Bottom-up vs. Top-down

Abstract

The emergence of tools that support fast-and-easy visualization creation by non-experts has made the benefits of InfoVis widely accessible. Key features of these tools include attribute-level operations, automated mappings, and visualization templates. However, these features shield people from lower-level visualization design steps, such as the specific mapping of data points to visuals. In contrast, recent research promotes constructive visualization where individual data units and visuals are directly manipulated. We present a qualitative study comparing people's visualization processes using two visualization tools: one promoting a top-down approach to visualization construction (Tableau Desktop) and one implementing a bottom-up constructive visualization approach (iVoLVER). Our results show how the two approaches influence: 1) the visualization process, 2) decisions on the visualization design, 3) the feeling of control and authorship, and 4) the willingness to explore alternative designs. We discuss the complex trade-offs between the two approaches and outline considerations for designing better visualization tools.

iVoLVER

“…This is a difficult challenge as the advantages of bottom-up visualization tools-a deep involvement and understanding of the visualization process-seem to be incompatible with the benefits of top-down visualization tools: the quick and easy visualization, even of large datasets, made possible through automation [44]. To investigate possible solutions to this challenge in-depth and from multiple perspectives, we followed a mixed design methodology driven by: (1) our own expertise as visualization researchers with significant experience in designing visualization tools and UIs, and (2) the ideas from a larger group of people with varying degrees of expertise in visualization and visualization tools, who took part in four design workshops we organized.…”

“…Such tools do not require textual programming skills as opposed to specialized languages and libraries (e.g., D3 [7]). Méndez et al [44] discuss approaches supported by visualization tools as a continuum between two ends: bottomup and top-down. The former, represented by tools such as iVoLVER [45], promote a hands-on, constructive visualization process [28,30,32], but are limited in terms of scalability.…”

“…In visualization tools, automation facilitates visual mapping of large datasets enabling designers to rapidly explore different visual representations [44]. Many existing tools feature recommendation modules that suggest designs as users manipulate data elements (e.g., Tableau's "Show Me" [40] and similar strategies in other tools [8,18,56,58,72]).…”

“…People experience different levels of control and authorship of the visualization design process when working with tools that implement opposite levels of automation [44]. Tori and Möller's taxonomy emphasizes the human aspect of visualization techniques by considering "how much the designer chooses display attributes" [69, p. 154].…”

“…Some existing evidence shows that requiring more effort from the designer has advantages in terms of learning and understanding [30], as well as an increased feeling of authorship and transparency [44]. However, the fundamental limitation of these constructive tools (e.g., [45], and their physical counterparts [6,29,32]) is that they do not scale to larger datasets and can be tedious and repetitive to use [44]. In this paper we investigate how to reconcile automation and computer-assisted design with the benefits of constructive visualization.…”