Balancing Design Intentionality and Serendipitous Emergence: How Visual Designers Navigate Generative AI Co-creation

Abstract

Generative artificial intelligence (GenAI) tools including Midjourney, Stable Diffusion and Comfy user interface (UI) have evolved from experimental novelties to widely used commercial tools integrated into professional visual design workflows. Yet diffusion models’ probabilistic nature creates persistent tension for designers between preserving deliberate design intent and embracing unplanned emergent visuals. Unlike conventional digital tools enabling predictable direct editing, GenAI often yields unforeseen outputs, swinging practitioners between creative frustration and fortuitous inspiration. To explore how designers resolve this conflict, this qualitative empirical study recruited 10 experienced commercial visual, illustration and UI designers. Drawing on dual evidence: Stimulated recall interviews and process records including prompt logs and draft visuals, alongside thematic analysis, it develops a three-stage spiral model: “Loss of Control - Taming - Symbiosis”. Results: (1) Designers’ cognitive-emotional journey is not a simple linear shift from anxiety to joy, but spiral cognitive reappraisal triggered by chance discoveries. (2) Four operational strategies manage algorithmic variability across a control-divergence spectrum: prompt denoising, regional inpainting, reverse engineering, hybrid collaging. (3) Designer agency persists, shifting from low-level manual manipulation to high-level meta-control over style rules, constraints and selection. This framework explains reshaped human creative agency in human-AI co-creation and informs future AI interface and design education.