Be Mindful of Misrepresenting Data in Visualization (2023)

It can be challenging to find creative and effective ways to present data to an audience. Some folks who participate in research and study data enjoy chewing on pages and tables full of numbers, but most general audiences do not. They are looking to glean and retain the most critical pieces of information as fast as they can.

When you’re choosing how to share your data in the most effective way possible, it’s vital that your summarized version still tells an accurate and complete story.

Data visualization provides a simple way to see and understand trends and patterns in data and can make it possible to analyze vast amounts of information at once. Since our eyes are drawn to colors and patterns, data visualization helps keep an audience focused on the message. It’s storytelling with a purpose.

With the freedom and flexibility that comes with communicating data through visuals, there are guidelines to follow, or you could end up causing more harm than good. Unintentionally—or even intentionally—misrepresenting or manipulating data jeopardizes the credibility of your visualization, and it is unethical to leave out pieces of data because it disproves your point or doesn’t jive with your narrative.

And yet, manipulation of data in visuals is a common practice.

What is Data Visualization

Data visualization is a hybrid form of art and science. Here, how the data is presented is as important as the data itself because an image invites the imagination to participate in the conversation. Typically in the form of charts, tables, graphs, maps, infographics, and dashboards, data visualization provides a way to organize numerical data—sometimes large amounts of complex statistical data—into digestible visual formats. This helps a viewer better understand information and to see patterns in data.

Some of the commonuses for data and information visualization are:

• Presentationsfor understanding: Some ideas are too awkward to express in words, buttranslate well in graphic formats. Things like maps or city plans.
• Explorativeanalysis: Using charts and graphs to uncover patterns that require furtherresearch.
• Confirmationanalysis: Helps to confirm an understanding of data and to explorerelationships and correlations further.

Visualizing big data can reveal hidden patterns and uncover lesser-known relationships to improve decision making. But with so much data coming from so many sources, it can be challenging to provide accurate, real-time visualization to explore big data. The nature of big data makes accuracy and completeness difficult to identify and track, which increases the risk of false discoveries, or skewed findings.

In the fields of science and mathematics,data can be hard to communicate to an audience that’s unfamiliar with thelanguage. Although historically very limited, graphics and images have becomeessential translators of scientific matters to the general public. The problemis that science communicators are typically nottrained in visual communication and unintentional errors andmisrepresentation can happen. Pair that with the knowledge that the tools forvisual storytelling change every day, it makes it more difficult forresearchers to keep up with thechallenges of scientific data visualization.

While it’s true that the numbers don’t lie, they can be usedto misleadand tell half-truths. Often, it’s assumed that companies or stakeholders intentionallymisuse statistics because they have something to gain from distorting the truth.Sometimes though, it’s design mistakes.

How Can Visuals Be Manipulated?

The global accessibility of data means it can be harder tofind the right information to tell a particular story, so a visual storytellerhas to be very careful where they get their data and how they share it.

Assuming you’ve used a reliable data source, some of the ways perfectly sound and useful data can be manipulated are by;

• Misrepresenting data: Beware of mislabeling data or using inaccurate descriptions for a visual. Or even, using the wrong chart. And when presenting data that shows a correlation, don’t imply the relationship. Don’t be afraid to add context because that may help the viewer understand what they’re seeing.

• Using too much data: Your target audience needs to be able to look at your graphic and find what they are looking for without clutter. When a graphic is too busy, it can be hard to find the main points. So keep it simple and narrow the focus of each graph.

• Distorting data: Again with using the wrong chart type, not charting to scale, omitting data which may hide trends while creating trends that don’t exist, and truncating the y-axis are all ways that data can become distorted. These distortions make it hard to compare data points.

Graphics created electronically carry more authority thanimages drawn on paper;people expect computers to be right where hand-drawn images are expected toleave room for a bit of error.

There has also been concern among the science industry thatcomputer-generated graphs and graphics could be tricking the public—and eventhe scientists who use them—into believing that theories or forecasts were provenfacts because of their finished presentation. Speakers at the Conference of theAmerican Association for the Advancement of Science notedin 1994 that in “an image-hungry world, a computer forecast of patterns ofair pollution was more effective in influencing policymakers and politiciansthan dry tables of numbers and charts.” This caused the speakers to worry aboutthe quality control of graphs or graphics, because the finished products wereprofessional enough to pass for the truth, without the same intense peer reviewas scientific papers.

Examples of Inaccurate Visuals

The ability to use graphic software to create visuals givesscientists a way to explore their data in methods that may have previously beencost-prohibitive. Now they can create graphics themselves! Cost-effective itmay be, but cutting out the trained graphic designer can also lower thestandards of their completed work.

On the flip side, a cartographer drawing a map uses their years of science experience and training to create the most accurate map they can. A graphic designer without cartography, archaeology, or GIS training creating a map may produce a better quality or cleaner version. Still, without that training, it’s more likely that the designer creates a beautiful but misleading map.

Spend any time on the internet and you will eventually comeacross an inaccurate graphic with potentially misrepresented data. Here are afew examples that have made the rounds.

1.Prison/Princeton

This graphic created by PublicAdministration.net made the viral circuit a few years ago but is incorrect. Believe it or not, the problem with this otherwise beautiful graphic is that the numbers for Princeton and prison are flipped. The annual tuition cost at Princeton in 2010 was $48,500, and the cost of incarceration in New Jersey in 2009 was $38,700. They’ve got it backward.

In this example, someone had a specific story they want to tell and manipulated the data to tell that story. “Gasp! We spend more on prisoners than an Ivy Leauge college!” was the viral response. If that were true, this would be the wrong chart to prove it.

2.Gun deaths in Florida

This Reuters chart is trying to show how the number of gun deaths in Florida has changed since the ‘Stand Your Ground’ law was enacted in 2005. The chart is actually upside down. The x-axis starts at 0 and goes down to 800, making it look like the number of gun-related deaths has declined when actually, it’s gone up.

Why might someone present a chart this way? In all probability, this was the graphic designer taking a bit of a creative license to allow for the dramatic red background and was not designed with the intent to deceive. But again, it might depend on the stakeholders behind the chart and what message they’re trying to send. If they’re supporting the law, they want it to look like the number of deaths has gone down—so they point the chart that direction.

3.Global cooling

It is not uncommon for stakeholders or organizations tocherry-pick data to support otherwise unsubstantiated claims.

The average temperature of our planet naturally varies alittle from year to year. Climate change naysayers take advantage of this realityby showing temperature data in small clusters of years instead of a larger span,like decades. Filter out the natural pulsing of the planets, and an overallrise in temperature is apparent.

This line graph, which tries to communicate that there isn’t a climate problem and that temperature just varies a bit, leaves out some critical information. Primarily, those decades of climate variances we talked about.

4.The truth about false accusation

Sarah Beaulieu’s viral infographic from 2012 was intended todrive home the point that in spite of what the media might suggest, false rapeaccusations are extremely rare. It does make that point; however, the rest ofthe graphic is inaccurate.

There is a different between “false reporting,” claiming an assaulthas happened; and “falsely accusing,” which is pointing out and identifying aperpetrator. It’s important not to confuse the two terms. Research shows that 2-8percent of reported rapes are false, but the number of accusations is even smaller.

This graphic assumes one rape per rapist, but at the time it was created, the average rapist had six victims. By using the label “rapists” instead of “rapes,” this chart might give the impression that every-other person you meet is a rapist when in reality, it’s an estimated 6 percent of men, repeating their assaults.

Also, the Rape, Abuse and Incest National Network (RAINN)estimates 3out of 4 sexual assaults go unreported. This chart overestimates thatnumber.

Why? Is it to point out how rare it is for an assault victim to report? Or is it to raise the conversation? Although it is correct that most sexual assaults go unreported and the chances of being falsely accused are small, the data in this graphic is telling an inaccurate story.

5.2017 NRL Draft

Sometimes a chart is incorrect because it’s just the wrongformat for the information provided. This chart from the NFL draft is a goodexample.

As a pie chart is used to show percentages, the sectionsshould add up to 100 percent. This percentage chart totals at 289. Based on thesize of the parts, we’re to believe that 60 is less than half of 69, and that USCis making a haul.

The thought is that the designer wanted to make all of thelogos visible in their section, but the chart grossly misrepresents the truth. It’sa great candidate for a bar chart.

6.The Big Cheese

Speaking of pie charts, this pizza version looks cool and meansto tell a story about the executives who’ve served as CEO at Papa John’s, but aviewer would never get that without reading every bit of text. The sections areunbalanced and this information would be much better presented as a timeline.Points for trying to be witty and topical, but it’s also not the right formatfor the information.

7.Summer Olympic Medal Count

This graphic clipped from an edition of the Erickson Times shows a deceiving chart withits iconography not being to scale. Sure, there are more medals to show the U.S.A.awarded the highest medal count, but the number of medals shown in no waycorresponds to the number of medals awarded.

We’re expected to believe that the two medals pictured forGermany represent their 499 medals. Then it instantly contradicts thatassumption with three medals pictured for France’s 523. And six medal icons represent1,975 medals for the U.S.A. There is very little relationship between the imagesand the actual number of awarded medals. So why bother?

8.Governor’s Race

And then there was that very exciting Governor’s race whenone candidate’s 37 percent was visibly larger than another candidate’s 37percent. There’s no reason to go into what the problem is with this chart.However, since we’re working with percentages and we add up to 100 percent (asis accurate), this might have been an ideal place for a pie chart.

9.Scotland gave 110%

Equally exciting was the incident when CNN gave us votingresults as a data visualization in the form of simple numbers…and they were wrong.

The graphic showing the results of the voting for Scotland’sindependence was 52 percent ‘no’ and 58 percent ‘yes,’ for a total of 110percent—a worrying result for a countrywide vote.

Not only was the data wrong, but it was wrong in the wrongdirection. The actual vote was 52 percent ‘no ’and 48 percent ‘yes.’

This example goes to show that it’s not just bar and piecharts and infographics that can get it wrong and present inaccurate data;sometimes it’s as simple as not checking your numbers.

Reminder about the ethics of using your power wisely

Seeing is believing, and people want to believewhat they see. Seeing data solidifies it in the mind’s eye and makes itmore real. Literate communication presented in tables or a line of text isboring, so regardless of how abstract or complex an immersive data presentationis, we are less skeptical of data presented in an immersive visual format. Datavisualizations are so persuasive because they are both literate and immersive formsof communication, and we like them. So the information you present in visualformats needs to be accurate.

As data sets become more complex, displays should become increasinglyinformative, illuminating relationships that would be inaccessible fromtables or summary statistics.

The multitude of ways data visualization can go off therails is not a good argument for not using them. They can make complex informationmore transparent and are easily accessible. They just need to be accurate.

So before creating or sharing a data graphic, take athorough look at it. Do a bit of mental math. Evaluate where the graphic—and eventhe data—came from. Is it from a reliable, well-regarded source? Are there anysources at all? Who are the stakeholders? Do yourself the courtesy of notsharing inaccurate data.

WorksCited

Allen, Elena A, et al. “DataVisualization in the Neurosciences: Overcoming the Curse of Dimensionality.” Neuron, vol. 74, no. 4, 24 May 2012, pp.603–608., doi:10.1016/j.neuron.2012.05.001.

Beaulieu, Sarah. “The Truth About False Accusation.” Sarah Beaulieu, 1 Dec. 2012, sarahbeaulieu.me/the-truth-about-false-accusation.

“Data Visualization Beginner’sGuide: a Definition, Examples, and Learning Resources.” Tableau Software, 2019, www.tableau.com/learn/articles/data-visualization.(Module 3)

Esteban, Chiqui. “A Quick Guideto Spotting Graphics That Lie.” NationalGeographic, 19 June 2015, www.nationalgeographic.com/news/2015/06/150619-data-points-five-ways-to-lie-with-charts/.(Module 7)

Estrada, Fabiola CristinaRodríguez, and Lloyd Spencer Davis. “Improving Visual Communication of ScienceThrough the Incorporation of Graphic Design Theories and Practices Into ScienceCommunication.” Science Communication,vol. 37, no. 1, 26 Dec. 2014, pp. 140–148., doi:10.1177/1075547014562914. (Module6)

“False Reporting Overview.” National Sexual Violence Resource Center,2012, www.nsvrc.org/publications/false-reporting-overview.

Gleick, Peter. “GlobalTemperature Anomalies from 20th Century Average 2007 to 2011 (Degrees C).” Forbes, 5 Feb. 2012, www.forbes.com/sites/petergleick/2012/02/05/global-warming-has-stopped-how-to-fool-people-using-cherry-picked-climate-data/#22d4aef42b5c.

Hepworth, Katherine. “Big DataVisualization.” Communication DesignQuarterly Review, vol. 4, no. 4, 27 Dec. 2017, pp. 7–19.,doi:10.1145/3071088.3071090.

“Information Visualization – A Brief Introduction.” The Interaction Design Foundation, 2019, www.interaction-design.org/literature/article/information-visualization-a-brief-introduction. (Module 3)

Janssen, M. P. “What Does YourData Tell You? How Transfusion Chain Data Can Support ManagerialDecision-Making.” ISBT Science Series,vol. 12, no. 1, 21 Nov. 2016, pp. 38–45., doi:10.1111/voxs.12323.

Kirsch, Noah. “The Inside StoryOf Papa John’s Toxic Culture.” Forbes,Forbes Magazine, 19 July 2018, www.forbes.com/sites/forbesdigitalcovers/2018/07/19/the-inside-story-of-papa-johns-toxic-culture/#3ed71a53019f.

Lebied, Mona. “MisleadingStatistics & Data – News Examples For Misuse of Statistics.” Datapine, 8 Aug. 2018, www.datapine.com/blog/misleading-statistics-and-data/.(Module 7)

Macaulay Millar, Thomas. “MeetThe Predators.” Yesmeansyesblog.wordpress.com,16 Nov. 2009, yesmeansyesblog.wordpress.com/2009/11/12/meet-the-predators/.

Marsh, Samantha. “When DataVisualizations Mislead (And How to Prevent It).” IDashboards, www.idashboards.com/blog/2019/02/27/when-data-visualizations-mislead-and-how-to-prevent-it/.

McArdle, Megan. “Ending theInfographic Plague.” The Atlantic,Atlantic Media Company, 24 Dec. 2011, www.theatlantic.com/business/archive/2011/12/ending-the-infographic-plague/250474/.

McCormack, David. “CNN Reports110% Turnout in Scottish Independence Vote.” Daily Mail Online, Associated Newspapers, 18 Sept. 2014, www.dailymail.co.uk/news/article-2761778/Something-doesn-t-add-CNN-Reports-110-turnout-Scottish-independence-vote.html.

Miller, Paul, and JulianRichards. “The Good, the Bad, and the Downright Misleading: ArchaeologicalAdoption of Computer Visualization.” ComputerApplications and Quantitative Methods in Archaeology, 1995, pp. 19–22.,proceedings.caaconference.org/files/1994/03_Miller_Richards_CAA_1994.pdf.

Reuters. “Gun Deaths in Florida.”LiveScience, 16 Feb. 2014, www.livescience.com/45083-misleading-gun-death-chart.html.

Robbins, Naomi. “MisleadingGraphs: Figures Not Drawn to Scale.” Forbes,Forbes Magazine, 16 Feb. 2012, www.forbes.com/sites/naomirobbins/2012/02/16/misleading-graphs-figures-not-drawn-to-scale/#49ccb9bd15ef.

Sloan, Christopher. “VisualStorytelling for Science.” ScienceVisualization, 2019, sciencevisualization.com/.(Module 6)

Yang, Chaowei, et al. “Big Dataand Cloud Computing: Innovation Opportunities and Challenges.” International Journal of Digital Earth,vol. 10, no. 1, 3 Nov. 2016, pp. 13–53., doi:10.1080/17538947.2016.1239771.

Zen, Pola. “Storytelling SecretsFor Creating Images That Connect.” Yotpo,27 Dec. 2017, www.yotpo.com/blog/5-visual-storytelling-secrets-to-improve-your-marketing-images/.(Module 5)