One struggle in how we talk about data visualization is how we understand it: that goals, abstraction, and numbers intersect to make a whole. It’s not just for the mathy types, but for anyone who seeks to understand their world through data.
Data Viz Philosophy: Better than Bar Charts
I’m lucky this year to surround myself with various coffee chats with others. This post owes a debt to some of them, so thank you to Allen Hillery, Lilach Manheim Laurio, and Vidya Setlur for feeding this brew.
Too often, data visualization ends up in box. It’s for the mathy types or I don’t understand how you play with numbers all day… or it should have been a bar chart or if LinkedIn tells me right, you need 10 years’ experience writing SQL queries. Except, that’s not it.
W. E. B. Du Bois, for example, put together a whole library of charts to highlight systemic racism.
Go read more at the Smithsonian or through Allen Hillery’s analysis. There’s a lot here, particularly with this chart. And, of course, there’s iterations of this as a bar chart.
Du Bois’ chart forces you to work. It’s not built to allow quick assessment of scale, but rather, functions in part as art. You cannot look at this and immediately judge percentages or how many times more or less one group is to another easily. It’s hand-drawn with some of the linkages angled, adding a level of imprecision. That doesn’t seem to be the goal anyway. A part of this chart is how you feel: that you see the red spiral, the jarring path to it, and the various abstract metaphors you can find within it. Is it alluding to blood or a noose or more abstractly a vortex of a system that pulls you in? All could be appropriate interpretations.
Parts of the chart are unsettling. Why those angles and what logic drives them? Often, we’re looking at these charts one by one as analysts through a computer screen. Du Bois’ audiences saw this as part of a composition. Note the potential size – these look much larger than a standard 8.5×11 sheet of paper, but looking at single-chart images, it’s easy to forget the scale of these.
They could be seen, potentially turned, and were compositionally arranged. I imagine some were touched, lines like the spiral traced over the glass to understand the magnitude. The exhibit was meant to surround, inform, but also affect change.
Beyond Perception, Intent
“It should have been a bar chart” relies on our understanding of perception. We see the precision of bars better than most things. Except, there’s usually some caveats based on the task. Are you trying to understand which is more and by roughly how much? Bar chart. What about part-to-whole? Gasp, pie chart?! No!!
Task is also code for intent. We do these tasks because we intend to walk away with certain types of understanding or even do something.
Lilach Manheim discussed evaluating tasks and reframing design critiques at the DC TUG recently. It’s well worth the watch.
A portion of her talk showcases iterating through a visualization to something that better lines up with the intent. It’s not what you’d expect, but as she breaks it down, each added item totally makes sense. The best part of all? Use the viz without thinking about it too much and all the pieces support you.
Within the viz, the scatterplot is familiar. The marginal strip plots fall away to the background, reiterating parts and shaping the path for interactivity. Reference lines draw you to the center, creating a quadrant and providing context. Now, here’s the fun part: click.
So much happens without being overwhelming. Small arrows highlight where the country is with the reference lines moving to support. Additional information shows up at both axes to further facilitate understanding this country’s position in relation to the others.
Could it have been a bar chart? Sure, but as Lilach shows in both the workbook and her presentation that particular chart is too limited for the task. Further, she’s not thinking of this as a series of charts, but a whole semantic unit designed around interplay and building context. The end result is superb, compact, and elegant – a tidy message built to clarify and support the task of understanding one country in relation to others.
Philosophy in Design
It’s time to change the discussion in data visualization. Lilach frames it one way with her tool. She provides an assessment framework that allows iteration and design that supports a given task. It’s human-centric and allows users to mature their designs. I can’t wait for her to share more about this.
One struggle in how we talk about data visualization is how we understand it: that goals, abstraction, and numbers intersect to make a whole. It’s not just for the mathy types, but for anyone who seeks to understand their world through data. It’s an artifact, one curated and logged, manipulated via calculations and comparisons, and finally sent out to the world as a message.
The days of charts in isolation are over. In many ways, they never were isolated. W. E. B. Du Bois relied on a wall – almost a mural – of charts to convey his message. They were understood in the context of one another, each building, expanding, or taking an alternate look at a whole. Our discussions about charts – and visualization – needs to change.
Vidya Setlur and I propose a 3-legged stool approach, one that addresses perception, semantics, and intent. Instead of competing, we argue these items work together in design. Perception, we know, is a key component to how we understand and evaluate information. It often ends up the as our precision instrument – which chart is more concise to task? Semantics embeds information, providing clues and rules to how we can interpret what we see. From semantics, we delve into literacy and why some people hate line charts, but understand trends with bars. We can also use semantics to go beyond the chart to units of information or chart clusters, as Lilach so brilliantly demonstrated. Intent, as we see, frames the goal of the message and informs and questions design throughout the process.
Together, these form the core of functional aesthetics in data visualization. Design can be functional, merely boring fold-up furniture that is as disposable as it is memorable. So often, bars are functional. They achieve the bare-minimum task. Aesthetics are designed to evoke emotions, to tease out our small-self and realize the depth of this information. W. E. B. Du Bois does that, while preserving the function of the visualization. His work still has us examining it a century later. Lilach Manheim Laurio provides an excellent interactive example.
Beyond Bar Charts: Functional Aesthetics
Does it seem like there should more to this, that maybe this idea could be a whole book? Stayed tuned, because it will be!
