Bit by bit: how to show frequencies with animated unit visualizations
Frequencies like 200 coffee cups sold today or 1 out of 300,000 people develop cancer from a toxin seem deceptively simple. Data visualization enthusiasts could tell you a story about both of those statements with pictograms, icons arrays, and waffle charts. But these get complex if you want to animate them and create more than one at a time.
In fact, I did this for a class project on a research paper a few years ago.
Here’s the icon array I drew in PowerPoint:
I visualized the risk of pedestrians dying from distracted walking. The units, once you crunch the numbers may seem strange. 23.1 micromorts means a miniscule risk of death and the icon array shows the skull representing the 7,668 deaths among 320 million people in the US. In my icon array, the skull was actually larger than it needed to be so. To correct that, I animated it shrinking to the point where it’s barely visible (0.00002 deaths out of 1,000 dots). That’s how low-risk distracted walking is for US pedestrians based on 2010s data.
What my icon array doesn’t show is the hard work it took to line up the dots in PowerPoint, color them, and shifting them into place as my visualization concept evolved. Animating the skull’s shrinking was much easier, but I would have like to do that without getting tangled up in PowerPoint’s point-and-click interface.
That’s where the DataParticles paper from the CHI 2023 conference becomes useful. The researchers behind it created a special software program just to create animated unit visualizations (AUVs).
Helpfully, there’s a video walkthrough.
Instead of pointing and clicking a bunch of options in PowerPoint or Excel, you can just type commands in plain language like Here are 28 popular Nespresso coffee pods (Fig. 3a below). Users will see their unit visualization as organized dots in the center (b) and animation options to the right (c). The simplified UI lets you animate the dots as they enter and exit. I especially like that the dataset grounding the visualization stays in view (e).
From a research perspective, I appreciate the work the authors did combining methods. They used:
Semi-structured interviews: interviews with designers, journalists, animators, and others showed that it’s hard to create animated unit visualizations because there are too many tools that need to be coordinated
Content analyses: by studying 40+ diverse examples of unit visualizations from the web, the authors discovered that modular content, aligning text and visuals, and creating a story from overview to details was common
System design: the DataParticles software program was designed and developed as a new way to quickly and easily write visualization commands, see results, and animate with dropdown menus
User study: nine experts participated in a virtual usability testing session to do the same tasks and report their experiences on a questionnaire and an interview
My main gripe with the user study or evaluation is that nine experts don’t represent the potential universe of users. Why not also test how usable this tool is for novices like undergraduate students in different majors and employees with low data visualizations skills? I’d like to see someone with limited data visualization experience feel empowered with DataParticles. In general, I want to see researchers test their inventions on a range of users with varying backgrounds like skill levels, interests, and occupations.
As a data visualization enthusiast, I’m curious about how well these tools can infuse best practices from psychology and data visualization. If there are more specific ideas about which icons to use and animation practices, I’d like to see the system use the most effective solutions. What if the psychological studies on communicating frequencies and risk could be used in DataParticles? Considering the strong lack of connections to psychological studies in the human-computer interaction field, I’ll keep this in mind.
The user interface reminds me of the best parts of PowerPoint (modularity, text view, visual view, and dataset view) without the task switching that makes PowerPoint tedious. It excited the participants in the user study because it presented the text commands, visuals, and data in coordinated parallel views.
I don’t expect any of the major data visualization tools from Apple or Microsoft to change in this way, though AI-authored data visualization may be possible given Microsoft’s savvy licensing of ChatGPT.
Research projects like DataParticles may be best for third-party and open-source developers who can move more quickly and be more flexible with product development. I’d like to see a set of tools like these for the most common data and visualization types. Authors and audience would both appreciate it.
References
[1]
Cao, Y., E, J.L., Chen, Z. and Xia, H. 2023. DataParticles: Block-based and Language-oriented Authoring of Animated Unit Visualizations. (Apr. 2023). DOI:https://doi.org/10.1145/3544548.3581472.
