We were very lucky to have a joint talk given by Miriah Meyer and Bang Wong as they were passing through Europe, on their way to the EuroVis conference in Bordeaux.
Miriah and Bang, who co-ordinate that Data Visualization Initiative over at the Broad Institute, talked to us about the Visual Representation of Science for Communication and Research. They explained how they have been able to bring the ability to deal with large datasets together with creativity and artistic flair, and create a design methodology that they can apply to the challenges of scientific visualization.
Below is a write-up of my notes from their talks.
Speaker #1: Bang Wong
Download Bang’s talk (ZIP file)
The art of visualization
Bang Wong
Starting from a scientific background, Bang developed his skills with pen, pencil and carbon dust, and is now an accomplished illustrator. He has been able to bring these skills to bear in working on the meaningful representation of data.
Duhrer, da Vinci, Galilieo, Brödel… they were all trying to represent and capture scientific insights with their illustrations and sketches. We do the same, although perhaps with the advantage of more advanced tools. As Bang remarked, “…illustration allows the artist to edit the scene, set focus, and show something that cannot be seen.”
But scientific illustration and data visualization, although often beautiful, are not art. The role of design is to address the need to figure out what is primary info, what is secondary, and what is extraneous, and how to represent that.
“Visual conventions are a learned language”
Bang’s work does not only deal with on-screen or paper visuals; he has also been responsible for creative direction for the DNAtrium installation, which sits in the lobby of the Broad Institute. This is an innovative way to allow the public a rare chance to see real data. It includes physical devices such as window-mounted controllers; driving past, you can see big headlines; walking past you can read some detail; going in, you can interact and unfold the survey.
Insights through visualization
Importantly, scientific visualization can allow us to clarify scientific topics and engage people in discussion. Echoing Edward Tufte, Bang used the example of Anscombe’s Quartet to demonstrate the value of visualizing (in this case, graphing) data as a key part of its analysis.
As a designer, one is able to influence the salience of certain points, and to highlight what is important, sometimes even showing what might otherwise be hidden (e.g. the scope of possibilities available to a medical illustrator vs a photographer).
Towards the end of his talk, Bang spoke about visual encoding – the careful use of colour, alignment, proximity, and the idea of linearising complex pathways to allow their comparison on a common axis.
Speaker #2: Miriah Meyer
Download Miriah’s slides as a PDF
The value of a defined methodology
Miriah Meyer
Miriah outlined for us the method or approach that she (often in collaboration with Bang and others) will take to solving visualization and communication problems through design.
The approach incorporates principles of user centered design (UCD) that will no doubt be familiar to readers of this blog! 
She stressed that having a defined methodology to apply to projects can help to consistently yield results. It further helps to demonstrate that considering usability can bring scientific reward, and is very important progress [1].
“a methodological approach to viz development makes effective design decisions salient”
As part of her talk, Miriah demonstrated the Pathline application that she has worked on, pointing out how certain key pathways could be linearised and then compared side-by-side, amongst other things.
She also showed us MizBee, demonstrating its application in comparative genomics, and the visually encoding of conservation relationships. This example allows researchers to see that there is noise in their algorithm, and suggest that they need to rethink that before dealing with the results.
Miriah’s design process:
- understand (questions, constraints, measurements, goals)
- abstract (data types, tasks)
- design (visual encodings to support those)
- implement
- validate
Understand
Throughout the design process, we should aim to gather information about users and stakeholders, and feed that into the development. This is an iterative, non-linear process ["Step away from the waterfall!" - Ed].
(Sounds familiar, right?!)
Abstract
[using Pathline as an example]
What exactly is the data?
Metabolic pathways; gene expression, similarity scores, phylogeny.
Tasks and goals – What do users do? What do they want to be able to do?
Studying expression data as a time series – peaks, valleys, time shifts
Comparisons of time series and similarity scores
Design
Proper iterative design process; brainstorming
All before writing any code….
(reminiscent of Jesse James Garrett’s surface layer in his Elements of User Experience)
Implementation
Bring all that together in a working product (prototyping).
Validation
was the abstraction correct? did we understand tasks? right data?
(recommends having a look at “A Nested Process for Viz Design and Validation” – T Munzner, IEEE Infovis 2009)
Have there been efficiency gains? Have we been able to gain new insights?
Reflections (from both Bang and Miriah)
- Having a defined process helps to focus development
- Fast, nimble prototyping is essential
- Refine methodology for viz design – partuicularly the abstract and validate stage
- Working without images, or at least in lo-fi for as long as possible, to avoid directing the collaborators and users.
see also Seán O’Donoghue et al, Nature Methods, vol. 7, no. 3s, “Visualizing biological data – now and in the future“, (2010)
Tags: anatomy, art, data, design, illustration, medical, methodology, Visualization
This reminds me of a podcast I heard recently on uxpod.com entitled Visual Communication – an Interview with Dave Gray: