IdeaBlocks: Expressing and Reusing Divergent Intents for Graphic Design Exploration using Generative AI

TL;DR

<3-5 sentence high-level summary>

Abstract

While designers increasingly leverage Generative AI for divergent exploration, current interaction is optimized for convergent refinement, forcing users to specify fixed targets rather than open-ended search spaces. Based on a formative study (N=7), we define the anatomy of Divergent Intent, comprising property, direction, and range, and identified two critical barriers: the lack of mechanisms to explicitly shape the parametric boundaries of exploration and the difficulty of reusing successful search strategies. We present IdeaBlocks, where users can modularize divergent intents into Exploration Blocks. Users can reuse prior intents at multiple levels (block, path, and project) with options for literal or context-adaptive reuse. In our comparative study (N=12), participants using IdeaBlocks explored 2.13 times more images with 12.5% greater visual diversity than the baseline, demonstrating how structured intent expression and reuse support effective divergence. A three-day deployment study (N=6) further revealed how different reuse mechanisms allowed distinct creative strategies, offering design implications for future intent-aware creativity supports.

Figures (9)

• Figure 1: Conceptual distinction between IdeaBlocks and prior works in supporting divergent intent. (a) Expression: Current GenAI interactions force users to specify a single point. Prior research supports broader exploration by adding divergence, but often lacks user control over the boundaries. IdeaBlocks enables users to shape parametric boundaries (Property, Direction, Range) of exploration. Within a selected property, users can freely change the direction of exploration to inspect different points in the property space. They can also control the range to widen or narrow the exploration boundary, allowing them to determine the desired level of variation within that property. (b) Reuse: Prior methods support reusing static assets or prompts. IdeaBlocks supports reusing the exploration vector (Direction + Range). By decoupling the strategy from the content, users can efficiently transfer successful exploration logic to new design contexts.
• Figure 2: Screenshot of the IdeaBlocks system with (a) the Blocks Storage Panel and (b) the Exploration Canvas. Users express their divergent intent through structured inputs in Exploration Blocks (F1), which can be chained to build on prior ideas across different properties. Users can reuse individual blocks (F2-1), entire exploration paths (F2-2), or the whole project, with optional adaptation.
• Figure 3: Users create an Exploration Block for the specific design property they want to explore, enter the exploration direction, and adjust the typicality level to control the range of the generated suggestions. IdeaBlocks generates four suggestions, either in text or image, depending on the property type. Users can refine these suggestions to adjust the exploration directions. The resulting images are shown in the adjacent Result Block. Users can link multiple blocks to build on prior exploration, and each Exploration Block and Result Block becomes the context for generating the next results, allowing users to maintain previously explored intentions while exploring new properties.
• Figure 4: IdeaBlocks's four features for reusing prior exploration. (a) Users can reuse a previously created Exploration Block from the sidebar, placing it on the canvas with its original settings or adaptively changed suggestion set aligned to the new context. (b) Users can copy and paste the entire Exploration Paths, choosing between literal copy and context-adaptive versions that are updated based on the new context. (c) Users can import an earlier project they worked on with a different topic and adapt it to the current one, or literally import another user's project.
• Figure 5: Survey responses on users' perceived support for intent expression (left) and intent reuse (right), with 95% confidence intervals. Participants rated IdeaBlocks significantly higher in their ability to express a range of intent and to reuse previously expressed intent.