SARC 5400: Data Visualization
I took this course last spring. The best data viz course you can take. No, it does not focus on technical tools. The class gets at deep philosophical questions about data, display, and perception. What is data? How do you show it? What do you see? The metaphysical and philosophical questions on Technology, Data, Visual Forms, and perception were some of the coolest parts of the course. But then going on and converting that to actual visualizations in the world through D3.js, Tableau, and standard Python libraries was even cooler.
Introduction
informare: to give an essential or formative principle or quality to
technology: the application of scientific knowledge for practical purposes
visualization: a tool to discover something that the author does not already know
Gehry’s Disney Concert Hall, Pedagogical Venn-diagram, (syn)thesis, Japanese Scroll, 1 + 1 = 3
In the very first 10 minutes of the class, the lecturer pulled up a Data Viz of the class containing an interactive bubble-plot visual that showed the categorical groupings of the class by major, year, ugrad/grade, and other info. Then, it proceeds in the most interesting fashion of introducing interesting course material. First, the Oxford Dictionary and the etymology of the words being covered. Then, showing us all the visualizations possible, challenging what it means to visualize, and tells us simply to see.
https://benfry.com/zipdecode/ Enable the zoom option!
Language, Graphics, and Communication
language: a system of arbitrary signals, to which we assign meaning through convention
prosody: the patterns of stress and intonation in a language
A course in visualization needs to spend its time covering semiotics and language. The ability for humans to convey themselves to others. And as a corollary, just as humanity has developed linguistic grammar, we also have visual grammar. As an example, think of not only how SOS sounds but also what it would look like.
The Material of Information
tectonics: using material and construction methods in such a way that they form an integral component of the design and actively help to shape it
data: the reconciliation of ourselves with our surroundings and environment
“graphics is one of the major “languages” applicable to information processing”
Telescopes, School of Athens, Pythagoras Theorem
Karl Popper and his three worlds were invoked. The implications are on what we as humans use to perceive and on what there is in the world. And to further that, the history of film was taught. Who knew what daguerreotypes were? The rest of the lecture went on to depict the different ways the five fundamental human senses, sight, sound, touch, taste, and smell, interact with “raw materials” to create perception. It tends to end with the Magic Number 7 as a fun exercise for the class.
Structuring Information Aspects
“How can we map retinal variables with informational variables to construct an experience, to illicit spatial-temporal-visual-causal intuitions and insights?”
The course’s fundamental question is posed: How can we map retinal variables with informational variables to construct an experience and illicit spatial-temporal-visual-causal intuitions and insights? Through this, we are introduced to InkScape, Observable, and essentially the “engineering“/technological aspects of the course.
Escaping Flatland
“because you simply cannot think with images.”
“I do not paint things, I paint only the differences between things.”
“What we seek instead is a rich texture of data, a comparative context, an understanding of complexity revealed with an economy of means.”
Galileo, Schreiner, IBM Series, Mississippi River, Goals Poster, US Memorial, Wharf 5 Tide Prediction, The Matrix
Referring to the film Flatland. This lecture focused on such escape through a visual display of more than two dimensions of data, more than the x-y scatterplots. Before that, though, there is such a fun shortcut to the musings of Postman in his book “Amusing Ourselves to Death,” which illustrates a critique of visual society. The notion of images and appearances has become reductive for actually “thinking” or any novel interpretation. The shortcut serves as a counter-point to consider because the point of the class is to show how visuals as a medium can carry a lot of novel knowledge. However, visuals should not aim to show “truth” or “raw information” but rather the insights and relations beyond the surface.
An interesting tidbit to take away from this is that unorderly visuals are not an overload of data, just a failure of design.
Edward Tufte: the Grand Principles of Analytic Design
- What is the question? - “Do no harm!”
- Show relationships and comparison. - what is this in relation to that?
- Think and show causality. - Problem solving is causality thinking
- “To clarify, add data!” - be multivariate
- Integrate the evidence. all marks are visual, text, number, diagram, image
- Document everything. take ownership
Techniques:
- Juxtapose: Layer and Separate or Layer and Separate
- Small Multiples
- Micro / Macro Scales
- Parallelism
- Minimalism: Less is more
- Standardize
Tableau and Adobe Color
It was mostly a workshop in Tableau, but it did start off with some basic color theory: triadic, analogous, complementary, monochromatic, and how those inform choices in visualizations.
Graphical Rules
UVA Team Example
How do you assign meanings to the retinal variables? The shape of a line, the color of an object, the gestalt ordering of the collection. What are the ways you can manipulate the DOM with these rules to generate visuals? Tableau is way too limiting in a lot of ways, and that’s why the class has moved to D3.js. This gets to the more technical part of materializing the theoretical founding of the visualization. An important thing to remember is to keep the technology from binding the imagination. Imagine, with theory, Work in praxis.
Games and Visual Information System
“Tools are great. But don’t let the tool drive the process. Because tools really aren’t great.”
play: to engage in activity for enjoyment and recreation rather than a serious or practical purpose
simulation: a structured set of units and relationships that resolve to a result
uva bay game, alphafold, wow plague virus
Lecture starts off with a conversation about Big Data and how correlations will replace theory. Even giving examples of many “high-ranking” individuals are advocating that there is no need for models, just data and math. However, the counter force to that is highlighting data without a model is just noise. Correlation is not causation. Then with the conversation shifted to models and specifically mental models, we are given an epistemological system of proximal and distal. Connecting that back to visual information, is the idea of creating a system that allows for “play” but also contains core concepts.
Spotify overview - Working with Big(-ish) Data
Describes the dataset for the class midterm project. Everyone contributed their Spotify history, and the goal was to have a large dataset with which to tease out some relations.
Tableau and Spotify
Tableau workshop w/ the Spotify data
between
“ubiquitous, incessant surveying smothers knowledge with noise”
ecotone: a transition area between two biomes
311 Complaints
Pi is a number, yes, but it has a deeper visual relationship.
Da Vinci wrote the word dimmi, show me, on his pages. Painting was a means for him to understand the world and science.
Dynamics and Interaction
“increase the capability of a man to approach a complex situation, … to derive solutions to problems”
1 Pixel Moon, Kepler’s Planets
Visuals need not be static. In fact, many good visuals are ones where people can interact with them. I was actually surprised a UI/UX lecture came this far into the course, but I was also very intrigued by how it fits together with visual design. The end takeaway is to deeply consider the possibilities for new technologies as a way to change how we think of visualization. Visuals have the power to be dynamic.
The Data in D3
D3 workshop
Maps
map: symbolic depiction emphasizing relationships between elements of some space, objects, regions, or themes
Nat Geo Inequality, Stereotype Map, Reconstructing a Bike Crash, DC Metro
Maps were the original, most useful form of visuals. It all started from cartography to planning/mapping. As with any other lecture in this course. It starts with asking the question: what even is a map? and then showing the historical progression from Columbus to now. Google Maps is very much a map.
Maps as any other visuals can use the same techniques, relational variables, and visual form to convey information/relations. But, it works specifically if you are looking at a geospatial question.
Gave a very interesting notebook that shows the different kinds of projections into a 2D space. https://observablehq.com/@d3/projection-transitions
Maps - Tableau and D3
Workshop for Tableau and D3 for geoms
D3 Layout lib
D3 Workshop
Wicked Problems
The Ignorance Project, NCASE.me Trust
It begins to tie in and recenter the key goal of the course, which is the ability of visualization to inform on hard problems. But you have to realize that the way you interact with the data affects what you can see. Don’t start from the data. Start from the problem! The rest of the lecture shows very good case studies in how they challenge and bring creativity to visualization.
A thing I reflected on with the course is that data is not neutral, data visuals are even more unneutral, and data interpretation is even more unneutral.
Integrating Multiple Orders
Workshop for Tableau and D3 to make them dynamic
Conclusion
“The challenge for this generation is the synthesis of information”
I’m very glad I took this class.