OpenVis 2017 — a recap
My summary of OpenVisConf 2017, a conference all about visualizing data on the web. 24./25. April 2017 in Boston.
OpenVis — what’s that?
OpenVis are two days of talks about data visualization. Applied data visualization rather than academic research — like IEEE VIS for example. It’s about tools, as well as data projects, and ideas. This year it happened for the 5th time. It’s always set in Boston, though the venue changes — this time it was in the beautiful State Room in beautiful downtown Boston.
It’s OpenVis, because everybody can submit an abstract to give a talk, and all talks will be published soon after the conference, also it’s all live transcribed and live streamed — kudos Amanda Lundberg, what a crazy job!
Irene Ros and her team at Bocoup, plus some more people from the community organize the whole thing.
My 4 favourite talks
All talks are online, and all of them were super insightful, entertaining, and/or inspiring. My top 4 picks:
A data point walks into a bar: Designing data for empathy
Lisa Charlotte Rost does so many things! She started off with creating illustrations for German publications such as Spiegel, was an OpenNews Fellow for the NPR Visuals team in Washington, she organizes the Berlin DataVis meetup, she writes about tools, experiments, visualizations on her blog — like, every day, and now she put a lot of thinking into how emotions and data go together.
Because fact is: we care more about things that we have emotions about, we have emotions about things that we can relate to, and the empathy we have is rather limited. Still we don’t want to be lulled into something just because we have feels. So we want rational facts too!
So, as always, find a middle ground. Get to the reader’s empathy by pointing out the relevance of one single data point and bring it into perspective with the big numbers — for our rational side.
Video | Slides | Lisa Website
Keynote: About Uncertainty
Amanda Cox, New York Times, need I say more? She gave a great keynote that shed some light on how to deal with uncertainty in data — and that we should do it more often. It’s in a prediction’s nature that it contains uncertainty, but it is a topic that tends to be neglected especially in news graphics. Because you want to say something, and not just be like: Well, anything could happen really.
During the last big fateful US election, the New York Times published the “Live Presidential Forecast” which many readers had strong feelings about.
The needle was jittering between the 25th and the 75th percentile to demonstrate the uncertainty around the forecast. After receiving a lot of emotional feedback, Gregor Aisch outlined their decision in a blog post afterwards. It was an interesting demonstration of how people don’t like to see uncertainty. They want answers, not probabilities — especially in these very particular times that was. Luckily Amanda Cox assured us that she would do it again, maybe with even more jitter.
How spatial polygons shape our world
You might have heard about the spatial weighting problem in maps. This describes the problem that we can easily be misled by the size of the actual geographic area and the weight it has.
There are many big states that have a rather small population, whereas small states can be dense and represent way more people. The visual weight is on the big states though.
Amelia McNamara pointed out a few ways to tackle this:
- You could use percentage, instead of absolute numbers, but if the absolute number is rather small, outliers weight pretty heavy and that can be misleading again.
- We could do some Bayesian Weighting
- Or use cartograms! But, problem here: geographic features like bordering areas will likely get lost or mixed up.
Also I learned more about a new trend word: Gerrymandering. The relevance of this problem was just demonstrated lately in the US elections. John Oliver explains it best.
What story does your timeline tell?
Matt Bremer aka Mr Time, former UBC Phd student, now a Microsoft Researcher, is digging into time — big times. He et al. surveyed 263 timelines and defined a design space for storytelling with timelines that balances expressiveness and effectiveness. They called it Timelines Revisited.
There’s also a tool: Timeline Storyteller. There you can build your own timelines and switch their representation from a straight line, to spiral, to a free style shape. You can also change the scale of time, from chronological, to relative, to logarithmic, or sequential and apply filters, add annotations.
And if you’re really into the whole timeline thing, Matt is also part of the Timeline Consortium, whose goal it is to find a widely-agreed-upon standard for timelines. That’s the current proposal.
Video | Slides | Matt Website
Tools to check out
- VegaLite: a high-level visualization grammar from the University Washington Interactive Data Lab guys.
- d3.express: Mike Bostock’s new project. An integrated discovery environment. Sounds exciting and will be out soon(?)
- REGL: a high performance functional abstraction over WebGL
- D3 & Canvas: Not a tool per se, but a comprehensive walkthrough of how to use D3 with canvas by Kai Chang.
