top of page

Testing guardrails for graphical information

Updated: Aug 30, 2023

Communicators sometimes think that the public cannot “handle” numbers. However, when provided them, people can understand more and act in healthier ways. For example, providing numeric likelihoods of a medication’s side effects increased risk understanding and willingness to take it. Similar positive numeric effects emerged in vaccines, cancer screening, and treatment advertising. Further, people trust and engage more with messages that include numbers. How numbers are formatted though can matter as much as what numbers are presented.

In this project, eight studies were conducted in order to develop a set of concrete recommendations about data visualizations. The recommendations include tested guardrails (sets of do’s and don’ts for visualizing data). We recruited 1,648 online participants divided about evenly between younger adults (25-34 years), middle-aged adults (35-54 years), and older adults (55+); we further recruited some participants aged 18-24 years. All participants completed baseline measures in an initial session and then responded to eight experiments divided into two subsequent sessions. In each experiment, participants were given graphical information or a no-graphic control condition and answered questions assessing comprehension of the numeric information.

For example, in Experiment 1, some participants saw a graph like the one below. In other conditions, instead of authorized immigrants, participants were shown the total number of immigrants. In the no-graphic condition, they only saw the text. For other participants, the graphs were included but the text was omitted. Participants were asked questions such as “True or false: The number of unauthorized immigrants has increased more than the number of authorized immigrants over the past three decades.”

From these data, we developed overall recommendations for people of moderate numeracy (i.e., those that answer two out of four numeracy questions correctly). We summarize three of those recommendations here.

1. Provide data visuals. In three experiments, we included control conditions where graphical information was not provided—for immigration, air quality, and COVID-19 vaccination rates, respectively. In all three cases, comprehension was higher when graphics were provided vs. not.

2. Provide less information to increase comprehension. We sometimes gave participants cues—such as color, bubble size, labels, and trend-lines. Providing cues helped comprehension of questions related to their content (i.e., size of changes in wages) but hurt comprehension of questions unrelated to their content (i.e., current wages). For example, participants viewed figures depicting changes in wages by industries over a 15-year period. Figures showed industries ordered by the highest wage in 2019 or the size of the change over time.

To half of the figures, we added category labels indicating whether the change was a “large increase,” “small increase” or “decrease.” These labels helped participants identify the industry with the largest decline (below, left), but made it more difficult for them to identify the industry with the highest wage (below, right).

3. Provide information in a format that matches the communication goal. The format of information often increased understanding of one aspect of the data while hurting comprehension of another aspect.

For example, all participants viewed graphs with the number of authorized immigrants and either the number of unauthorized immigrants or the number of total foreign-born.

When it was graphed with total foreign-born people in the US, participants understood the total better (green bar on the left below) and understood less about the number of unauthorized immigrants (green bar on the right below). The opposite occurred when we graphed the number of authorized immigrants and unauthorized immigrants (gray bars below). In other words, graphing two categories (authorized and unauthorized immigrants) helped comprehension of the categories, but hurt comprehension of the total (relative to graphing one category and the total).

For more on how to present numbers so that they matter, check out Innumeracy in the Wild: Misunderstanding and Misusing Numbers (Oxford University Press). “Policies in health and financial domains have shifted toward giving consumers and patients more information (which is often numeric). These changes are intended to empower individuals to take charge of their own welfare. The evidence is clear, however, that not everybody is prepared to use this information effectively and that those who are less numerate tend to make worse decisions unless provided adequate support. The book discusses … methods that exploit existing evidence and enable those who are less comfortable with numbers to use them more effectively and make better choices in an often numeric world.”

This work was supported by grants provided by the National Science Foundation (SES- 2017651) and USAFacts. The content of this post is solely the responsibility of the authors and does not necessarily represent the official views of the National Science Foundation or USAFacts.

209 views0 comments


bottom of page