A Design Space for Multiscale Visualization

TL;DR

Multiscale visualizations face challenges when the range between largest and smallest items is large. The authors introduce a design space with three dimensions and eight subdimensions to capture low-level design decisions across scales and validate it with a coded corpus of 52 examples. They identify four high-level strategies, analyze missed opportunities for alternative encodings or navigation, and compare patterns across analysis versus presentation contexts. The work demonstrates descriptive and generative power, provides supplementary materials and a public website for practitioners, and discusses limitations and avenues for future validation.

Abstract

Designing multiscale visualizations, particularly when the ratio between the largest scale and the smallest item is large, can be challenging, and designers have developed many approaches to overcome this challenge. We present a design space for visualization with multiple scales. The design space includes three dimensions, with eight total subdimensions. We demonstrate its descriptive power by using it to code approaches from a corpus we compiled of 52 examples, created by a mix of academics and practitioners. We demonstrate descriptive power by analyzing and partitioning these examples into four high-level strategies for designing multiscale visualizations, which are shared approaches with respect to design space dimension choices. We demonstrate generative power by analyzing missed opportunities within the corpus of examples, identified through analysis of the design space, where we note how certain examples could have benefited from different choices. We discuss patterns in the use of different dimension and strategy choices in the different visualization contexts of analysis and presentation. Supplemental materials: https://osf.io/wbrdm/ Design space website: https://marasolen.github.io/multiscale-vis-ds/

Figures (27)

• Figure 1: Nine representative examples from the corpus, stylized for clarity. Blue indicates marks or features in the visualization. Orange indicates user interaction. Examples A-G are all on one screen. In examples H and I, the arrows indicate changes to what is shown on screen. When the scale counts here differ from those of the original examples listed in Table \ref{['tab:coded-examples-dimensions']}, our stylization involved a change to scale counts for simplicity. A) xkcd Money munroe2011xkcdmoney B) MizBee meyer2009mizbee. C) Rivet (multi-tier strip chart) bosch2000rivet. D) Melange elmqvist2008melange. E) DeLVE solen2024delve. F) Large Viewing Vis isenberg2013largeviewingvis. G) The Mandelbrot Explorer bau2009mandelbrot. H) Powers of Ten eames1968powersoften. I) EVEVis miller2011evevis.
• Figure 2: The four strategies for designing multiscale visualizations.
• Figure 3: Proposed redesign of Hierarchical Route Maps wang2014hierarchicalroutemaps. A) Stylized original. B) Stylized redesign, with second separate scale showing an overview connected using association by marks.
• Figure 4: A proposed redesign of Chromoscope lyi2023chromoscope. A) Stylized original. B) Stylized redesign, with association by both marks and channels.
• Figure :