May (A)I Beautify Your Visualization? Expert Judgments of Acceptable Aesthetic Alterations

Jul 16, 2026

Upload by Logic

In 3D visualizations of natural phenomena, aesthetic improvement can deliver measurable gains in memorability, learning, credibility, interpretability, and usability, yet it often relies on transformations that change how data is perceived. An expert survey of 95 visualization researchers, practitioners, and domain scientists evaluates fifteen alterations spanning presentation-level edits (e.g., lighting, camera position) and data-level modifications (e.g., denoising, removing or filling gaps) by humans and AI. Acceptability depends on the transformation’s meaning rather than its pipeline level. AI-generated versions receive consistently lower acceptability than identical human-produced changes.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages5

Reading Time13 min

Introduction

Related Work

May (A)I Beautify Your Visualization? Expert Judgments of Acceptable Aesthetic Alterations Kalina Borkiewicz* University of Utah Jixian Li University of Utah Joshua A. Levine University of Arizona Katherine E. Isaacs University of Utah arXiv :2607 .00239v1 [ cs .HC] 30 Jun 2026 Figure 1: Examples of aesthetic alterations from our study. Top row: a subset of abstract transformation categories. Bottom row: pairs of original (lower-left) and AI-generated (upper-right) visualizations. These examples illustrate a range of alterations that may preserve, enhance, or distort the message or features of the data. AI images generated with Nano Banana Pro and Qwen-Image. ABSTRACT In 3D visualizations of natural phenomena, improving aesthetics can provide measurable benefits, but often involves transformations that affect how the data is perceived. As a growing range of tools—including AI-based methods—make visual design and modification more accessible, it is increasingly important to understand tradeoffs and concerns when making these changes. We conducted an expert survey (N=95) with visualization researchers, practitioners, and domain scientists, investigating reactions to fifteen alterations spanning presentation-level adjustments (e.g., lighting, camera position) and data-level modifications (e.g., removing errors, filling gaps), applied by both humans and AI systems. Results show differences in perceived acceptability are driven by the transformation’s meaning, regardless of whether it operates at the presentation or data level. Additionally, certain modifications were consistently judged as more permissible than others regardless of human or AI authorship. While this relative ordering remains largely stable, AIgenerated transformations are consistently rated as less acceptable than identical human-produced changes. These results reveal a distinction between more permissible and more sensitive alterations, and suggest the need for both designers and AI-assisted visualization tools to incorporate constraints and guardrails that reflect these differences. Index Terms: Cinematic scientific visualization, science communication, visualization aesthetics, generative AI. 1 INTRODUCTION Improving the appearance of a visualization can enhance memorability [4, 5], learning [31], credibility [7, 19], interpretability [2], and usability [8, 34] . These benefits are especially relevant for communicating scientific results to audiences broader than domain * e-mail: [kalina@sci.utah.edu](kalina@sci.utah.edu) experts. However, transforming a traditional or analytical visualization into one suitable for public communication may involve a wide range of modifications, from relatively superficial changes atthe level of presentation (e.g., camera position, lighting, or background) to alterations at the data level (e.g., denoising, feature omission/addition) . Some of these changes may affect how faithfully the visualization represents the underlying data or phenomenon. Recent advances in artificial intelligence (AI) introduce new ways to apply transformations (e.g., [20, 23, 32]) . These systems lower the barrier to producing aesthetic modifications, but also introduce new challenges such as limited controllability [26] and potential hallucination [14] . As AI tools become integrated into visualization workflows, there is a growing need to define the constraints and guardrails for applying such transformations. Rather than imposing assumptions about which AI features should be supported or restricted, we ground this problem in expert judgment by examining which aesthetic alterations are considered acceptable. We use acceptable to describe judgments about whether a transformation is appropriate for a given context, recognizing that such judgments may reflect a combination of concerns including accuracy, interpretability, and the potential to mislead. This framing also allows us to examine if these judgments shift depending o

FAQs

什么因素决定审美变换在科学可视化中的可接受性？

可接受性主要取决于该变换如何改变可视化所表达的含义，而不是变换发生在呈现层还是数据层。研究还观察到更可接受与更敏感的变换之间存在稳定分组。

研究比较了哪些类型的变换，以及由谁来执行？

研究考察了15种可视化变换，覆盖呈现层调整（如光照、相机位置）与数据层修改（如去噪、删除或补全特征/缺口）。这些变换由人类设计者执行，并在同样条件下由AI系统生成后进行评估。

AI生成的审美改动与人类改动相比，可接受性如何？

即使变换本身相同，AI生成的版本也会被一致评为更不可接受；总体排序在一定程度上保持稳定，但AI相关判断总体更严格。

研究提出的应用含义是什么？

结果表明需要将“更可接受”和“更敏感”的差异转化为约束与护栏，促使设计者和AI辅助可视化工具使用可解释、可验证的变换策略。

Jul 16, 2026

Upload by Logic