Evaluating Artistic Creativity Through Neural Models

The increasing overlap of artificial intelligence (AI) with artistic creativity is an exciting prospect of learning more about how machines may be used to copy, enhance, or even recreate human creative behaviors. In this paper, the neural frameworks of self-assessment of artistic creativity are discussed with reference to the generative models of Generative Adversarial Networks (GANs), diffusion models, and transformer-based frameworks. Although the art created by AI is still developing and getting more advanced and diverse, the issue of evaluating creativity remains a multi-dimensional problem. We suggest a holistic assessment system, which combines both quantitative and qualitative assessment methods in order to measure creativity in visual, textual, and auditory contexts. The quantitative analysis is based on objective measurements, i.e. originality, diversity, aesthetic coherence, to which statistical standards and comparisons with other models are applied. Instead, the qualitative framework focuses on a human based assessment that considers the expert judgement, crowd-sourced assessment, and cognitive views of emotional and psychological involvement. The study will fill the gap between algorithmic generation and perceptual appreciation of art by taking a hybrid approach to it by combining computational and humanistic perspectives. Experimental results show the different neural structures that exist to communicate different creativity types and how these can be experimentally tested in a systematic protocol. Finally, the study leads to a better comprehension of the creative potential of AI and creates a systematic approach to assessing artistic innovativeness in the computational systems

Cognition · Comprehension · Computational Creativity · Creativity · Creativity technique · Generative grammar · Perception · Aesthetic Perception and Analysis · Creativity in Education and Neuroscience · Music Technology and Sound Studies