Data Visualization’s Breakthrough Moment in the COVID-19 Crisis
The discipline has made some important contributions over its 300-year history, but perhaps none more so than now
During the COVID-19 crisis, data visualization researchers and professionals rose to the challenge, delivering widely used tools for public explanations, pandemic modeling, and government policy-making. These interactive data visualizations inform the public and guide decision-makers to save lives. Spirited debates center on “flattening the curve,” which is a clear reference to visual representation of the rise and fall of case numbers.
Johns Hopkins University engineering professor Lauren Gardner and her team built a prominent visualization dashboard that shows current worldwide country data, down to the county or province level. Its evolving multiple windows show a great deal of data by way of clickable tabs for those who want more information. Newspapers, TV news programs, and bloggers feature this dashboard, giving it even wider usage.
The dramatic rise of data visualization could be traced to hardware factors such as widespread use of high-resolution large desktop displays tied to powerful computers. Other important trends are the increased availability of vast data resources, familiarity with data management software, and innovative web-based software that support rapid display and update of visual information.
Steven Drucker (Microsoft Research) says “the COVID crisis has generated a real need to understand and communicate vital information about data, models, and outcomes. We’ve needed it to persuade, understand current conditions, and predict future outcomes based on behaviors. I don’t think there’s ever been a moment where data, models, and hence visualization has been thrust so much into the center of everyday life.”
The 300-year-old strategy of drawing charts is usually traced back to William Playfair’s 1786 book with economic analyses of world trade patterns. An 1854 cholera epidemic in London provoked Dr. John Snow to draw a map showing deaths in the Broad Street area, revealing a concentration around the local water pump. When the pump handle was removed, deaths dropped, supporting the idea that cholera was transmitted by sewage-contaminated water. In 1869, Joseph Minard’s anti-war map powerfully showed how Napoleon’s army died off during its tragic winter attack on Moscow. These precedents set the stage for the current explosion of interest.
Data visualization blossomed during the past 30 years as an academic highlight with conferences, journals, and textbooks filled with innovative designs, theories, and evaluation studies. Many of the currently popular visualizations follow the well-established Information Visualization Mantra: “Overview first, zoom and filter, then details-on-demand.” This guideline suggests that designers begin by showing an overview, such as a world map, and allow users to zoom in on what they want, and filter out what they don’t. Then they can select an item, such as a city, to get just the details that interest them, such as a city’s COVID-19 case, death, or recovered numbers. Users might filter by age range, ethnic groups, or severity. These easy-to-use designs were widely applied, laying the foundation for the current strong interest.
The early academic research on data visualization triggered a growing number of commercial successes, which were put to work by business analysts, scientific investigators, and by tech-savvy journalists to investigate and present complex stories.
The New York Times was an early adopter of interactive visualizations and continues its tradition of excellence with daily updated COVID-19 data and compelling scrolling stories on specific issues such as disease exposure for different jobs. The Washington Post, The Guardian, and many other publications also had original and revealing data visualizations showing animations of disease spread and how different countries were flattening their curves.
Government agencies such as the U.S. Centers for Disease Control and Prevention show the disproportionate disease rates across ethnic or age groups and Canada’s Public Health Services shows graphics for the number of cases by date. Many U.S. states produce popular charts and maps collected into a dashboard, such as this one from Maryland including county level dashboards, such as this one for Prince Georges County, outside of Washington, DC.
Corporate visualization leaders also stepped forward to provide COVID-19 insights, while featuring their products, such as on the websites from Tableau and Spotfire. Other companies show how their customers put their tools to work during the COVID-19 crisis, such as Weather.com’s use of IBM’s Cognos or Narrative Science’s use of Microsoft’s Power BI. Even small visualization companies are finding distinctive angles, like Visual Action’s recovery rates by country.
Geographic maps were produced by ESRI using its geographic information systems tools and also by its customers, such as this dashboard made by the British Columbia, Canada Center for Disease Control.
University data visualization groups, such as the University of Maryland’s, made guides to the numerous visualizations, while our colleagues put their skills to work, producing specialized visualizations for topics such as data on social distancing indexes, percent staying home, and work trips for each state.
The Data Visualization Society seeks to promote awareness of the topic, especially with its free online publication Nightingale. That name refers to Crimean War nurse Florence Nightingale, who used charts and graphs to dramatically show that more soldiers died from disease than from fighting. Executive Director Elijah Meeks wrote about how data visualization hit the mainstream in 2019. A recent Nightingale article described how the Financial Times staff developed new insights from visualization, which showed the rate of growth of COVID-19 in different countries.
These widely used COVID-19 visualizations depend on maps, bar charts, line charts, and scrolling lists, however designing them to fit on small mobile-device displays requires skillful programming. Some websites are overwhelmed by the demand, which results in very slow service, or worse, crashes. Many websites need to be tested in different browsers to ensure that all the features work. Many websites could be improved by usability testing to guide designers to needed fixes and refinements, while helping to ensure accessibility by users with disabilities. Just as the U.S. Centers for Disease Control and Prevention provides guidance in as many as twenty languages, website designers should try to allow users to select from several languages.
An enduring issue for any statistical analysis or visualization design is uncertainty in the data, especially when showing predictions of future cases or deaths. Good design and professional practice is to show the sources of data, name the contributors to the visualization, and provide an email address for comments, corrections, and questions.
Many websites could improve their handling of missing data, such as cities or counties that skip reporting on certain days, and changes to metrics such as when the rules of confirming COVID-19 cases change. Comparisons across cities, counties, states, provinces, and countries needs to account for different counting strategies and availability of testing. Such comparisons are improved when designers shift from absolute case numbers to cases per 100,000 people.
Information designer Paul Kahn and PhD student Janice Zhang at Northeastern University are adding daily updates to an open source database with 600-plus COVID-19 visualizations. Zhang is developing a coding scheme that captures metadata such as publisher type, date, language, and country of origin. She is recording the intended message, types of charts (maps, line charts, scattergrams, etc.), data sources used, the ways they visualize uncertainty, and many other features.
The complexity and importance of COVID-19 has put data visualization center stage in worldwide discussions. The free public websites have boosted data visualization literacy, which could lead to more people using interactive tools to explore data for many applications and then presenting their results to more receptive visually literate audiences.
Ben Shneiderman is a Distinguished University Professor Emeritus in Computer Science at the University of Maryland. He is a member of the U.S. National Academy of Engineering.