InfoVis 2013 is this week in Atlanta

InfoVis is the premier annual conference on information visualization. It is part of the annual VisWeek, a whole week of visualization research presentations including VIS, showcasing scientific visualization, VAST, showcasing Visual Analytics research, and InfoVis, focusing on more abstract visualization techniques. We are frequently asked by clients to build internal dashboards and consumer-facing branded utilities that deal with volumes of data, decision making, explanation, and analysis. So we like to participate and keep up on advances in the field.

This year, I helped to organize a full-day workshop at the conference around NewsVis, an emerging application of visualization to journalistic practices. We’re specifically focusing on how mobile (mobile reporting, citizen reporting, mobile consumption of news) and how breaking news are influencing and being influenced by visualization. We have talks and a panel in the morning and a hackathon this afternoon.


Also, I wanted to point interested readers to other developments in Infovis. Here’s the Infovis 2013 program so that you can follow along too. Our bias at THINK is toward systems and frameworks that can help us make better design choices for infographics and visualizations. We are also following interaction and exploration models that move beyond the desktop — the internet of things means that phones, clothing, walls, and spaces will convey information and allow interaction with visualizations. A quick word of thanks to Zhiyu Ding at MIT who’s collected many of the pre-press papers. All of my links to the PDFs (and all the others) are thanks to his hard work.

What Makes a Visualization Memorable? [PDF]
Michelle A. Borkin, Azalea A. Vo, Zoya Bylinskii, Phillip Isola, Shashank Sunkavalli, Aude Oliva, Hanspeter Pfister


cc4ab9ca5354cec909ebebb1e3ae8234It’s not earth shattering, but it turns out that graphs with images produced by professional graphic designers are, lo and behold, more memorable than those that are not. When the creator is someone affiliated with a scientific paper or the government (hardly ever produced by a graphic designer), then the memorability goes down. Note the difference between “bad” data-to-ink ratios between a professional vs non-professionals (in the left plot above).

I’ll also note that the authors did little to take the conclusions of the study to heart and attempt to create better charts for their paper by basically any measure they found as important in memorability in their work. But here are some scientifically validated data graphics that are particularly memorable. The paper has some specific guidance too.


An Interaction Model for Visualizations Beyond The Desktop [PDF]
Yvonne Jansen & Pierre Dragicevic

This is a really cool paper showing off a generalized interaction model that can apply to screen-based visualizations, as well as larger wall-sized, 3D, and even ‘one shot’ visualization sculptures. It also has a set of innovative and beautiful new systems that are exploring this space.


A Deeper Understanding of Sequence in Narrative Visualization [PDF]
Jessica Hullman,Steven Drucker, Nathalie Henry Riche, Bongshin Lee, Danyel Fisher, Eytan Adar

Narrative visualization is a relatively new subtopic for visualization researchers. As visualization has gotten more casual, it’s become clear that providing a narrative scaffolding for viewers and users is one of the best ways to increase engagement and retention. One of the early papers in this space is Segel and Heer’s from 2010. This year, there is follow-on work to understand how best to help creators make narrative order, ‘sequence’,  part of the design space. They performed an experiment to find out what useful and natural transitions make the best stories. The conclusion: Temporal transitions are preferred to transitions between dimensions (how outcomes differ for two populations), and preferred to Measures, exploring all of dependent variables (the authors call this a ‘measure walk’). Finally, all of these animations are preferred to transitions between levels of granularity, such as when a visualization zooms from the world to the US to a particular state.



Information Visualization and Proxemics: Design Opportunities and Empirical Findings [PDF]
Mikkel R. Jakobsen, Yonas Sahlemariam Haile, Soren Knudsen, Kasper Hornbæk

Nice paper that uses proxemics, the idea that personal space is culturally and context dependent (originally from Anthropology and it’s now used a lot in HCI research). The authors built a set of different interaction techniques for body movement to control a visualization on a large screen. What the video for a demonstration of the systems. Human-zooming, i.e., walking toward the screen to zoom, an interactive visualization is pretty neat!



Visual Sedimentation [website / demo]
Samuel Huron, Romain Vuillemot, Jean-Daniel Fekete

So many things on the web are in the form of a stream of items. Here’s some really nice work to deal with ‘streaming data’ in an infovis context. It’s hard to visualize both the items themselves and the attributes of the stream (its flow rate, trends, comparisons to other streams). Huron et. al., introduce visual sedimentation, an animated way to show these features, and it also builds more common infovis chart types such as bar charts, stacked bars, and even pie charts using the same sedimentation metaphor.

The software they have is available on github (and based on D3, a popular way to get visualizations out onto the web via javascript and a cool declarative syntax).



SketchStory: Telling More Engaging Stories with Data through Freeform Sketching
Bongshin Lee,Rubaiat Habib Kazi,Greg Smith


Another system showing the future possibilities of visualization. SketchStory uses a large multi-touch monitor to build a new kind of visualization presentation system. The cool part is that the system recognizes the semantics of visualization, so that a user can draw axes, and then ‘sketch’ in the data. The system retains the whiteboard style visual tropes, for really informal and collaborative visualizations. But while appearing informal, the visualizations are backed in real (or even real time) data, and use infovis smarts for things like linked views, brushing, and highlighting.

Visualizing Fuzzy Overlapping Communities in Networks [PDF]
Corinna Vehlow(Universität Stuttgart),Thomas Reinhardt, Daniel Weiskopf (Universität Stuttgart)


Here’s some work I like because it’s so simple (I know that the image above is anything but!). In many network diagrams, nodes are not definitively segmented or clustered. Instead there are nodes that are fully in one category or ordering and others that are really properly situated between two or more categories. In most network diagram systems, spatial position is the only way to show this property. But this paper shows a technique for creating an abstract ‘fuzzy’ visualization of a network diagram. Looking at A, B, and C above left, it’s easy to see how much better looking it is than D, where all of the data is visible.


At the most fuzzy, only the categories are displayed — and a starburst is used to denote how complete a particular cluster or category is. Then at progressively more detailed, the network nodes are made visible,but with a halo to denote their degree of membership.