How We Know It, Is What Really Matters: Incorporating UI Components with Distinctive Variations

UI components for front-end developers can be widely reused by parameterizing visual and behavioral properties. Although flexible, this makes assumptions difficult, as developers have to think about many property values ​​and interactions. Essentially, they should explore the larger design space of the component and assign realistic and natural values ​​to the structures. To address this, we introduce a differentiating variable: a variable that is both imitative and unique. We develop a generation classification framework as a sample space, combining symbolic assumptions to identify key structures with mimetic sampling driven by LLM to generate realistic hypotheses from its world knowledge. We're introducing segmentation variations to Celestial, a tool that helps developers explore and visualize segmentation variations. In a study with lead engineers (n=12), participants found this variable useful for comparing and mapping component spaces, reported that the mimetic measurement was consistent with the domain, and confirmed that Celestial transformed part of the action from a manual process to a planned, experimental task.