“[T]he choices we make about the colors, shapes, words, and representations in our data analysis and visualizations can affect how people perceive the final results, how change might be implemented, and how that change will impact different people and communities ” (Schwabish and Feng, 2021).
It is important to be inclusive in all visual aspects of a publication, not just photographs or other images of people. In their guide Do No Harm: Applying Equity Awareness in Data Visualization, Schwabish and Feng provide the following recommendations.
Demonstrate empathy. Help readers connect with the data and remember that it represents the lives and experiences of people. This might be done by pairing data-driven charts with personal stories or by making charts interactive. Also, recognize the needs of your audience. Is the content accessible? Would it be more useful if it were translated into a different language? Is the issue being framed in a way that introduces bias? Empathy can also be embedded into the graphics themselves. For example, consider using icons for data points that represent or reflect people instead of dots or bars on a graph.
Reflect lived experiences. As mentioned in the writing section, whenever possible it is best if data can be developed in conjunction with communities instead of about them. Ideally this would be considered from the beginning of the project.
Consider the data. Are the data biased or the product of oppressive data collection systems? Data visualizations amplify data and may therefore amplify the harm that bias can do. Consider how and why the data were collected and generated. Who is represented and who is missing? Who will benefit from or be harmed by the data? Schwabish and Feng note that applying the principles outlined in their guide will not fix data or analysis that is inherently biased.
Use language with an equity awareness. Titles, text, and labels are important as they are often the first thing readers see when looking at a graph or chart. The principles outlined in the language section of these guidelines should be considered when writing text to accompany data visualizations. For example, use person-first language and inclusive language in chart labels. Schwabish and Feng use the example of a graph showing the relationship between race and poverty where the axes were labeled “More Black” and “More Poverty.” They note that, “A more inclusive way to label the legend might be ‘Larger proportion of people experiencing poverty’ and ‘Larger Black population.’”
Order data purposefully. Consider the order in which data is presented, particularly if it presents people in an implied hierarchy. For example, do all the charts depicting demographic information start with “White” or “Male” as the default or first position? Consider starting with the particular group the study is focused on or sorting the data alphabetically or according to sample or population size. This is another example where interactive charts may be useful.
Consider missing groups. Noting information that is missing or limited acknowledges the missing communities and may encourage others to include them in the future. Conversely, if the data is not directly relevant to the study or the point being made, does it need to be collected at all? Also consider whether aggregating smaller sample sizes is the best approach as it may inadvertently harm or erase those communities. If necessary, Schwabish and Feng offer several alternatives to the label “Other,” including “another race,” “additional groups,” “all other self-descriptions,” and “identity not listed.” Regardless of the term, the AMA recommends noting which groups are included in such categories (Frey). Finally, if data about a group is purposely excluded, it may be worth mentioning. For example, in situations where sample sizes are too small, this may be noted instead of omitting the category.
Using colors with an equity awareness. Color palettes should meet basic accessibility guidelines, offer sufficient contrast for readers with low vision or colorblindness, and avoid reinforcing gender or racial stereotypes. Schwabish and Feng provide the example of a chart showing racial and ethnic data where “White” is represented in blue, “International” and “Unknown” in shades of gray, and “Black or African American,” “Hispanic or Latino,” “American Indian or Alaskan Native,” Native Hawaiian or Other Pacific Islander,” “Asian,” and “Two or more races” are all varying shades of red. The effect of this graph is that “White” is seen as default and all other categories are either lumped together or fade into the background. Also consider the connotation that certain colors may have and consider omitting them. Orange, yellow, green, and red may fall into this category (Calder).
Using icons and shapes with an equity awareness. Consider whether the icons and shapes used are appropriate given the context or if they might reinforce stereotypes. Schwabish and Feng use the example of an icon of a baby in a chart about infant mortality. Other examples are similar to those for photographs. Are nurses always represented as women and doctors as men? Are Black people always in images representing poverty or crime? Also consider whether the use of a specific icon might cause readers to misrepresent the data.
Embrace context and complexity. According to Schwabish and Feng, data should not “speak for themselves” because data are not neutral. More complex designs should be used when necessary to foster better understanding of the data. Conversely, Schwabish and Feng also note that not all data needs to be visualized.
Accessibility. It should be noted that many data visualizations are exclusive of people with disabilities. Marriott et al. notes that such designs are “premised on implicit assumptions about the reader’s sensory, cognitive, and motor abilities.” For example, in the spring of 2020 during the early months of the COVID-19 pandemic, a popular data visualization encouraged readers to “flatten the curve.” The graphic was a quick and simple way to understand the situation, but such depictions are not accessible to people who are blind or have low vision (Ehrenkranz, Marriott). Marriott et al. points out that interactive graphics may also be inaccessible to users with motor disabilities. See the next section for more information on accessibility.
Calder, C., & Supadulya, V. (2019, January 9). Inclusive color sequences for data viz in 6 steps. Medium. https://medium.com/design-ibm/inclusive-color-sequences-for-data-viz-in-6-steps-712869b910c2
Ehrenkranz, M. (2020, April 9). Vital coronavirus information is failing the blind and visually impaired. VICE. https://www.vice.com/en/article/4ag9wb/vital-coronavirus-information-is-failing-the-blind-and-visually-impaired
Frey T, Young RK. Race and Ethnicity. In: Christiansen S, Iverson C, Flanagin A, et al. AMA Manual of Style: A Guide for Authors and Editors. 11th ed. Oxford University Press; 2020. https://www.amamanualofstyle.com/view/10.1093/jama/9780190246556.001.0001/med-9780190246556-chapter-11-div2-23
Marriott, K., Lee, B., Butler, M., Cutrell, E., Ellis, K., Goncu, C., Hearst, M., McCoy, K., & Szafir, D. A. (2021). Inclusive data visualization for people with disabilities: A call to action. Interactions, 28(3). https://interactions.acm.org/archive/view/may-june-2021/inclusive-data-visualization-for-people-with-disabilities
Schwabish, J., & Feng, A. (2021, September). Do no harm guide: Applying equity awareness in data visualization. Urban Institute. https://www.urban.org/research/publication/do-no-harm-guide-applying-equity-awareness-data-visualization