BOF: On Standards for Accessible and Interpretable Visualizations

[atuldeshpande]

Hello, I would like to propose a BOF on the following topic:

Input/Topic: Color is becoming an increasingly powerful data visualization tool to represent the dimensionality of our complex and rich scientific data. However, these visualizations and the patterns therein are also becoming increasingly difficult to interpret for individuals with color vision deficiency in particular, and other vision impairments in general. Color vision deficiency affects a significant percentage of the population, and leads to difficulties in perceiving patterns in multi-colored figures. In many cases, the perceived patterns; e.g. in heatmaps and reduced dimension plots, could differ between individuals with normal and color deficient vision.

Figure: An Ishihara's test where people with normal vision see the number 74, but people with color deficiency see either no pattern, or 21/81. (From Wikipedia(https://en.wikipedia.org/wiki/Ishihara_test) and an unscientific survey on slack).

Visualization tools currently provide accessibility to colorblind developers and users through the option of choosing colorblind-friendly palettes. However, if the default visualizations in our R packages are not color-blind friendly, a vast amount of our published results risk being inaccessible to individuals with color vision deficiency. One solution to address these issues is to include colorblind friendly strategies (Color Blindness: A Point of View) as a default setting in our visualizations. Several colorblind-friendly palettes exist (e.g., see R packages viridis and dittoSeq) and can be integrated into data presentation as the default option. Even with these palettes in place, it is desirable to limit the number (about 8-10 at a maximum) of colors in visualizations. To reduce the dependence on colors, one solution would be to include additional visual cues to differentiate regions (hatched areas) or cells (point shapes). Overall, a broader discussion regarding the accessibility of our figures that is not just limited to color vision deficiencies would be greatly beneficial towards improving data accessibility.

Note: Some of the thoughts above have also been noted in the whitepaper here: https://birsbiointegration.github.io/whitePaper/

Output: We would like to identify and recommend strategies for making standard visualizations provided by Bioconductor packages more accessible to people with vision impairments, including, but not limited to:

• Incorporating colorblind-friendly color palettes as a standard recommendation/requirement in the R package submission guidelines.
• Discussing effective visualization techniques which could reduce the dependence on color as the explanatory factor in visualizations (e.g, include point shapes, hatched areas, limiting the number of colors, etc.)
• Harnessing interactive plots to enable the choice of color palettes and other visualization tweaks to the reader.
• Addressing broader accessibility with the inclusion of "accessibility captions" accompanying figures which "guide" the reader's perception of the images, and how they could be incorporated as part of standard visualization tools.

[atuldeshpande]

Here is a link from the Department of Interior with recommendations on how to add descriptive, screen-reader friendly Alt-text to figures in PDFs: https://www.doi.gov/ocio/section508/video2

[csoneson]

👋 @atuldeshpande - your BioC2020 BoF session has been scheduled for Mon, July 27, 12-12:55pm EDT. A Zoom meeting has been created and will be accessible from within the conference platform (please make sure that you are registered for the conference in order to get access to the platform). There will also be a person from the organization team present during your session to help in case of any technical issues. Don't hesitate to let us know if you have questions or need additional support for your BoF. Thanks for contributing to the conference!