data visualization | faculty | pedagogy

What a Sinking Ship Can Tell Us About Teaching Data Visualization

It was an icebreaker idea I first heard about on a podcast.


I planned to use it during my first Visualizing Data class, itself the first required class in the new Journalism + Design program, bringing students together from across The New School. I wanted to get the students talking, and to get a sense of what the hell I was supposed to be teaching brand new students in a brand new class in a brand new kind of journalism program.


Here’s the idea: After I’d gotten the preliminaries done — jibber-jabber about myself, point at the syllabus, ask them about themselves, hold up the assigned reading, any questions? any overtallies? — I’d take out a few carefully-chosen examples of data visualizations and show them to the students. One of them would be a classic, one of them would be terrible, one would be really good, and etc. During the exercise, they were to gather into groups, talk for a little while about the visualizations, and then present their ideas to the class.

I liked the idea because it would give me a first crack at talking about why we visualize and what makes a good visualization. Maybe the conversation would be about data quality, or perceptional accuracy, or journalistic honesty. There could even be, I hoped, teachable moments!


Florence Nightingale’s coxcomb diagrams, courtesy of

So I pulled some examples from old presentations I’d done — the classic Florence Nightingale “coxcomb” charts from the Crimean War made my list very early, as did a terrible Good Magazine graphic from 2009 of corporate bankruptcies represented by sinking ships.


To the collection I added a map from the 1910s that I found on Visualoop, and a few other graphics — one of them, called “Vampire Energy,”about inefficient electronics, which was essentially a bar chart drawn on top of a vampire holding open his cape. I barely looked at it when grabbing it, but thought it would be funny. Another was a Good Magazine graphic about the regional popularity of illicit drugs drawn as a stacked bar chart made out of psychedelic letters. (For what it’s worth, I wasn’t really picking on Good — they just make a lot of graphics and have a very particular visual style that I thought would spur conversation).


I didn’t tell them anything about the visualizations. I just said I wanted to know whether they liked the graphic, or if they’d stick around to read a story attached to the graphic. I asked, “Would it have been better as a table of numbers?”


They broke up into groups. I stood at the front of the room and prepared to dispense wisdom. After some discussion, I asked the groups to present their findings.


First up, the group who looked at the Nightingale roses. It’s a classic in the data visualization canon, oft-cited and taught reverentially in data visualization classes. I’m pretty sure Alberto Cairo has a t-shirt of it.


“It’s ugly,” said a student. “And the reproduction quality is awful.”


I explained Florence Nightingale, the Crimean War, and the history of data visualization in about 30 seconds.


“They didn’t have Illustrator then,” I pointed out. I asked them to cut it some slack because people were still inventing the rules of data visualization in the 1850s, and that polar area charts are an advanced, somewhat discouraged, technique even today.


But the students pointed out that it was in the wrong order, with the mortalities between 1845-55 shown to the right of the mortalities from 1855-56. Even in the mid-19th century, I had to admit, people read from left to right. And unless you read the annotation, the graphic was incomprehensible.

The “Vampire Energy” graphic got higher marks. The students first observed that their cellphones did most of the things the graphic split among devices — radio, cordless phone, game console, etc.


GOOD Magazine’s Vampire Energy Infographic

They also understood, without knowing about “data-ink ratio,” that the image of the vampire played a role other than displaying the data. I made the case that the image was superfluous and ultimately distracted the reader from understanding the data. I may even have used the word “chartjunk.” The students disagreed, arguing that cleverness drew in the reader and made the graphic better than a simple bar chart could have. They pointed out, in a way I’d never quite thought about, that visual puns and playfulness give readers a reason to stick around to explore a graphic instead of quickly extracting its meaning and moving on.


Another group looked at the graphic about corporate bankruptcies. The graphic is a staple of presentations I give in front of grown-up journalists. I call it, with added histrionics, “the worst graphic I’ve ever seen.” Through it I explain a host of graphical “anti-patterns” — mistakes you should never make.


A brief overview of the visualization’s flaws: years are arrayed on the y-axis, though the distance between tick marks is uneven to simulate three dimensions. It’s only the baseline of each bar — where the sinking boat is entering the water — that is related to the y-axis. The height of the boat from waterline to bow is not; rather, the volume of the boat represents the pre-bankruptcy assets of each company. The graphic is heavily labeled and would be incomprehensible without its labels.


We had a good laugh at its expense. I was able to explain how volume grows exponentially compared to length, and how bad humans are at accurately perceiving area, and so on. But to my great surprise, the graphic had its defenders.


“I don’t think this is as bad as you think it is,” said a student sitting near the front. Other students nodded in agreement. They stipulated that it was flawed, but thought it was successful at explaining how GM’s bankruptcy was part of a spike in enormous bankruptcies during the financial crisis, and at giving some context about the size of GM’s bankruptcy relative to some truly titanic cases, like Lehman and WaMu.


It was the first time I thought about the graphic this way, and I had to admit they had a point.


The exercise was my way of observing them — where they were starting from in terms of graphical literacy, where their tastes were, etc. We’d known each other for 90 minutes. Everything we were going to learn lay ahead of us.


And though as I write we’re still only a few sessions into the class, it’s an anecdote that’s representative of the way the students look at the world. They’re damn quick, and even if they don’t know yet know what a histogram is or if they haven’t yet read Cleveland and McGill, they’re courageous and whip-smart, and bring with them a keen visual insight and a fresh set of eyes on conventions I’d thought were long ago settled.


And it also reinforced why teaching data visualization is necessary. Though we sometimes assume we’re born with the ability to read charts and graphs, we’re not — though it does benefit from innate cognitive abilities.


Data visualization is a field whose basic vocabulary of bars, circles and lines were invented a little more than 200 years ago. It took a century to establish the basic rules we now take for granted. It’s a field that is still inventing itself.

Tomorrow’s practitioners will stand on the shoulders of people who stand on the shoulders of William Playfair, Charles Joseph Minard, Florence Nightingale, and even Good Magazine, but what the class made me realize is that there is always a new generation coming that will see things with fresh eyes, connect with visualizations in their own way, and that will take the craft in new directions.

[author] [author_image timthumb=’on’][/author_image] [author_info]Scott Klein is Distinguished Journalist-In-Residence at the Journalism + Design Program. He is the Assistant Managing Editor at ProPublica.  [/author_info] [/author]