[VDS Learning] Design and Data Science for Visual Analytics

The first week of this 4-weeks series will focus on the concept of visualization research and Visual Analytics.

Design and Data Science for Visual Analytics

Since I've started writing this newsletter, most of my energy has been dedicated to data storytelling and its specific instantiation in the news, which only answers to one precise function of visual representations:

Among the three functions of visualization, "communicate" is best exemplified in the news (from J. Bertin, Semiology of Graphics)

As 2023 comes to an end, I would like to use the few remaining weeks til the holidays to reflect upon the lessons learnt from these past months I spent as a PhD student at the frontier of computer networks and visual analytics.

Visual Analytics

I have noticed how hard it was for different research communities to share findings outside of their zone of influence, and many of my ex-colleagues at the CS department did not even know there was research being done in visualization.

Since 2021 the Area Model for VIS groups different research topics in Visualization and Visual Analytics into six areas

Schematically, one of the umbrella under which visualization research can fall is information visualization, another is visual analytics. If the focus of information visualization is on the spatial layout, visual analytics is interested in developing interactive systems in a real-world context. One major output from such research are design studies, a thorough report of the design rationale from observations in the wild.

The model for Visual Analytics was first formalized in "Visual Analytics: Scope & Challenges" (D. Keim, 2015), with a broad scope indeed:

Such breadth may partly explain the limited success of Visual Analytics beyond the academic community, which is usually limited to dashboards or business intelligence.

At the time of writing “Semiology of Graphics”, Jacques Bertin could not have a clue how the software revolution would transform the way we question the data. This additional layer of interactivity made the "processing" function of visualization subject to real time updates by readers.

To illustrate how visual analytics deals with such problems, I have chosen to share a recent and excellent example of such a project published this year at EuroVis, which I had the chance to assist to.

Data + Design = Visual Analytics

Yesterday on X, I mentioned "ChemoGraph", a visual analytics system presented at Eurovis 2023. It provides a good example of what a visual analytics system may be in the context of chemoinformatics and drugs discovery, and does a very good job at coming up with design requirements from qualitative observations.

Data are produced and consumed by humans, and a discipline which has always focused on the end-user needs is design. At heart, visual analytics is the merger of data and design.

Design leverages insights and creativity, and is the process to transform qualitative observations into a problem before finding solutions. This broad definition has made design an elusive concept to understand, because what is the common characteristics of systems design, information design, graphic design, or interaction design?

All these disciplines abstract a set of problems into useful concepts which form basic blocks to solve a problem.

Everyday objects like digital software should present some attributes to be interacted with (The Design of Everyday Things, Don Norman)

Interaction Design is the general abstraction to think about user interaction in the context of digital solutions.

If a visualization were to be interactive, what should it be able to do? How do we come up with such requirements?

Does It Relate To Data Storytelling and Why Should I Care?

Interactivity in the news is perhaps best exemplified by the short-lived hype around the concept of "newsgames", birthed and dead in the 2010s.

Much energy was given to mimicking the aesthetics of 8-bit computer games of the 80s, but these attempts otherwise feature only a flat range of interactions. The limited agency of the reader/player made them basically irrelevant to data exploration.

Like many of this kind, the Uber Game (Financial Times) is question-based.

Nonetheless, playful design is much more subtle, and among a few successful attempts are the following by the New-York Times, such as "You Draw It!":

Or this haunting Voronoi to predict the election outcome:

Probably the most meaningful example I have found about playful design for information design (The New-York Times)

Or the popular "Flashback" game:

Such playful design emphasizes the importance of interactivity to engage readers and create understanding. If that’s of any interest to you, this is what I will talk about in the following 3 weeks.

To know more about the topic:

I will discuss these articles on Twitter this week, so don’t forget to follow my account to not miss the deep-dive!

I’ve written this week issue with the goal to introduce the concepts I will develop in the following days. This is a draft to what I intend to be an online course about the topic, and is a very general overview of the subject. if it was overwhelming, I will repeat and develop many of these topics by the end of year. Feedback are welcome!

See you next week,

Mathieu Guglielmino

Join the conversation

or to participate.