chart junk? more like chart hunk!!!
November 2, 2021 6:09 PM   Subscribe

Declutter and Focus. This is your data viz mission, if you choose to accept it. A study on effective data communication (aka chart design) from the Visual Thinking Lab at Northwestern University via Policy Viz
posted by spamandkimchi (17 comments total) 33 users marked this as a favorite
 
Removing axis labeling is a cursed pattern. Fight me.
posted by sjswitzer at 7:26 PM on November 2, 2021 [14 favorites]


There are competing aims here: conveying information accurately or conveying it persuasively. Choose your team.
posted by sjswitzer at 7:31 PM on November 2, 2021 [4 favorites]


So, accurately = team declutter (to a point probably behind removing axis labels) and persuasively = team focus?

Nifty research! Steve Franconeri taught one of my favorite grad school classes, on visual cognition. As a data visualizer myself, it's interesting to think about the limits of these continua (declutter/accuracy and focus/persuasion), which I don't think are orthogonal to each other, though it's tempting to think of them that way. I often find myself in dashboard crit sessions being asked to add back in "accuracy" information so that viewers are persuaded that I know what I'm talking about, but if I've removed enough of it that it's just a persuasive sentence with colorful shapes, people totally buy it and no one asks for the axis labels back, or what the N was, or if the effect is statistically significant. Basically, declutter enough (with an eye to interpretation) and you'll be focusing, but no amount of focusing-without-decluttering will prevent people in the audience from trying to check your math in real time and asking questions that aren't relevant to the effect being described. Cluttered graphs make your audience feel dumb, mindlessly decluttered graphs are boring, and focused graphs are great as long as they're minimally accurate, but can be used for evil by the unscrupulous.

I'm a little thrown by the inclusion of interpretive text on the "focused" graphs in the article - surely this had an effect on how well people remembered the graph and could reconstruct it later? I'd like to see a version of this experiment where the focused graphs only used non-text visual elements as focusers instead of also telling people how to interpret the graph. Maybe that was in the pilot experiment, I didn't read the supplemental material (yet).
posted by All hands bury the dead at 8:28 PM on November 2, 2021 [3 favorites]


I'd like to see an infographic about the people suspicious of focused graphs and how many of them were acquainted with How to Lie with Statistics.
posted by ChurchHatesTucker at 10:52 PM on November 2, 2021 [8 favorites]


The link isn't working for me, but it sounds like others have been able to connect?
posted by Dip Flash at 5:08 AM on November 3, 2021


I didn't read the supplemental material (yet)

Viz top tips
posted by flabdablet at 5:21 AM on November 3, 2021 [2 favorites]


Link seems to be down/broken/gone for coffee.
posted by Thorzdad at 5:34 AM on November 3, 2021 [1 favorite]


There are competing aims here: conveying information accurately or conveying it persuasively.

Whoa! I do this for a living and well, hard disagree.

Persuading isn't about misleading at all. It's just about figuring out a main message, designing information so that visuals, text, and strategy all align, and letting your information exist in a hierarchy of importance so that you can communicate the most important details most clearly.

In fact, it's not really persuading. It's just clearly communicating. Working toward effective communication/cognition - ensuring that your audience understands what you're trying to say - is conveying information accurately. There's a lot of research on what design techniques work to do this.
posted by entropone at 6:51 AM on November 3, 2021 [9 favorites]


It's interesting to me that the paper (archive link) lays out a broad swathe of literature that explicitly teaches these principles for visual design (I'm a big fan of Cairo's work, and Dona Wong's WSJ visualization guide, myself), but says "if this practice were widespread, these books wouldn't feel the need to teach it."

Whereas from my point of view, from the data journalism side of things, what they describe is pretty standard best practice. The reason the books cover it is not because they don't think professional visual journalists declutter and focus their graphics--it's because their intended audience is made up of people who have not yet learned those lessons, because they're early-career or untrained.

Sure, you can find bad visualizations if you look for them (internal business slide decks are often littered with them). But if this is already a well-known recommendation from the literature, and it's already what newsrooms and popular communicators are doing, I don't really understand the point of the research paper except to say "we agree with the experts."
posted by Four String Riot at 8:52 AM on November 3, 2021 [2 favorites]


But if this is already a well-known recommendation from the literature, and it's already what newsrooms and popular communicators are doing, I don't really understand the point of the research paper except to say "we agree with the experts."

From what I've seen, a lot of scientific fields can be very one-track-minded, and do a poor job of incorporating research findings from other fields.

I work in public health, and the researchers that I work with aren't exposed to a lot of visualization research. So, they make bad visualizations. But, when working with them, it's really helpful to point to visualization research so that I can say "Despite the fact that this technique is strange or unfamiliar to you, there is a body of research that supports its use and shows that it is more effective."

Especially when contravening norms in science, academia, and publishing, it helps to cite sources.
posted by entropone at 9:12 AM on November 3, 2021 [6 favorites]


It's back up!

I've become a wee bit obsessed with data viz since my undergrads have to do a lot of case studies that use American Community Survey or Census data and oh my eyes, all the default Excel generated bar charts or unnecessary histograms. The one that broke me and made me devote weeks of my life to becoming a conduit of data viz design principles is the pie chart that showed population numbers from three different years. Three slices of the pie, each labeled with a different year.
posted by spamandkimchi at 3:03 PM on November 3, 2021 [3 favorites]


I'm trying to figure out how to make a graphic showing how five variables change across 2-4 axes.... over time...

So I can use all the tips I can get. thanks for the post.
posted by rebent at 7:23 PM on November 3, 2021


Isn't there some kind of irony behind the fact that they made this a video?
posted by Literaryhero at 7:32 PM on November 3, 2021 [1 favorite]


Looks over glasses to the bookshelf spying the Tufte books, doesn't read article. Chart junk.
posted by zengargoyle at 1:15 AM on November 4, 2021


I'd like to see an infographic about the people suspicious of focused graphs and how many of them were acquainted with How to Lie with Statistics.

I'd suggest reading this rather well written counterpoint to Huff's book.
posted by MattWPBS at 9:41 AM on November 4, 2021


I'd suggest reading this rather well written counterpoint to Huff's book.

So, don't paywall your graphs?
posted by ChurchHatesTucker at 12:50 PM on November 4, 2021


Dammit. Seems to work if you Google the title. Stupid FT.
posted by MattWPBS at 4:54 PM on November 4, 2021


« Older The United Sceptics of America   |   Mundane Halloween Newer »


This thread has been archived and is closed to new comments