Too Many Cooks: Exploring How Graphical Perception Studies Influence Visualization Recommendations in Draco

TL;DR

This work builds a replicable pipeline to translate a broad set of graphical perception results into Draco, a constraint-based visualization recommender, and uses a case study plus exploratory analyses to show how individual perception papers shift Draco's soft-constraint weights and, in turn, the resulting visualizations. By translating 30 papers from Zeng & Battle into Draco specifications and analyzing weight-shift and recommendation-shift patterns, the authors identify coverage gaps, consensus/discord among studies, and the conditions under which perception results reinforce or cancel each other. The findings demonstrate a quantitative meta-analysis approach for graphical perception, reveal practical implications for updating and improving visualization recommenders, and highlight opportunities to extend frameworks and automate knowledge incorporation as new perception results emerge. Overall, the paper advances understanding of how perceptual knowledge shapes downstream visualization guidance and provides a blueprint for integrating large perception literatures into recommender systems. The work also raises cautions about interaction effects, data-model alignment, and the need for richer design- decision variables in future perception studies and recommender frameworks.

Abstract

Findings from graphical perception can guide visualization recommendation algorithms in identifying effective visualization designs. However, existing algorithms use knowledge from, at best, a few studies, limiting our understanding of how complementary (or contradictory) graphical perception results influence generated recommendations. In this paper, we present a pipeline of applying a large body of graphical perception results to develop new visualization recommendation algorithms and conduct an exploratory study to investigate how results from graphical perception can alter the behavior of downstream algorithms. Specifically, we model graphical perception results from 30 papers in Draco -- a framework to model visualization knowledge -- to develop new recommendation algorithms. By analyzing Draco-generated algorithms, we showcase the feasibility of our method to (1) identify gaps in existing graphical perception literature informing recommendation algorithms, (2) cluster papers by their preferred design rules and constraints, and (3) investigate why certain studies can dominate Draco's recommendations, whereas others may have little influence. Given our findings, we discuss the potential for mutually reinforcing advancements in graphical perception and visualization recommendation research.

Figures (9)

• Figure 1: The visualization specification used in Zeng & Battle's dataset.
• Figure 2: A demonstration of the Draco visualization specification and how Draco-Learn can detect and then learn visualization preferences from a ranked visualization pair concluded from Saket2018taskSaket2018task.
• Figure 3: A demonstration of translating existing theoretical perception rules Mackinlay1986automating to Draco-Learn Yang2023draco2 training data. When comparing specifications encoding nominal data using position-x (positive) versus color hue (negative), Draco learns to avoid categorical color hue encodings (in red).
• Figure 4: We conduct mini-experiments to investigate how Draco-Learn shifts the weights of soft constraints based on different training data. The left side shows the four ranked visualization pairs applied in the mini-experiments, as well as the soft constraints that are only seen in the corresponding visualization. The right side shows how the weights of soft constraints shift compared to the baseline for each mini-experiment.
• Figure 5: The trend in weight shifts when adding duplicate single pairs. Including pair (a) multiple times increases the weights of the corresponding constraints almost linearly at first but levels out after about five inclusions.