Emil Polyak

Before the Prompt: Heteromorphic Imagination and the Cognitive Cost of Premature Crystallization in Human-AI Co-Creativity Generative AI systems have been shown to enhance individual creative output while simultaneously homogenizing collective creative diversity. Current explanations for this paradox focus primarily on output-level analysis, attributing the effect to design fixation, distributional convergence in model responses, or the narrowing effects of AI-generated examples on ideation. This paper proposes a complementary cognitive-phenomenological explanation grounded in the distinction between two modes of imaginative cognition: heteromorphic imagination, characterized by unconstrained, pre-linguistic associative play across real and unreal combinatory spaces, and creative imagination, where emergent patterns achieve sufficient resonance to warrant articulation. We argue that prompt-based interaction with generative AI systems imposes a demand for premature crystallization---a forced translation from the pre-linguistic heteromorphic domain into linguistically structured, model-interpretable expression---that structurally incentivizes bypassing the very cognitive phase most responsible for radical creative origination. Drawing on recent neuroscience demonstrating the causal role of the Default Mode Network in divergent thinking and the predictive significance of dynamic switching between spontaneous and controlled cognition, we advance the hypothesis that habitual prompt-based ideation may, over time, diminish the use-dependent neural infrastructure that sustains heteromorphic exploration. The implications for computational creativity research are substantial: evaluation of human-AI co-creative systems should account for effects on upstream human cognitive processes, and the heteromorphic dimension represents a missing variable in current models of co-creativity.