IdeaBlocks: Expressing and Reusing Exploratory Intents for Design Exploration with Generative AI

Abstract: Generative AI opens new possibilities for design exploration by rapidly generating images aligned with user goals. However, our formative study (N=7) revealed three key limitations hindering designers' broad and efficient exploration when interacting with these models. These include difficulty expressing open-ended exploratory intent, lack of continuity in exploration, and limited support for reusing or iterating on previous ideas. We propose IdeaBlocks, where users can express their exploratory intents to generative AI with structured input and modularize them into Exploration Blocks. These blocks can be chained for continuous, non-linear exploration and reused across contexts, enabling broad exploration without losing creative momentum. Our user study with 12 designers showed that participants using IdeaBlocks explored 112.8% more images with 12.5% greater visual diversity than the baseline. They also developed ideas in more iterative and continuous patterns, such as branching, chaining, and revisiting ideas. We discuss design implications for future tools to better balance divergent and convergent support during different phases of exploration, and to capture and leverage exploratory intents more effectively.