5  Ethical Data Visualization

Creating a high quality data visualization requires more than picking the right plot for your data and your research question. Certainly following visual design principles and utilizing design tools is a part of creating high quality visualization. However, further best practices are needed to ethically present complex data with clarity in a way that is accessible to audiences. This chapter will discuss ways to enhance the clarity and accessibility of data visualizations as well as important ethical considerations.

5.1 Learning Objectives

This chapter will demonstrate how to apply visual design principles to minimize complexity and improve the clarity of a data visualization, identify steps to enhance the clarity, accessibility, and understandability of a data visualization, employ ethical practices in constructing and sharing data visualizations, and how to evaluate whether a data visualization can be improved to enhance its clarity, accessibility, or accuracy.

5.2 Best Practices for Accessibility

5.2.1 What does accessibility have to do with data visualization?

Both accessible and ethical data visualization improve viewers’ understanding rather than confuse or mislead. When sharing a visualization widely, there are several audience types you may encounter.

  1. Audiences unfamiliar with your research.
  2. Audiences with visual impairments.
  • Low vision or blindness
  • Color vision deficiency
  1. Audiences with dyslexia or who are prone to distraction.
  2. Audiences with motor impairments.

The majority of this resource will focus on making visualizations more accessible for audiences who are unfamiliar with your research or who have visual impairments. Accessibility considerations for audiences with motor impairments mostly impact how to prepare interactive visualizations.

Resources for ensuring interactive visualizations meet accessibility standards include the WCAG Understanding Documents that provide guides for understanding and implementing the World Wide Web Consortium (W3C) Web Accessibility Initiative (WAI) Web Content Accessibility Guidelines (WCAG) (Participants, n.d.b). An example of a relevant guideline is the Keyboard Accessible Guideline (Guidelines 2.1) (Participants, n.d.a).

The University of Washington, as part of their Digital Accessibility Guide, provides a guide related to ensuring keyboard accessibility that references the relevant WCAG Guidelines as well. (Washington 2021)

Harvard University also offers guidance for producing accessible interactive visualizations (Technology, n.d., ​).

5.2.2 Audiences unfamiliar with your research

The rules for increasing the accessibility of your visualizations for audiences unfamiliar with your research are united by the principle that you know your research better than your audience will know it. Therefore, you need to guide the audience through the visualization, making sure that the stand alone plot can communicate whatever you would show or tell if you were presenting it to them.

Rule 1: Provide descriptive, understandable, and visible text.

Rule 2: Reduce the complexity of visualizations and be intentional in how you share them.

Rule 3: Don’t over clutter plots.

Rule 4: Make comparisons easier.

Rule 5: Use a layout that follows typical reading order and aligns with expectations.

5.2.2.1 Providing descriptive, understandable, and visible text

  • Use a large, visible text size
  • Consider bolding specific words for emphasis
  • Avoid specialized language such as jargon, acronyms, and abbreviations. If you must use them, define acronyms and abbreviations in your caption.
  • Use a declarative and descriptive title that communicates the main takeaway.
  • Include a detailed caption that answers the following questions:
    • What is the data?
    • Where did the data come from?
    • What special wrangling, transforming, filtering, etc. was done to the data?
    • What’s being shown?
    • What do the colors or shapes stand for?
    • What is the overall trend or pattern?
  • Include axis labels that are accurate but not too detailed.
    • Especially avoid jargon and undefined acronyms in axis labels.

5.2.2.2 Clearly presenting complex data

An overly complex plot can be misleading for a number of reasons. The main takeaway could be obscured because too much data is present and therefore the main takeaway is non-obvious, hidden, or confused with another pattern that is there. In addition, unnecessary or overly distracting visual design elements could unintentionally be drawing away attention from the main point to another area of the plot. The solution to this issue is to use your visualization to answer only one question at a time.

Accessibility can be increased by decreasing the complexity of a plot, especially decreasing the complexity of the message. If a multi-line title is needed to explain every possible conclusion, there are too many takeaways in the plot.

Sometimes visualizations are showing enough data that multiple conclusions or hypotheses for further explanation can be made. This happens especially when the visualization is showing connections or relationships among multiple aspects of the data. Frequently survey data may be visualized within context, showing how certain populations or demographics answered questions differently. In these cases, several additional conclusions may be possible from close inspection of a visualization. Another example of this is a Circos plot used in cancer research to display genome wide variation (Krzywinski et al. 2009; Chia et al. 2016). Each ring within the circle displays a different piece of information (such as copy number variation or intra- and inter-chromosomal rearrangements/structural variants) for aligned regions of the genome. This is an especially useful plot following catastrophic genomic breaking and error prone reassembly due to chromosome instability in cancer (e.g., Chromothripsis)(Simovic-Lorenz and Ernst 2025).

In these cases where a visualization could suggest multiple conclusions, be intentional in how the complex plot is shared.

  • Use the title to focus the takeaway message.
  • Use the caption and any accompanying narrative to describe exactly where the audience should be looking.
  • Label different aspects such as facet labels, the rings of a Circos plot, axis labels, a few specific data points.

Strongly consider using a less complex plot or a series of less complex plots. Make one or a set of a few and show them to team members and ask which is more accessible while still conveying the point.

If you are sharing a complex visualization in a talk be sure to set aside enough time to walk people through the figure. If there are elements in the visualization that you do not feel need to be mentioned, perhaps they can be removed. Consider having at least one slide beforehand that orients people to what the figure will look like. Using toy data (e.g., simulated, skewed, unrealistic, etc.) explain what different patterns would look like and mean – specifically what conclusions could be made. Once displaying the visualization with your real data, considering using animation to gradually display parts of the visualization. This can be done in a variety of ways such as by utilizing cropping; manually adding shapes that are filled with the background color to cover data and removing them as you proceed through describing the visualization; manually adding helpful annotations on slides; potentially building different aspects of the plot separately and combining them together to form the whole visualization; or building different variations of a visualization depending on what specifically needs focused on or annotated.

If you are sharing a complex visualization in a report or paper be sure to use descriptive titles and captions that describe the main message and what pattern of interest to look for. Consider using multiple panels to showcase more focused aspects individually, supplementing the more complex figure. Reference and describe the plot in the accompanying narrative.

If you are sharing a complex visualization on a poster be sure to make the plot large so that your audience can read it, seeing details and distinguishing key elements. Also be sure to include the narrative or “elevator pitch” you would use to walk people through the visualization visible in case they are viewing the poster without you. In general it’s good practice to streamline the text on a poster, however including descriptions and explanations for complex plots that you can’t simplify is a good idea.

Finally, use visual design elements to reduce complexity. It may seem like adding these elements is adding complexity because it’s adding clutter or ink. However, certain elements are worth the addition because they can increase clarity. Use annotations or labels such as y- or x- intercept lines. Use facets or subplots. Though sometimes splitting a plot into multiple separate plots will be more useful than increasing the number of facets in a single plot. Notice that annotations, labels, and facets may be some of the elements that you gradually reveal or “animate in” during a talk. Finally, ensure clear labeling and names to reduce the complexity of a plot. Make sure the text is readable (large) and understandable.

In the same way that undefined specialized language may not be understandable, it can be misleading. Acronyms and abbreviations can be polysemantic and ambiguous with specific abbreviations having more than one possible meaning (Barnett and Doubleday 2020). Given the context or larger paper, the intended meaning will likely be conveyed, but data visualizations should be able to stand alone. And as such, using undefined specialized language can be misleading to a viewer.

Examples of acronyms with multiple meanings

MS stands for

SD stands for

RA has at least 8 different clinical meanings (Pakhomov, Pedersen, and Chute 2005).

5.2.2.3 Reducing clutter in plots

To reduce the clutter in plots consider the following things:

  • Use facets or subplots
  • Increase the transparency for overlapping data
  • Make several related plots

While it is necessary to not over clutter plots, it is also necessary to minimize unused space. It’s important to strike a balance between these two considerations. Some iterative polishing and team review may be necessary to find a proper balance. This is a case where data visualization is a bit more of a subjective art form.

5.2.2.4 Making comparisons easier

To make comparisons easier for the audience consider the following steps:

  • Reduce unnecessary or unused space. For example, a bar plot with too much space in between the bars makes it more difficult to compare the bars.
  • Outline bordering geometric marks to add a visual distinction between pieces.
  • Order categories based on frequency. For bar parts especially, or any plot where you want to show a ranking, show the categories ordered according to the ranking or frequency.
  • Label with the amount of represented data or include an accompanying table with this information.

5.2.2.5 Arranging data to tell a story

Use a layout that follows typical reading order. Progress through the story of a plot or diagram from left to right and top to bottom. If your visualization is supposed to be showing an increase it should visually look like an increase. If your visualization is supposed to show a decrease, it should look like a decrease. The “highest” point shouldn’t be at the bottom of the plot. If showing a ranking, the highest rank should appear one of the edges with the lowest rank at the other edge.

5.2.3 Rules for increasing accessibility for audience members who have visual impairments

The rules for increasing accessibility for audience members who have visual impairments are all united by the principle that plot elements should be distinguishable.

Rule 1: Use color vision deficiency friendly color palettes.

Rule 2: If using your own color palette, use a package or a simulator to check the accessibility of the color palette.

Rule 3: Don’t use color alone to represent groups.

Rule 4: Use a proper contrast ratio.

Rule 5: Don’t over clutter plots (also Rule 3 for audiences unfamiliar with your research)

Rule 6: Use a sans-serif font.

5.2.3.1 Using accessible color palettes

The most well known example of a color vision deficiency friendly color palette is viridis.

5.2.3.2 Checking accessibility for color palettes

colorblindr is an R package that simulates color vision deficiency directly on a plot in R, returning what it might look like for different types of color vision deficiency. You can directly examine your visualization to make sure that plot elements and groups are distinguishable and make changes to reassess as needed.

5.2.3.3 Using shape, patterns, or labels in addition to color.

An example of this technique can be found in recent editions of Uno card decks that include geometric shapes on the cards to communicate what group or color that card is.

  • Red: Circle
  • Green: Triangle
  • Blue: Square
  • Yellow: Star

Using color and shape together allows the audience to distinguish groups no matter what the color palette is. Uno cards use this method, including geometric shapes on cards to redundantly communicate groups. To those without color vision deficiencies, this appears redundant. However, for those with color vision deficiencies, anyone can distinguish the groups using the geometric shapes. Red has circles, green has triangles, blue has squares, and yellow has stars.

By redundantly providing a shape as well as the color, groups can still be distinguished because of the shapes.

The shapes are still distinguishable without color, ensuring accessibility and a way to tell the categories apart.

5.2.3.4 Checking accessibility for contrast

5.2.3.5 Using sans-serif fonts

Common sans-serif fonts include Verdana, Arial, and Calibri.

5.2.3.6 Descriptive Alternative Text

5.3 Best Practices for Ethics

5.3.1 What does ethics have to do with data visualization?

Ethical visualizations improve viewers’ understanding rather than confuse or mislead. Misleading messages are not always intentional on the part of the researcher putting together a data visualization, and mistakes happen (e.g., bugs in software, errors in data entry, etc.). It’s the responsibility of the researcher to carefully evaluate all data visualizations before sharing them to make sure that the visualizations clearly and accurately present the data and communicate the overall goal.

Several of the topics we’ll provide within this section about ethical considerations are related to tips to reduce complexity and enhance accessibility. This is not surprising given that clear communication is a necessity for both ethical and accessible visualizations.

There are 5 main categories for practicing ethical data visualization.

  1. Championing Safety & Transparency
  2. Ensuring Accuracy
  3. Preventing Data Distortion
  4. Maintaining Originality
  5. Practicing Empathy

5.3.2 Ethical Data Inclusion: Safety & Transparency

The rules within this section are aimed to safeguard patient privacy and provide transparency.

Rule 1: Don’t include identifiable data

ADD thoughts here

Rule 2: Include all relevant data or subgroups

Pie charts may incorrectly portray or give the impression that all data is included in the visualization.

Rule 3: Describe removed samples or dropouts

Captions are the primary place for this. Perhaps a title could very explicitly describe which groups are being visualized. The accompanying narrative should also include this information. However, because data visualizations should be able to stand alone, be sure to include this information with the visualization as well.

Some plot types, especially diagrams showing a process may provide a way to show which categories of samples are removed or how many patients dropped out (e.g., Sankey Diagrams).

5.3.3 Ensuring Accuracy

The rules within this section are aimed to encourage careful verification of visualizations.

Rule 1: Verify that visualizations accurately represent the data

ADD thoughts about data that may be missing, or sizes/area that don’t align with data. Do visualizations show a consistent message if you check visualizations against each other? Repeat the idea of checking twice before sharing the visualization

Rule 2: Use clear and accurate axis labels

ADD thoughts especially related to axis labels. Often having simplified axis labels help audiences. Doing this ethically such that the simplified label is what is being plotted is really important. If the label doesn’t match the data, why not plot the data for the simplified label?

5.3.4 Preventing Data Distortion

The aims for rules within this section include to avoid misleading messages and to compare groups fairly.

Rule 1: Use fair axis limits rather than truncating axes

ADD about starting axis limits at 0 because otherwise differences will look starker than they are

ADD about axis limits when you have facets and how audiences may expect those subplots to use the same axis limits

Rule 2: Order axes in a logical way that corresponds to the plot’s message

ADD info about the overall message. If a line is going up over time, but the message is related to decrease, why is the line going up? This can be misleading.

Rule 3: Separate overlapping plot elements

ADD info about how true info can be obscured if plot elements aren’t separated.

5.3.5 Maintaining Originality

ADD info about plagiarism and what that means for data visualization

ADD info about inspiration

Rule 1: Don’t take someone’s visualization (with their data) without attribution.

Rule 2: Use your own data in making a similar visualization, modifying where you can.

Rule 3: If using a type of plot and are in doubt, attribute.

ADD info about attribution if you’re concerned.

5.3.6 Practicing Empathy

Tips for showing empathy while constructing data visualizations based on suggestions from the Urban Institute. These are more tips and not rules because they are more subjective.

Tip 1: Visualize each data point

Point to jitter

Tip 2: Avoid stereotypical colors

Tip 3: Use person-first language when appropriate and avoid insensitive language

Tip 4: Expand an “Other” category

5.4 Improving Visualizations

5.5 Summary

Barnett, Adrian, and Zoe Doubleday. 2020. “The Growth of Acronyms in the Scientific Literature.” Edited by Peter Rodgers. eLife 9: e60080. https://doi.org/10.7554/eLife.60080.
Chia, Puey Ling, Craig Gedye, Paul C. Boutros, Paul Wheatley-Price, and Thomas John. 2016. “Current and Evolving Methods to Visualize Biological Data in Cancer Research.” JNCI: Journal of the National Cancer Institute 108 (8): djw031. https://doi.org/10.1093/jnci/djw031.
Ford, Kaitlyn, Elena Zuin, Dario Righelli, Elizabeth Medina, Hannah Schoch, Kristan Singletary, Christine Muheim, et al. 2024. “A Global Transcriptional Atlas of the Effect of Acute Sleep Deprivation in the Mouse Frontal Cortex.” iScience 27 (9). https://doi.org/10.1016/j.isci.2024.110752.
Gingell, Luke F, and Janna R McLean. 2020. “A Protamine Knockdown Mimics the Function of Sd in Drosophila Melanogaster.” G3 Genes|Genomes|Genetics 10 (6): 21112115. https://doi.org/10.1534/g3.120.401307.
Krzywinski, Martin, Jacqueline Schein, nan Birol, Joseph Connors, Randy Gascoyne, Doug Horsman, Steven J. Jones, and Marco A. Marra. 2009. “Circos: An Information Aesthetic for Comparative Genomics.” Genome Research 19 (9): 16391645. https://doi.org/10.1101/gr.092759.109.
Napoli, Francesca, Ciro Vella, Luca Ferri, Marco B. Ancona, Barbara Bellini, Filippo Russo, Eustachio Agricola, Antonio Esposito, and Matteo Montorfano. 2024. “Rheumatic and Degenerative Mitral Stenosis: From an Iconic Clinical Case to the Literature Review.” Journal of Cardiovascular Development and Disease 11 (5): 153. https://doi.org/10.3390/jcdd11050153.
Neurological Disorders, National Institute of, and Stroke. n.d. “Multiple Sclerosis (MS).” https://www.ninds.nih.gov/health-information/disorders/multiple-sclerosis-ms.
Pakhomov, Serguei, Ted Pedersen, and Christopher G. Chute. 2005. “Abbreviation and Acronym Disambiguation in Clinical Discourse.” AMIA Annual Symposium Proceedings 2005: 589593. https://pmc.ncbi.nlm.nih.gov/articles/PMC1560669/.
Participants, Accessibility Guidelines Working Group. n.d.a. “Understanding Guideline 2.1 Keyboard Accessible.” https://www.w3.org/WAI/WCAG22/Understanding/keyboard-accessible.html.
———. n.d.b. “WCAG 2.2 Understanding Docs.” https://www.w3.org/WAI/WCAG22/Understanding/.
Rajan, V., K. Beharry, H. D. Modanlou, J. Parker, Y. Akmal, and J. V. Aranda. 1998. “Effect of Morphine Sulfate (MS) on Endothelin Production and Receptor Responses in Newborn Piglet Brain Endothelial Cells 349.” Pediatric Research 43 (4): 6262. https://doi.org/10.1203/00006450-199804001-00370.
Simovic-Lorenz, Milena, and Aurlie Ernst. 2025. “Chromothripsis in Cancer.” Nature Reviews Cancer 25 (2): 7992. https://doi.org/10.1038/s41568-024-00769-5.
Son, Ahrum, Woojin Kim, Jongham Park, Yongho Park, Wonseok Lee, Sangwoon Lee, and Hyunsoo Kim. 2024. “Mass Spectrometry Advancements and Applications for Biomarker Discovery, Diagnostic Innovations, and Personalized Medicine.” International Journal of Molecular Sciences 25 (18): 9880. https://doi.org/10.3390/ijms25189880.
Swords, Christina M., Jerlym S. Porter, Amy J. Hawkins, Edwin Li, Melissa Rowland-Goldsmith, Matthew D. Koci, John T. Tansey, and Nicole C. Woitowich. 2023. “Science Communication Training Imparts Confidence and Influences Public Engagement Activity.” Journal of Microbiology & Biology Education 24 (2): e00037–23. https://doi.org/10.1128/jmbe.00037-23.
Technology, Digital Accessibility Services Harvard University Information. n.d. “Data Visualizations, Charts, and Graphs.” https://accessibility.huit.harvard.edu/data-viz-charts-graphs.
Washington, University of. 2021. “Keyboard Accessibility.” Accessible Technology. https://www.washington.edu/accesstech/guide/keyboard/.