Amanda Cox: Making Illustrations Better with an Annotation Layer

At Eyeo, the afternoon sessions split into 3 tracks.  I attended a presentation by Amanda Cox, who does Data Visualization for the NYT.  She makes interactive charts you can click into to get more information.

Pamplona Spain is the biggest conference for news graphics.  And someone there pointed out a key question in the field:  What’s the difference between illustration and data visualization?

Amanda says: The claim that I want to make today is – there are a lot of lessons in data visualization that are the same kind of idea as the illustrators have.  Something is there in the front because you care about it, something is in the background.   Some detail you want to leave out.  There’s a lot of stuff I can learn from the guys who have spent their career working in 3D and are better illustrators.  There’s a lot of data that should be left out.  How many “backup singers” should there be?  Provide context, make you sound good, but don’t put them in front.

She goes on:  At the Times there are mythical stories with diagrams where every bolt shows up in the diagram of the helicopter.  You could make your own.  Is that really necessary?  People have been playing with this idea for decades. I just need to give you the essence of the understanding. But the converse is sometimes the detail really helps you.  So the idea is “show a pattern, leaver some detail.”  This isn’t a new idea.  Reporters will start with an anecdote.  Or add one in the middle.  It makes it easier to care.

She shows an example.  A great video explaining the pitches of Mario Rivera’s pitches that shows how at the point he hitter must decide how to swing, all his pitches are in the same place and show a confusing spin pattern.  By the time the pitches go over the plate, they are all over the place, so the hitter has very little idea where to swing.  Shows a pattern, leaves some detail.

Then she goes into a list of data that shows republican vs. democrat expectations for senatorial votes.   The table shows something the map doesn’t show. She likes the idea of surfacing interesting things if you show a different pattern, throw something away.  Then she shows a map of Times readers’ comments on a grid to see whether the commenters are positive or negative about the death of Bin Laden, and whether they thought it was a significant or insignificant world event.  There are lots more data points in the top right quadrant (positive that he was killed, think it’s significant) than anywhere else.  And in a ski jump video, she points out that the important part of the video is telling the viewer where to look when. Now look at his arms.  Now look at his hips.  It’s a layer on top of the video that seems simple but leads to increased understanding.

There are interesting things in data that are not entirely obvious.  How can I add things that people might have otherwise misinterpreted?  It sometimes ends up feeling a little like a scavenger hunt, and we don’t want to make scavenger hunts.  The danger online when you don’t show discipline it becomes like a free for all.  “Here is some data, hope you find something interesting.  If you do, could you tweet about it?”  Pushing this on other people is a mistake.  There needs to be a proper annotation layer.  Don’t say “Here’s an interface, now go ahead and browse for the rest of your life.”

Another comment she made: “I think it’s important to really sketch this stuff – on paper.  It can help.  You have to work through a few different forms, try stuff on till you find something that works.”