Trustworthy by Design: The Viewer's Perspective on Trust in Data Visualization

TL;DR

This study addresses a gap in trust research by centering the viewer’s perspective on data visualization trust through a qualitative mixed-methods study. It identifies three core analytic themes—internal consistency within individuals, divergent priorities across users, and overarching trends—then derives actionable designer guidelines (Present Data Clearly, Choose the Right Type of Chart, Invest in Aesthetics, Leverage Familiarity, Educate Where Necessary, Cite Credible Sources). The findings underscore that readability, chart type, source credibility, and familiarity robustly shape trust, while aesthetics play a secondary but nontrivial role, informing practical guidance for crafting trustworthy visualizations. Although exploratory and limited by sample size and scope, the work offers a user-centered framework to guide visualization design and highlights avenues for empirical validation and broader literacy-focused metrics.

Abstract

Despite the importance of viewers' trust in data visualization, there is a lack of research on the viewers' own perspective on their trust. In addition, much of the research on trust remains relatively theoretical and inaccessible for designers. This work aims to address this gap by conducting a qualitative study to explore how viewers perceive different data visualizations and how their perceptions impact their trust. Three dominant themes emerged from the data. First, users appeared to be consistent, listing similar rationale for their trust across different stimuli. Second, there were diverse opinions about what factors were most important to trust perception and about why the factors matter. Third, despite this disagreement, there were important trends to the factors that users reported as impactful. Finally, we leverage these themes to give specific and actionable guidelines for visualization designers to make more trustworthy visualizations.

Figures (5)

• Figure 1: The categorization of keywords in our codebook. For definitions of each keyword, see \ref{['tab:codedefs']} in \ref{['appendix_codebook']}.
• Figure 2: The keywords used by each participant are colored based on the keyword category. Although the order of the rounds was randomized for each participant, we have aligned them in this figure to illustrate what participants attended when ranking the same charts. Participants used a small set of keywords, though the specific set varied between individuals.
• Figure 3: The frequency of different keyword categories broken down by education level, design experience, sex, and age. Each bar represents the average number of responses per participant that were associated with each category.
• Figure 4: The 17 visualizations used in the studypandey2023youborkin2013makes. These visualizations include line, bar, infographic, and more visualizations from news networks, scientific publications, and government sources. Each is given a letter ID from [a] through [q].
• Figure 5: These histograms represent the rankings given to each visualization in each round (with 1 as most trusted and 6 as least trusted). The visualizations in each round are ordered by sum-product score, with the left having the best (lowest) score and right having the worst (lowest) score.