A Design Science Framework for the Multidimensional Evaluation of AI-Generated Drawing and Painting

Abstract

The rapid advancement of Artificial Intelligence (AI) has transformed visual art, drawing, and painting. Creativity in artistic expression used to be exclusive to humans; however, with advances in AI, it is augmented and, in some cases, performed autonomously by algorithms. This study presents a comprehensive investigation into the evolution, mechanisms, and implications of AI-generated drawing and painting technologies, employing Design Science Research Methodology (DSRM) as its guiding framework. The study conducts a structured analysis of generative models, including Generative Adversarial Networks (GANs), diffusion models, Variational Autoencoders (VAEs), and Neural Style Transfer (NST), to examine how these systems simulate, replicate, and, at times, transcend conventional artistic processes. A custom evaluation framework is developed and applied to assess AI-generated artworks across six criteria: aesthetic quality, technical fidelity, stylistic consistency, semantic coherence, originality, and emotional resonance. The results demonstrate that state-of-the-art diffusion-based models achieve low Fréchet Inception Distance (FID) scores compared to a human reference set in controlled evaluations, while also revealing persistent limitations in semantic coherence and cultural expressivity. In addition, the study addresses the ethical, philosophical, and socioeconomic consequences of AI-generated artworks, including questions of authorship and copyright.