Ethical Design Recommendations for COVID-19 Visualizations
The stakes are high, and the data is complex. Here are some practical recommendations for data designers working on COVID-19 visualizations
In March 2020, COVID-19 was classified as a global pandemic. Every day, we learn something new about the novel coronavirus and updated case numbers appear in datasets managed around the world. And every day, it seems, we also learn something new about data visualization.
Data Visualization Society members have been closely following the massive number of public visualization efforts using COVID-19 data. Some charts have informed and enabled understanding, while others have misrepresented the seriousness of the illness, unintentionally dehumanized or sensationalized represented subjects, or ignored the significant data quality issues around COVID-19 case and death data. Data visualization has played a key role in making sense of the current pandemic globally, and it will continue to do so.
The pandemic has resulted in an unusually large number of people, from a wide range of backgrounds, industries, and professions designing visualizations quickly, based on very similar, if not identical, underlying data. In many ways, COVID-19 data presents the same ethical visualization challenges as in pre-pandemic days. There’s a lot of advice around on what to do, and what not to do in visualizations, user experience, and in design and tech more generally that can be applied to any individual visualization. While all of these efforts have merit, they can be hard to find and piece together, let alone accommodate in a single workflow for the time-pressed designer.
The current context creates even greater complexity; the stakes are high and the data is complex. Case and death data have an unusually high degree of uncertainty for such a widely used data set, as do the forecasts and models built with that data. Visualizations are informing individual decision making and the sweeping policies that can literally influence life and death. Disproportionately impacted communities, in particular, may be fatigued or suffering from information overload, making it even harder to reach them.
Pause before you publish
The uncertainty, hunger for information, and sensitivity of data about human health and loss of life all come together to create an environment where visualization ethics considerations are more important than ever before. Many excellent resources offering guidance on visualizing COVID-19 data have come out recently (see list below). Their strongest recommendation? Don’t visualize. Or if you must, reflect deeply on your intentions first.
Despite the challenges and recommendations for caution, the reality is that many people need to create visualizations of COVID-19 data. Such people include activists, journalists, data analysts, health professionals, and policy advisors working on COVID-19 and related issues. With these people in mind and anticipating that the pandemic will have a long arc globally, we’ve assembled a streamlined, practical guide to using COVID-19 data ethically, outlined below.
Ethics Checklist for COVID-19 Visualizations
- Partner up
Find a subject expert to partner with. This is a person with in-depth subject expertise on healthcare, epidemiology, biostatistics, pandemics, and/or public policy.
Understand your audience and represented subjects. Consider bringing the voice of the patient into your process, particularly when visualizing sensitive data stories, including the human toll of COVID-19.
- Get strategic
Identify your specific audience. Determine what they need to know that you can say with the data at hand, and how important it is that they understand your message. People are making decisions with the information they’re seeing — do I continue to stay home and practice social distancing? Are things getting better near me?
- Identify stakes
Your proposed visualization could be misused or harm someone (particularly from a high-risk group). Charts may feel objective, but they can be easily misused particularly around a highly charged topic. Revisit this consideration throughout your design process as you make choices around the visual forms, colors, text, and narratives you reinforce with your visualization. If there is a strong potential for misuse or misunderstanding, revisit your intention in creating the visualization.
- Approach your data the way a journalist approaches a source
If you’re visualizing COVID-19 case or deaths data, take the time to understand how the data was collected and its limitations. While a number of companies and media outlets have made ready-to-visualize case datasets publicly available, the case data is complex.
- Relatable context
Avoid presenting case counts without additional reference information about who is impacted and how. Supplement COVID-19 datasets with information and/or data points that are relatable and meaningful to your audience.
- Appropriate comparisons
If including comparisons between different states or countries, ensure the definitions of your measures are the same. If there is variance (for example, one state includes probable cases in the ‘confirmed case’ count and another does not), state those differences explicitly.
- Avoid calculations
Epidemiological math is hard. While they may seem straight-forward, calculations such as summary statistics and case fatality ratios require a more nuanced understanding in a pandemic. Leave it to your subject matter expert, or draw only from existing published measures from reputable sources like the WHO or CDC.
- Title clearly
State your main finding in the title. Use language your audience will understand, but avoid unnecessary jargon.
- Annotate, annotate, annotate
Explicitly emphasize key takeaways with words. Make the key points easier for your users to understand quickly.
- Include totals
Explicitly state totals related to your charted data. Good example of denominator in a caption: ‘This represents 7k cases with the required detailed data out of 122K total cases.’ Bad example: ‘This represents 7K cases’ (not enough detail).
- Include definitions
Provide definitions of specialist terms. Footnotes can be used to clarify technical economic and medical terms where necessary.
- Timestamp and sign your work
Include authorship as well as date and time of sourced data on the chart image. Be transparent about your affiliations, expertise, funding, and biases. With information being shared so rapidly, it’s important to be able to rapidly identify who visualized what, and when.
- Design mobile-first
Most people are checking COVID-19 data on their phones. Consider how your work will display on mobile, and design mobile-first where possible. If you will not support smaller device formats, then acknowledge this with a note.
- Prioritize accessibility
Many visualizations of COVID-19 are not accessible to people with visual impairments or cognitive disabilities. Use visualization tools that are compatible with screen readers or other assistive devices, and conduct accessibility checks before publishing. Consider how the full range of human diversity will be able to interpret and understand your visualizations.
- Color restraint
To emphasize key data points use high-saturation color. De-emphasize other features with low contrast colors, in accessible color palettes. Pay attention to local color meanings in the location where your visualization will be seen.
- State limits
Clearly state uncertainties in footnotes. This is particularly important when plotting correlated variables (which may imply causal relationships to some people) and when significant numbers of cases are missing data for more detailed analysis of outcomes.
- Track impact
Follow discussions about your visualization. Be prepared to adjust and/or update your visualization to mitigate any potentially harmful assumptions being made based on your chart or to clarify misinterpretations.
- Update manually
Don’t passively allow a script to update your numbers. Monitor incoming trends for any anomalies or outliers, follow updated documentation on your data source(s), and correct for harmful assumptions.
Balancing ‘Effective’ and ‘Humane’
These recommendations reflect principles of data humanism: to be ethical, a visualization must be effective and humane.
To be effective, visualizations must accommodate the limits of human sensing, cognition, and contextual capacity. To be humane, visualizations must employ tact about how the limits of those human qualities are stretched, honor the humanity of represented subjects, and acknowledge the potential harms that may be inflicted on people related to the visualized subjects (such as surviving relatives).
For example, policy conversations in many places have turned into tradeoffs of re-opening economies and the potential toll in human life. If visualizing data related to the economy and cases together, consider the story you are telling and how your language would feel to someone from a high-risk population.
Recognize that despite the perceived urgency of producing COVID-19 content, creating effective and humane visualizations about this topic takes time. Plan for multiple iterations and rounds of feedback from different people. While the publication of daily case counts may create a sense of urgency, the daily tracking needs for most countries are met with existing visualization tools from reputable sources.
We are still in the early stages of understanding this disease. We can celebrate those who have recovered but do not know what the long term morbidity associated with COVID-19 infections will be. Be mindful of the language you use to describe recovered cases.
What would you add?
The recommendations provided here is a first attempt to present a cohesive, actionable guide for visualizers using COVID-19 data. It’s based on professional and research experience, as well as reflections about the visualizations created in the first two months of this pandemic. If you have additions and corrections, please add any feedback in the comments below or reach out to the primary author directly.
Additional recommended reading & listening
For a deeper dive into ethical considerations of visualizing COVID-19 data, we recommend the resources below:
- The Ethics of Visualizing During a Pandemic , by Bridget Cogley— which includes a more comprehensive list of questions to consider if/when designing a COVID-19 visualization
- Why I’m not making COVID19 visualizations, and why you (probably) shouldn’t either , by Will Chase— which includes practical suggestions for alternative public outreach efforts, instead of COVID-19 visualizations
- Ten Considerations Before You Create Another Chart about COVID-19 by Amanda Makulec
- Ethics and What We Owe Each Other by Bridget Cogley
- Visualizing coronavirus data? Consider adding a disclaimer. by Amanda Makulec
- A conversation with an Epidemiologist: 5 things to keep in mind when you look at the numbers on COVID-19 by Amanda Makulec
- How to Slow the Spread of Misinformation by Anna Foard
The list of recommendations was framed using the work in the forthcoming conference paper ‘Make Me Care: Ethical Visualization in the Sciences and Data Sciences’ by K Hepworth, Ph.D., to be published by Springer in the HCII Conference Proceedings in July 2020. For more information and updates, see: https://kathep.github.io/ethics.
The framework was adapted with more specific recommendations for visualizing data related to COVID-19 by Amanda Makulec, MPH, drawing on her domain expertise in public health, related publications, and DVS discussions on COVID-19 visualizations.