Creative Ambiguity and Cognitive Tension in Generative AI Tools

Abstract

Generative AI tools are increasingly embedded in creative workflows, yet their outputs often misalign with user expectations, producing results that are ambiguous, contradictory, or oddly off-mark. Rather than treating such misalignments as failures to be corrected, this paper reframes them as sources of creative ambiguity and cognitive tension. Drawing on theories of cognitive dissonance, frame conflict, and prediction error, I show how these tensions can catalyze reflection, reinterpretation, and creative reframing. Through five cross-modal cases spanning writing, design, music, code, and prototyping I demonstrate how creative ambiguity and cognitive tension function as generative forces in human-AI interaction. This work contributes to HCI and creativity support by proposing a shift from optimization toward reflection, designing not to minimize friction, but to work with it.