Data Visualisation is Very Important and you **should** care about it: Part 2

| 07.30.2024

Part 2: Unforeseen circumstances and other complicating factors

In Part 1 I talked about Data Visualisation (DataViz), what it "is" and what it is meant to achieve. However, the picture I presented (he he he, pun not intended, but welcome) shows a rather simplified view of the matter. Data is not merely presented and observed in a vacuum. Your audience will see the data through some sort of medium and that medium may distort the message in ways you had not predicted, or they may see it differently to you because the way they see things is different to the way you see things. For example, take the famous Monty Python sketch about the Spanish Inquisition (bet you weren't expecting THAT to show up here! Ho ho ho ho.)

About a quarter of the way into the video the following exchange occurs (t=166)

Ximinez (aka Palin): Fear, surprise, and a most ruthless-- Ooooh! Now, Cardinal -- the rack!

Ximinez: You....Right! Tie her down.

Written down, like this, the joke is meaningless, but if all you had to read was a transcript that's all you might see. If you are watching subtitles in another language the whole thing may be even more confusing because your language may not include the word play that "rack" is both a fearsome torture device and shorthand for something used to dry dishes in the kitchen. The Monty Python sketch, however, is intended to be enjoyed visually in its original language and I'm sure the authors would welcome translation into other languages and media but ultimately these aspects were not considered when it was originally created.

In engineering, however, the data we work with is international and multi-cultural to a degree that entertainment does not typically have to abide by and we do need to consider these aspects when presenting data. In this part I want to talk about the many complications that come with presenting data.

Complication # 1 - Colorblindness/Color Vision Deficiency

Colorblindness, or "Colour Vision Deficiency" (CVD) as it is generally called these days, is something many individuals in the science and engineering industries seem to not be aware of, or don't think about as much as I think they should (the DataViz and UX fields by contrast are hyper-aware of it).

"Colour vision deficiency (colour blindness) is where you see colours differently to most people, and have difficulty telling colours apart. There's no treatment for colour vision deficiency that runs in families, but people usually adapt to living with it."

Inevitably, this will change how people view the "same" plot. OK, well how many people are affected by CVD then? Interestingly, CVD disproportionately affects men compared to women. 1 in 12 (~8%) men are CVD compared to 1 in 200 (~0.5%) women. Considering STEM is a male-dominated industry (Women in Tech: Why is there a small amount of women in STEM?) we should be considering it even more given our audiences are more likely to be men than women.

CVD typically affects people by reducing their sensitivity to a particular color, or colors. There are three prominent types of CVD and one extremely, super, rare type: full monochromacy (achromatopsia). Monochromacy means the perceiver can only see a single color and it is very, very, rare affecting approximately 1 in 33000 people. See https://achromatopsia.org/ for more information. Otherwise, the three main types are below.

ansys car cvd

Deuteranopia

  • This is a reduced sensitivity to green light.
  • This is the most common form of CVD

Protanopia

  • This is a reduced sensitivity to red light.

Tritanopia

  • This is a reduced sensitivity to blue light.
  • Barring monochromacy this is the rarest form of CVD

Looking at the figure above you can see how drastically the image is altered depending on the type of CVD a person has. In particular, it is worth thinking about what the images might be communicating relative to the standard image. Are the peaks and troughs very clearly in the same place? If red is no longer "red", then are people's eyes still drawn to it in the way you might want them to be?

There are many sites out there which you can use to see how images will look to people with these CVD types. Check out the section at the bottom of this article for various resources on CVD and where to go for more information. In terms of DataViz you will never be able to escape CVD completely and when communicating data it is important to keep it at the back of your mind at all times.

Complication # 2 - Format: presentations, printouts, papers, posters

The second complication to DataViz is the medium through which the data is presented. For example, if you are creating a plot for a PowerPoint presentation at a conference, it is very likely that you are going to use a projector and have to display this plot to a room of people sitting at a variety of distances from the screen in unknown lighting. In addition, each person should be able to see the plot the same way.

If you are creating the plot for a publication in a journal you may have limited page space to work with, ensuring that whatever you create can not be too big and must be quite dense. There's an added possibility to consider as well: 'will your plot get printed out in black and white?'. Some journals still print in black and white so it's not unheard of.

Perhaps your plot is in 3D but the audience won't be able to interact with it. Have you chosen the best angle to display the data you need or perhaps you realise you need to show off multiple angles to accommodate. Does your 3D plot have shadows from rendering that might interfere with the color itself? These are all things you might have to consider and should think about when creating DataViz.

Complication # 3 - "Great! Now, make it 'pop'." - Client

The last complication is much more of a circumstantial issue but it arguably has the largest effect on the final output. Often our clients/supervisors/managers step in and request things for unknowable and unforseeable reasons, but they want the things and they do pay us after all, so we have to make sure our visualisations include their suggestions or otherwise don't pique their interest in the first place.

For example, "jet" is a color palette that is extremely problematic in the DataViz world (oh boy, we're getting there), but part of its broad appeal is that so many scientists and engineers love it in the first place. It's not a good color palette but because it's well-known it is not uncommon for people to expressly request it because they like it. It's familiar and "pretty" (personally I don't think it is pretty).

It's easy to do the right thing and make your DataViz options the most inclusive and accessible they can be, but it is hard to do this whilst keeping everyone you report to, not to mention customers, happy. Especially when they are united in begging you to turn things back to "jet" because they like it and it's familiar. And... it's true, people love it. But if you ever get confronted by this attitude you can always share the following paper with your detractors!

Evaluation of artery visualizations for heart disease diagnosis. - Borkin, MA; Gajos, KZ; Peters, A; Mitsouras, D; Melchionna, S; Rybicki, FJ; Feldman, CL; Pfister, H 2011#

This paper talks about how they compared rainbow colormaps with perceptually uniform ones and found doctors were more like to make errors when interpreting the data in rainbow colormaps compared to perceptually linear options (amongst other things). Give it a read!

Come back for Part 3 where I'll discuss the solution to the various issues I've demonstrated so far.

CVD resources

The Series

This is part of a series of four articles on the importance of quality data visualization. Find links to the rest of the series here (as they are published):

Connect with Ansys