Innovations in Visualization

InfoVis 683: Discussion of Week 2 Readings (Sep 19, 2011)

Readings Assigned

  1. Sheelagh Carpendale. Considering Visual Variables as a Basis for Information Visualization.
  2. Ben Shneiderman. The Eyes Have It: A Task by Data Type Taxonomy for Information Visualizations. Proc. 1996 IEEE Visual Languages, also Maryland HCIL TR 96-13
  3. Melanie Tory and Torsten Moller. Rethinking Visualizations: A High-Level Taxonomy. Proceedings of InfoVis '04, Austin, TX, Oct. 2004, pp. 151-158.

“The Eyes Have It: A Task by Data Type Taxonomy for Information Visualizations”

Ben Schneiderman

Summary of the Discussion by Mohammad Albaba:

The main point of the paper as stated by the author is to offer a task by data type taxonomy with seven data types (1-, 2-, 3-D data, temporal and multidimensional data, and tree and network data) and seven tasks (overview, zoom, filter, details-on-demand, relate, history, and extract).

Some participants had a different view about the main point of the paper. They believed it tells the different UI interaction tools that make communication more effective. In other words, they considered the seven tasks listed by the author as ‘interaction tools’ more than tasks. For example, ‘finding details’ could be a task while ‘zooming’ is one interaction tool to carry out the aforementioned task. Another view is that the main point of the paper is to provide guidance about discovering data. The data discovery process starts by gaining an overview of the data, then focusing on interesting parts while filtering out irrelevant stuff. One opinion deemed the tasks described in the paper as ‘the basic functionalities that should be supported by any visualization’. Some participants also found the line to be blurred between ‘interaction techniques’ and ‘tasks’. Some views considered the current taxonomy to be confusing taking into consideration the fact that overlapping can exist in the tasks and data types and no crisp classification can be made in that case.

The discussion then diverted to ‘The Visual Information Seeking Mantra’ mentioned in the paper which is: “overview first, zoom and filter, then details-on-demand”. Dr. Carpendale wasn’t sure about the absolute correctness of ‘overview first’. She said that in many occasions or examples from our real life, we start by the details first then we gain an overview about the subject of interest. For example, we study the details of programming in the first years then we study a class about comparative programming or programming languages at late years. Her point was: “Why should details be an afterthought?” Also, one downside of the current Mantra is that it creates a frozen mindset about the tasks that should be performed when seeking information whereas there might be better alternative ways that exist at different situations and contexts.

The discussion also addressed ‘history’ element and its role in information visualization. History is considered the most neglected aspect in InfoVis. It provides a different type of context to InfoVis and is deemed as one of the challenges in the field.

Last but not least, the discussion mentioned the shortcoming of the usability study conducted by the author as it lacked a lot of details and context. Also, the agreement of the participants doesn’t always correlate with statistical significance.

Class Discussion on “Considering Visual Variables as a Basis for Information Visualization” by SheelaghCarpendale

1) What was the main point of the paper?
Amira: The main points of the paper were to unpack and discuss Bertin’s concept of visual variables, show how to use them in representing information, and show how each variable can affect the success of this.
2) What did this paper mean to you?
Ahmed: We should use visual variables to link aspects of the data we are visualizing, i.e. size. The explanation provided for each visual variable can help us choose which one(s) is the best fit for representing our data sets.
Ovo: This paper presents visual variables so we can see, stage by stage, how a visualization can be formed.
Sheelagh: These visual variables are basically an alphabet for how we work in this class. This alphabet lets you in on some way of representing data, not necessarily a good way or bad way.
Amira: It seems basic at the beginning but it can actually be quite difficult to choose a strategy to apply to a data set.
Sheelagh: Like any alphabet, learning to apply it can take time.
Mohammed: It’s interesting to discuss Bertin in the context of computational displays, in particular, the visual variable of motion. (Bertin worked with static, printed 2D maps.)
Sheelagh: Motion is powerful a powerful one and we are very sensitive to it. For example, we can’t stand flickering websites. However, if there is a slow change over time, such as an animation of a gradual change in the data set, it can be very helpful. We can watch this become that.
3) What practical use can you make of the contents of this paper?
Ovo: Greyscale and texture can be applied to a topographical map of a wind farm showing levels of pressure, which would tell us where the area of maximum wind speed is. The data represented here can be overwhelming, but motion could be a natural visual variable used to represent the data, given that wind moves.
Sheelagh: Greyscale is a monochromatic scale that gets continuously lighter (or darker) in one direction, i.e., red to white.
Mohammed: Are the qualitative visual variables mentioned in this paper applied to humans viewing the infovis or computers viewing it?
Sheelagh: Humans. Although a computer can detect the small differences in size, colour and position through machine vision. We could actually provide more information in an infovis that is intended for a computer because it can detect these differences.
Mohammed: Position is the most versatile and powerful visual variable. On what basis?
Sheelagh: It addresses all of the characteristics of visual variables with the most noticeable difference. Cleveland came up with this same assertion. Whatever aspect you want people to understand the most about your data, try representing it using position first. Cognitive science suggested that this is the most salient visual variable.
Matthew: It would be interesting to redo a cog sci study about this now that people have grown up with technology.
Sheelagh: Cog Sci has made universal statements about human perception, mainly that fundamental perception has not changed. Sociologists, however, are currently looking at how people are changing because of the web.
Matthew: Like we were talking about earlier with motion, we’ve learned to ignore flashing things on screens.
Sheelagh: There are also studies about how we can reduce the cognitive load of a visualization because we have learned to ignore or overcome busy visuals. That being said, we as humans have the tendency to make do with poor tools. If you give someone poor tools to complete a task, that person will probably complete the task anyway.