I love this slide from @jburnmurdoch's #rstudioglobal keynote: when people read a chart they aren't reverse engineering the visualisation to extract the data, they're expecting to be told a story. I think this is exactly right, but as an academic I'm left with a puzzle...

In academic papers we're "supposed" to encourage people to do the former not the latter. But everyone who writes papers *knows* that human readers (even scientists) read charts as stories, so academic papers and charts are storytelling exercises. It's a property of the readership

"Overselling" a claim in an academic paper is genuinely a bad idea (and so, so common). But I also read (and sadly, write) so many papers with ineffective data visualisations precisely because the papers *don't* follow the principles John's talk outlines. That's not good either

I see a lot of quite vitriolic comments on twitter about it being a "questionable research practice" for scientific papers to tell stories rather than dryly recount nothing but technical detail. I'm... not so sure anymore?

Creating unreadable charts and writing papers with poor narrative structure don't seem like stellar scientific practice to me. The data don't "speak for themselves". As scientists we do have a role as *interpreters*, and yes... storytellers