I recently attended Bruce Morser’s “Drawing: A Creative Approach” art class through Seattle’s School of Visual Concepts. Our teacher, Bruce Morser, introduced us to some tongue-in-cheek “rules” that can be used in the drawing process. These rules inspired me to come up with my own hints that can be applied to not just drawing, but to my domain of work: data analysis. These lessons are not required rules or absolutes, but more like helpful hints, so you can consider and develop your own practice.
Lessons from art class:
*Pick low hanging fruit right away. Easy first.
This one is easy, so I’m putting it first. Get it?
Do the smaller, quicker, easier things first to bag some quick wins and build momentum. Get a surplus of productivity early, and use that to carry you over while you are working on more challenging aspects.
Think about transitions. Know smooth sequels and sharp breaks, and when to use them.
In visual storytelling, the viewer is following our lead and their eyes are moving along a path we set out for them. In a line chart, are all the lines going up and only one going down? In a bar chart, are the bars sorted in ascending or descending order? In a scatter plot, are all dots the same opacity or do they pass through objects behind them? In a line scale, are lines smooth and gradual or extreme and choppy?
In the digital painting below, you can see examples of different kinds of transitions, some sharp, some smooth. Some tell a realistic story (more gradual subtle highlights on cheekbones), and some which do not (sharp transitions on face with overlapping color, or rough shading in the neck or sternocleidomastoid area). On the outer forehead and the hairline behind the ear, I can see lines from my original drawing that I actually colored over in a rush (people familiar with digital manipulation might even notice my sharp masking lines). These transitions, once identified, detract from the story the light is telling. In the same way, a data visualization might have a transition that detracts from the consistency of the overall message or story.
Transitions are not only things that can detract. Transitions are storytelling techniques that data professionals can use to communicate. For example, one data subset can be emphasized by filtering the rest of the data set to be more homogeneous, resulting in sharper relief. Another example is to change the y-axis scale on a line graph, to change the angle to make a slope look more or less significant (you might have heard the phrase “banking to 45”--this comes from William Cleveland’s book Visualizing Data in 1993, and was more recently examined by Justin Talbot, John Gerth, and Pat Hanrahan in their 2012 research paper, titled “An Empirical Model of Slope Ratio Comparisons”).
Plan out the space.
This is not necessarily “draw an outline” or do a mockup, but that’s one way to think about it. For those of you who have seen or coached children in sports, this is what happens when the players run after the ball, no matter where it is. The goal is to put a framework in place before focusing on details and adornments. John Singer Sargent supposedly said "Do not concentrate so much on the features. Paint the head. The features are only like spots on an apple.” In the picture below, I sketched the basic shapes first and plotted where things would go, and then came back and filled in the details later.
Do it at least twice, and find your own optimal design count (2,3,5, whatever)
In class, Bruce suggested that sometimes the first time is always a throwaway. This also allows at least 1 comparison a vs b test. Doing 3 designs means you can do 3 comparison tests (a vs b, b vs c, a vs c). For the self portrait above, I sketched two that didn’t look like me, then the third did. Then I drew a fourth, and a fifth, and the third, fourth, and fifth all looked similar, so I stopped. For myself, I find that 3-5 designs is the magic number (more than that is either detrimental or a no difference).
Use your eyes, not your brain. Know what symbolic representation is, and when you’re doing it.
Bruce called this “draw what you see, not what you know”: convey the patterns of light that are hitting your eyes, and forget about what your brain is telling you that you’re seeing. This is easier said than done, considering that “the all-important human trait of symbolic reasoning helped our species not only survive but thrive” (p. 7) according to John Medina (2014). Our eyes pass the brain information, and the brain takes that data and then makes things up (“I see two circular shapes with a dark spots in the middle of each, and they move with my eyes, so these must be iris and pupils and eyes”). The brain collects the data, analyzes it, applies a system to it, then chooses a match. This can get in the way when storytelling overwrites the source material.
On the left is a comic style symbolic representation. On the right, I drew myself more true to the shades of light and patterns that actually were being presented. The symbolic representation is more abstract and connects with viewers in a symbolic way, not based on the arrangement of light, shadow, shapes, patterns, and transitions. The representation that is closer to what my eyes saw, on the right, tells a more accurate story. The viewer can draw their own conclusions from the more shaded presentation, rather than relying on the symbolism and abstraction of the more comicy first one.
In data presentation, storytelling can be a seductive song, tempting us to over simplify and symbolize (“hey look at this gauge at 100%, we are crushing it! what do you mean this is missing data”), embellish (“our bar chart is double theirs! ignore the fact they do not both start at zero”), or omit data entirely. If data is not there then that should be communicated to the viewer so they can adjust their conclusions accordingly.
Consider a line graph with missing data. About the following charts, Stephen Few wrote “The best way to display the fact that values are missing is to omit the line altogether for those intervals, as I’ve done here.”
Posted by admin