Graphical Primitives versus Direct Visualization
A rectangle walks in a bar and orders a drink. An attractive circle sits on a bar stool next to the rectangle. ‘Hey Circle, want to get primitive?’ The Circle looks the rectangle slowly up and down and says: ‘Sorry, you’re not my type.’
The main characters in this joke can be seen as one of the stock characters in information visualization. They are also known as ‘graphical primitives’. They are used in diagrams, scatter plots, treemaps and bar charts. But thanks to modern software we are able to make an upgrade of these kind of elements in visualizations.
According to Lev Manovich (2010) there are two key principles that separate information visualization from other techniques and technologies and visual representation. The first principle is reduction: data is abstracted and summarized. This enables us to see patterns and structures behind a chaotic and enormous data sets. The second principle is the privileging of spatial variables. Spatial variables such as position, size and shape represent often key differences in the data. Other visual dimensions such as color, hue and transparency represent less important properties.
Manovich points out that both principles have their disadvantages. By abstracting data there is the risk of extreme schematization. A significant amount of information is lost. Furthermore, by privileging spatial dimensions over other non-spatial dimensions we lose possibly other things. And, as Manovich emphasizes, information visualization uses, unlike scientific visualization, ‘arbitrarily spatial arrangement of elements to represent the relationships between data objects.’ (10)
However, due to current advances in technology we able to overcome these challenges. We are now able to create to approach information visualization in new ways. One of these new approaches calls Manovich ‘visualization without Reduction’ or ‘direct visualization’. In direct visualization is ‘the data reorganized into a new visual representation that preserves its original form.’ (12) An example of this is the tagcloud or the index of a book. Manovich sees this as expansion: the typical graphical primitives are expanded into the actual data objects. Therefore it is unnecessary that in order to highlight patterns in the data one does not have to use reduction.
I find this personally an interesting idea and created a simple visualization to test out his theory about direct visualization. Through Processing, a language especially created for designers and artists, I accumulated images of the fifth season of the television series Dexter. Every minute an image is automatically saved, this process is also called sampling. Which is an important part of direct visualization. When all the data was acquired I developed a small program that would visually present the information.
The image looks interesting. However, I wonder what we can learn from this image. We can see that the episodes have different lengths. Also we can see that the images are quite dark which is not surprising for a show which main character is a serial killer. If you’re interested you can start counting, such as the scenes, characters, locations, and color brightness. As Manovich remarks: ‘by projecting time into space’ (18) we are able to study objects in a new perspective.
Scott McCloud’s Universality
However, by preserving the data objects without reduction we loose something else. In Understanding Comics: The Invisible Art by Scott McCloud, McCloud explains in a visual form how comics work. One of the points he makes in his book is relevant for the ‘graphical primitives versus actual objects’ discussion. See the following two images for his view on simplification.
McCloud sees cartoons as a form of amplification through simplification. By abstracting data, readers are naturally forced to focus on specific details. Furthermore, the more simplistic an image is, the more universal it becomes. For me this is an important aspect, the more people that understand a visualization, the better. And graphics are definitely more universal that the written word.
We could transform the visualization of the Dexter episodes into something more abstract, whereby the average color of the images are calculated. The visualization looks currently as an abstract art work.
This version has certain advantages over the direct visualization. The images in the direct visualization are in the current form too small to analyze. This could be fixed by adjusting the image size or making it interactive (and thus click able). However, for a quick overview the abstracted visualization is more comprehensible. The quantitative version is advisable for quantitative image comparison. It would be interesting to compare the abstracted stills of Dexter with different seasons. Did the use of color change through the years, has Dexter become darker or lighter or did it stay the same? It also could be worthwhile to compare different series with each other. For example, what would happen if you compare one episode of Dexter with one of Sex and the City or Glee?
The results from this series comparison are surprising for me. I would have expected that the colors of series like 30 Rock, Glee or Sex and the City would have more bright colors. We could conclude that this abstracted visualization looses important information. For instance, it does not represent the otherwise significant differences in the series. On the other side, this information loss could be nice for a privacy perspective. Abstracted data has often a higher level of privacy then direct visualization.
The Visualizations of Emotion
According to Gert Nielsen we loose more than information. In his presentation at the Infographics Conference Nielsen argued that statistics and visualizations remove the human aspect of the data:
‘We remove the human emotions and transform them into bar graphs.’
Which is true, it is hard to relate to a bar graph. Let’s do a quick test: an image of Dexter with a needle in his hand or an abstracted rectangle in the average color of this image. Which one do you find more emotionally appealing?
I understand that the previous question is unfair in this context. An image will always have in advantage in the current setting.
Both visualization approaches have certain advantages and disadvantages. If you want to visualize a vast amount of data without needing to know the specifics, you preferably use the primitive graphics. In this way the data is also anonymized, which in some cases of privacy can be an advantage. For a qualitative research the direct visualization approach is probably the best suitable. This approach is as default more capable to show the human factor in visualizations. When all things considered, I would suggest that the designer would first have a clear goal in mind before (s)he chooses an approach. And please remember the quote of Edward Tufte:
To clarify, add detail
Manovich, Lev (2010). “What is Visualization?”
McCloud, Scott (1994). “Understanding Comics: the Invisible Art.” New York: Harper Collins.
Nielsen, Gert (2011). “The Ugly Truth” presentation held at the Infographics Conference 2011 at Zeist
Tufte, Edward (1990). “Envisioning Information.” Graphics Press, Cheshire, CT