Estimated reading time: 6 minutes

Summary

Accessibility is not just about being inclusive; it is also about being transparent, for example, by citing the sources for your visualisation
Transparency is central to good research practice, and by being transparent when visualising your research you make it easier for your audience to evaluate your findings and process
In an academic context, graphic material must meet the same standards of source use as non-graphic material
Even in more popular contexts, such as in scientific animations intended for broader audiences, clarity and transparency in source use is ideal; it increases accessibility and establishes credibility

Accessible visualisation means more than adopting a norm-critical and inclusive perspective during the production of graphic material. It is also about transparency, that is, demonstrating through references what kind of data or information you have selected and based your visualisation on, as well as what has been excluded and why.

Academic contexts require source use transparency

Transparency is one of the basic principles of research ethics and good research practice, and by being transparent when visualising and communicating research data you will make it easier for your audience to evaluate your materials, processes and conclusions.

Certainly, any visualisation presented in an academic context falls under the same requirements regarding source use as non-visual material. For instance, when a protein structure is added to the Protein Data Bank (PDB), it is linked to the source article that presents those findings. An educational illustration published in the PDB's learning portal PDB101 likewise always has a list of sources that clarifies which structures, and therefore which articles, it is based on. In short, in any visualisation, the reference chain, from the published visualisation all the way back to the data, is ideally unbroken—or at the very least, maintained as much as is possible.

On that note, in your own visualisation work, you might want to give extra thought to descriptive legends, assistive texts and supplementary information. In the case of visualisations based on statistical data, you can provide a degree of transparency through supplementary tabular data (provided, of course, that this is possible). There are numerous ways to enhance the transparency of your data visualisation process. If your visualisation contains photographs that have been altered in any way (such as through cropping, colour adjustments, etc.), you could, for example, share non-processed full-size images openly as supplementary data, to make it possible to inspect and compare the processed and original versions. Or, if you have visualised your data with the help of statistical programming languages, for example R, you might share code to make it easier for others to reproduce and understand your work.

All scientific communication demands source use transparency

Even in more popular contexts, such as in scientific animations presented for broader audiences, clarity and transparency in source use is ideal; it increases accessibility and establishes credibility.

As Stuart G Jantzen, Jodie Jenkinson and Gaël McGill note in their 2015 commentary, "Transparency in film" (Nature Methods, 2015) audiences of in their case scientific animations "are often unable to determine the degree to which visualizations are informed by scientific evidence." If your intent is to educate the general public about your research, it needs to be clear how credible that research is, what sources or data it is based on, and so forth. Jantzen and his co-authors therefore recommend "providing a more detailed account of source use" so the evidence portrayed in the animation increases in credibility. This is, arguably, true for any visualisation of research data, no matter its context: to be credible, it has to be transparent and clear about its sources.

Transparency and the unknown/hidden

With all that said, however, we don't always have all the data we need at hand. In graphing, as in other parts of the scientific process, hypothesis-driven approximations occur from time to time. For example, it may be unclear how two proteins interact with each other, even though it is well-established that they do. Such uncertainties must be accounted for; it needs to be clear where there are gaps and what consequences those gaps might have for the conclusions being drawn in the visualisation. As early as 2007, David S. Goodsell and Graham T. Johnson addressed this problem in an article in PLoS Biology entitled "Filling in the gaps: artistic license in education and outreach." According to Goodwill and Johnson, the key to "filling in the gaps" effectively is to explain without distorting, even when a degree of uncertainty exists.

In other instances, it may not be necessary to show absolutely all the data. In graphical representations, as in any presentation of data, "noise" may need to be reduced and irrelevancies excluded in order to make the message clearer. However, in cases where what is excluded is necessary for understanding the context of the visualisation, then the exclusion in question needs to be accounted for.

For example, hydrogen atoms are excluded in a visual representation of protein and macromolecular structures retrieved from the PDB bio-database (mentioned earlier). This is because the exact position of the atoms is uncertain, a direct result of the need for the resolution limit to be 1Å or better, which is not always possible or appropriate. Therefore, in a given structure's associated information box, the total number of visualised atoms is indicated as number of modelled non-hydrogen atoms in the deposited model.

While not all scholarly journals provide detailed information on this aspect, it is relatively common for journals to include rules and best practices for image processing and manipulation in their submission guidelines and misconduct policies. In recent years, an increasing number of researchers have started to use generative artificial intelligence (AI) for data analysis and visualisation. As a result of the wide scale development and deployment of AI tools in society, many publishers have updated their guidelines to address concerns raised by the scientific community about AI-generated content, such as transparency, bias, integrity and responsibility. So, remember to check author instructions and software use disclosure policies early in your work process. For example, a publisher may require you to provide a detailed statement of what tools you used to create your visuals, including when, how, and why they were used. (Note: This paragraph was reviewed using the AI-powered writing assistant DeepL to improve its grammar.)

Transparency and the potential for misunderstanding

While visualisations can be powerful pedagogical tools, and are irreplaceable in certain types of communication, you can run the risk of creating a visualisation that inadvertently misleads the viewer. While a photographic image depicting an object provides a higher level of detail and accuracy, a schematic or illustration of that same object instead provides a higher level of artistic expression and design freedom, as it is an abstract simplification of reality. Furthermore, an image might look complete, but may actually be a part of a cell, organism or process.

However, this problem can be solved through clarity and transparency. Exploratory studies do this all the time, regularly reporting, describing and motivating both their selection and exclusion criteria. In your own visualisation work, you will want to consider whether you are deliberately excluding any graphic elements that are necessary for your visualisation to be properly understood. Then consider how likely it is that your intended audience has an adequate understanding and knowledge of the wider context, and whether the exclusion needs to be commented on.

As an example, let's consider how cytoplasm is often depicted in textbooks. The cell is often relatively empty; it may include a membrane, some organelles and a few floating proteins of various kinds, along with the nucleus in the center. For pedagogical reasons, much has been excluded and the scale distorted. Many people then proceed, throughout their educations, lives and so forth, to believe that this is what a cell looks like.

Contrast this with the image of the cytoplasm that appears in Sean R. McGuffee and Adrian H. Elcock's 2010 article Diffusion, crowding and protein stability in a dynamic molecular model of the bacterial cytoplasm,' where the authors present an animation showing macromolecules in the intracellular fluid of a coliform bacterium. We see a myriad of random molecules colliding in a fascinating jumble of Brownian motion. The cell, as we know, is anything but empty.

Makromolekyler i en kolibakteries intracellulärvätska.

Macromolecules in the intracellular fluid of a coliform bacterium.
Source: Cell Image Library, doi:10.7295/W9CIL28234. Creative Commons BY NC SA

An animation like Elcock and McGuffee's obviously serves a different educational purpose than the cytoplasm that appeared in textbooks, but the point is that visuals have a way of staying in people's minds. In your own work, therefore, be sure to comment on what has been excluded from your visualisation, especially if what has been excluded is important to fully understand the visualisation.

Recommended reading

Mike Rossner, Kenneth M. Yamada, "What's in a picture? The temptation of image manipulation". Journal of Cell Biology (2004).
Tracey L Weissgerber, Stacey J Winham, Ethan P Heinzen, Jelena S Milin-Lazovic, Oscar Garcia-Valencia, Zoran Bukumiric, Marko D Savic, Vesna D Garovic & Natasa M Milic, "Reveal, Don't Conceal: Transforming Data Visualization to Improve Transparency". Circulation (2019).