🤖 AI Summary
To address rigid interaction, high cognitive load, and inflexible iteration in AI-assisted creative writing, this paper proposes an infinite-canvas-based composable prompt workspace. Methodologically, it introduces a novel widget-based, modular prompt paradigm—supporting system-recommended, natural-language-triggered, and manual widget construction—and integrates dynamic widget architecture, context-aware prompt synthesis, and a human-AI collaborative interface to enable multi-granular text control, visual ideation orchestration, and low-cognitive-load real-time iteration. Empirical evaluation includes a controlled lab study (N=18), demonstrating statistically significant improvements in creativity support metrics alongside reduced cognitive load and frustration; and a field study (N=10), confirming enhanced perspective generation and improved human-AI collaboration efficacy. The work advances prompt engineering by redefining prompts as interactive, composable UI elements rather than static textual inputs, thereby bridging the gap between expressive intent and generative AI capabilities in creative workflows.
📝 Abstract
We introduce PromptCanvas, a concept that transforms prompting into a composable, widget-based experience on an infinite canvas. Users can generate, customize, and arrange interactive widgets representing various facets of their text, offering greater control over AI-generated content. PromptCanvas allows widget creation through system suggestions, user prompts, or manual input, providing a flexible environment tailored to individual needs. This enables deeper engagement with the creative process. In a lab study with 18 participants, PromptCanvas outperformed a traditional conversational UI on the Creativity Support Index. Participants found that it reduced cognitive load, with lower mental demand and frustration. Qualitative feedback revealed that the visual organization of thoughts and easy iteration encouraged new perspectives and ideas. A follow-up field study (N=10) confirmed these results, showcasing the potential of dynamic, customizable interfaces in improving collaborative writing with AI.