AI-Driven Digital Creativity: Conceptual Foundations, Theoretical Framework, and Future Research Directions

Abstract

With the emergence of AI-generated digital artworks in online creative platforms, computational creativity, market economics, and social engagement dynamics are unprecedentedly intertwined. Although generative AI systems and the emerging digital art market have attracted attention within the scholarly field, the quantitative connection between art style, creator type, platform context, and the dual scoring of market price and audience engagement is poorly understood. This paper introduces Creative Artwork Deep Neural Network (CADNN), a novel multi-task deep-learning network for simultaneously predicting the price of an artwork in the art market and the audience engagement score from a comprehensive feature set of 18 engineered features, extracted from the curated dataset of 5,000 artworks from five major online platforms, across 3 categories of creators and 3 styles of art. CADNN combines Multi-Head Self-Attention (MHSA) with four heads, three residual connection blocks, Swish activation functions, Batch Normalization, and a dual-output prediction head using a weighted multi-task loss function. We show that CADNN outperforms five machine learning baselines on both tasks through exploratory data analysis, PCA and t-SNE dimensionality reduction, K-Means market segmentation, ablation studies on five architectural variants, Monte Carlo Dropout uncertainty quantification and attention weight-based feature importance mapping. Logarithmic view count, platform encoding and virality index are the most prominent variables in predicting the videos, and four market archetypes were identified using clustering. We identified a practical predictive framework and theoretical understanding of the computational architecture of digital creative markets, which has applications for the valuation of AI-generated art, the design of platforms, and the creation of generative models. Beyond its empirical contributions, this paper advances a macro-level theoretical framework for understanding AI-driven digital creativity as a sociotechnical phenomenon. We conceptualize AI not merely as a tool for content generation, but as an active participant in the creative ecosystem one that reshapes the ontological boundaries of authorship, the epistemological criteria for aesthetic value, and the economic logic of creative markets. Three foundational theoretical lenses are introduced: (1) Computational Creativity Theory, which situates AI-generated art within Boden's taxonomy of combinational, exploratory, and transformational creativity; (2) Platform-Mediated Value Co-Creation Theory, which frames digital creative platforms as two-sided markets where algorithmic curation, social amplification, and creator identity jointly construct value; and (3) Human-AI Creative Symbiosis Theory, which reconceptualizes the creator-tool relationship as a dynamic, co-evolutionary partnership rather than a substitution dynamic. These theoretical pillars collectively provide the conceptual scaffolding upon which CADNN's empirical architecture is built, and upon which future research in computational creativity, AI aesthetics, and digital market design should be grounded.