VizCrit: Exploring Strategies for Displaying Computational Feedback in a Visual Design Tool

Figure 1. We introduce VizCrit, a design feedback tool that provides three levels of actionability: static, textbook-based feedback via text to awareness-centered (top) and finally solution-centered (bottom) feedback via adaptive visual annotations. To design our interactive annotation interface, we aimed to better understand how instructors provide feedback (left). Informed by our co-design with instructors, we designed a set of visual annotations (right) for design principles (pictured here is alignment) and developed algorithms for heuristically computing these annotations. We display these visual annotations as overlays in a visual design tool and study how they influence novices' design processes, learning, and creativity.

Abstract

Visual design instructors often provide multi-modal feedback, mixing annotations with text. Prior theory emphasizes the importance of actionable feedback, the “actionability” lies on a spectrum—from surfacing relevant design concepts to suggesting concrete fixes. How might computational tools support this range? And how does feedback style impact novice designers' interaction with feedback, creativity, learning, and design quality? We introduce VizCrit, a system for providing computational feedback that supports the feedback spectrum, realized through algorithmic issue detection and annotation generation. In a between-subjects study (N=36), novices revised a design with one of three conditions: textbook- based, awareness-centered, or solution-centered. Solution-centered annotations reduced design issues, but created a mismatch be- tween participants' self-sense of creativity and expert assessments. Awareness-centered annotations encouraged self-reflection. For all design novices across conditions, they risked productivity-focused mentality. We discuss implications for AI in creativity support tools to help novices not only produce visual designs but also develop their design awareness.