entropy is used as a quantitative measure of information richness to guide image selection, improve classification efficiency, and support objective evaluation in art education where traditional assessment struggles with subjectivity. The study also emphasizes the role of explainable AI (XAI) and concept based models to ensure transparency in AI-generated artistic outputs. A key concern is user trust, which is influenced not only by image realism but also by perceived AI involvement and cultural sensitivity. Empirically, the authors validate a 17-4-1 backpropagation (BP) neural network using NeuroSolutions software to evaluate art design professionals. Trained on expert judgments and enterprise criteria, the model achieves high accuracy, with MAE and MAPE of approximately 1.86%, RMSE of 2.27%, and a maximum error of less than 4.5%. This demonstrates strong generalization and reliability for talent assessment. The research proposes a unified conceptual framework linking digital inputs, entropy based analysis, AI processing, XAI, human perception, and creative outcomes. Ultimately, the study advocates for culturally aware, explainable, and humancentered AI systems that enhance rather than replace human creativity, originality, and pedagogical effectiveness in art and design.