- Data visualization practices can reinforce inequities and mislead, misrepresent, and harm communities that have experienced inequities and discrimination. Applying an equity lens is essential to more accurately representing and communicating with the communities served by researchers and others.
- As part of CORE's Insights blog series, we share six key steps to help apply an equity lens to your data visualization efforts.
- View the webinar below or read on for ideas and insights from the Urban Institute's Do No Harm guide, Catherine D'Ignazio and Lauren F. Klein’s Data Feminism, and real-world examples from CORE’s community partners.
Presenting data in visually compelling ways helps audiences understand it more easily and make decisions faster. Yet, according to the Urban Institute’s Do No Harm Guide, “When data are collected and communicated carelessly, data analysis and data visualizations have an outsized capacity to mislead, misrepresent, and harm communities that already experience inequity and discrimination."
In a recent webinar, Kristen Lacijan and Claire Devine of Providence CORE explained how applying an equity lens to data visualization can help you respond to these challenges, understand potential harms, and work to address them. As part of the webinar, Kristen and Claire shared six steps toward applying an equity lens to data visualization, leveraging insights from the Do No Harm Guide and Catherine D'Ignazio and Lauren F. Klein’s Data Feminism, along with real-world examples from CORE’s work and community partners. These are just a few of the resources we're exploring at CORE as we continue our learning around more equitable practices.
Read on for a summary of recommendations, or view the webinar video here.
1. Ask questions
Data users don’t always have control over how data are collected; for example, census data or Medicaid claims data. However, the more we can understand what’s behind the data and its collection, the more fully we can represent that context and its limitations when visualizing it. Questions to consider include:
- Why were these data collected?
- How were these data generated?
- Are the data demographically representative?
- Whose voices, lives, and experiences are missing?
- How we will communicate about missing voices, lives, and experiences?
- How much can these data be disaggregated by race, gender, ethnicity, etc.?
- Who stands to benefit from these data?
- Who might be harmed by the collection or publication of these data?
2. Include context
Once you understand that underlying context, communicate it in your graphics or figures to give audiences a more holistic understanding of the story behind the data. It is essential to be clear about what data are useful for, what they are not useful for, and what you are trying to say. Kristen and Claire highlighted several specific examples, including the following context from the Oregon Health Authority 2021 CCO Performance Metrics Dashboard’s section on race and ethnicity.
3. Choose language/imagery
Data visualization tends to reduce people and their experiences to bars, lines, points, colors, categories, icons, etc. Being conscious and careful about how you present data visually and with words can help build more trusting relationships with the communities behind it. Recommendations include:
- Always remember that data reflects the lives of real people.
- Use people-first language. For example, people in prison vs. inmates.
- Avoid using labels, hierarchy, color palettes, or icon selection that reinforce stereotypes or norms.
- Consider alternatives to using “other” as a catch-all category. For example, “another race” or “identity not listed.”
Claire pointed to an example from the Do No Harm Guide demonstrating how a graduated color palette can inadvertently suggest a hierarchy between races and reinforce White as a norm. 
Instead, use a color palette that avoids gradients and hierarchies, like this example from the King County Community Health Needs Assessment, 2021-2022.
4. Understand perspectives
While data are often perceived as objective, Claire emphasized that it’s impossible to avoid interpretation. She cited the following graphic from Data Feminism illustrating how the same data can tell different stories depending on their presentation. As you can see, both figures are factual, but each tells a different story.
The King County, Washington Communities Count page below offers a model example of how you can help viewers understand the perspective behind your data visualization. The subhead “Students of color and LGB+ students are less likely to feel safe at school” helps clarify the key takeaway right from the onset.
5. Involve communities
Relationships with communities are essential to building trust with research subjects. Connecting with people impacted by your research can also give meaning and context to quantitative data collected about those communities.
For example, Kristen pointed to the Best Starts for Kids program in Washington State, which convened members of underrepresented communities for a series of “data dive” events where Best Starts for Kids staff shared survey results with community members. One attendee’s comment captured this approach’s potential impact:
“We are in a developed nation and rich county, but we are at the lowest level economically. We don’t understand the system and there is a need for systems understanding for us. Coming to see us like you are doing right now helps us. We need a flow of continuous community. We need a flow of conversations that are coming back and forth not just one visit.”
Another example of a community-led research model comes from the Oregon Pacific Islander Coalition’s 2021 report: Pacific Islander Data Modernization in Oregon. As part of this approach, government representatives signed a data sovereignty agreement to ensure that the Pacific Islander community served as research experts, engagement experts, language experts, and writers; was properly compensated for the expertise that they provided at every stage of the project; and utilized the Oregon Health Authority’s Public Health Division as a technical assistance resource rather than as a governing body for the work.
6. Communicate with your audience
Lastly, data is most impactful if it can be shared and understood. That’s why it’s essential to determine which mediums, language, and ease of access will meet the needs of your specific audiences, including research subjects. The following diagram from We All Count shows the sweet spot at the center of this three-legged stool.
For more information and ideas about equitable data visualization, check out the resources below, including several referenced frequently in this blog post and webinar.
