User Tools

Site Tools


Visual Analytics Course


This first course presents the fields, and roughly follows Ward et al.'s first chapter.


  • What is InfoVis by examples
    • Directory tree dataset
    • Cars dataset
    • Poker player dataset
    • Air traffic dataset
  • What are we trying to achieve (as a user)? What tasks, why this visualization?
    • Three possible visualization scenarios: presentation, confirmation, exploration (Tukey)
    • Munzner's nested model (see also her book introductory chapter)

On Munzner's nested model for visualization

Munzner is on the way to publish a book on Information Visualization. I had hoped the book would be ready earlier :-) Munzner's book is structured around a new model of visualization that unifies design and evaluation issues in a single framework.

Munzner four levels model Visualization design is a complex problem that can be made easier by splitting concerns into four cascading levels, as shown in the left figure. The top level of problem characterization is to characterize the problems and data of a particular target domain. The next level of abstraction is to map those domain-speci c problems and data into abstract and generic tasks and data types. The third level, encoding and interaction, is to design the visual encoding and interaction techniques that use the data to support those tasks. The innermost fourth level of algorithm is to create an algorithm to carry out that design efficiently.

Designing and building visualization is not just about creating nice images. Munzner again:

Computer-based visualization systems provide interactive visual representations of datasets intended to help people carry out some task more eff ectively. When we design a visualization, how do we gure out if we have succeeded? There are many criteria we might use. We could ask whether somebody using the system can do something better.

  • But what does better mean?
    • Do they get something done faster?
    • Do they have more fun doing it?
    • Can they work more e ffectively?
  • But what does e ffectively mean?
    • How do we measure insight or engagement?
  • And better than what?
    • Another visualization system?
    • Doing the same things manually, without visual support?
    • Doing the same things completely automatically?
  • And to do something better - what sort of thing?
    • That is, how do we decide what sort of task they should do when testing the system?

Munzner four level model together with potential threats and validation issues

  • And who is this somebody?
    • An expert who has done this task for decades,
    • or a novice who needs the task explained before they begin?
  • Are they familiar with how the system works from using it for a long time, or are they seeing it the first time?
  • Even a concept like faster that might seem straightforward gets tricky. Are they limited by the speed of their own thought process, or their ability to move the mouse, or simply the speed of the computer in drawing each picture?

Threats and validation. When diving into the four levels while designing a visualization, design choice must be made. How are we sure these choices were the right ones to make? Munzner has an answer, and unfolds her four level model together with a list of threats and possible validation to these threats.

Information Visualization is a multi-disciplinary activity – and the model partly explains why. Designing a system or technique involves domain expertise, cognition, graphical design, human-computer interaction and dialog, and algorithmic efficiency. not to talk about multi-disciplinary issues that pops up when collaborating with experts from any specific domain (biology, social sciences, etc.).

Note (Munzner): This nested process model collapses algorithmic and implementation considerations together; a more detailed model could pull them apart into two separate levels.

Exercises / Assignments

  1. Give an example of each of the three possible visualization scenarios: presentation, confirmation, exploration. You should preferably find something on the web and document your findings by giving the links to your sources.
  2. Get the vehicle data from Ward, Keim and Grinstein book website. Explore the data without going into fancy visualization tools, maybe use Excel. Try to identify correlation between some of the data items (exploration), state a hypothesis if possible and prove it using standard statistical computations (confirmation), think of a way to present your conclusions.
  3. Choose an example visualization system or technique and discuss how Munzner's model has been (or has not been) applied.

The visualization pipeline

This is a rather common way of looking at things shared by the InfoVis community. The good thing about it is that everybody agrees on it. Another good side of it is that the pipeline tends to conciliate both the user's perspective and the engineering perspective -how you build a viz system). The bad side of it is it focused on the automated side of InfoVis (processes and algorithmic) rather than on cognitive processes.

Here are a few picks collected on the net:

 Card, S. K., J. D. Mackinlay, B. Shneiderman (1999). Readings in Information Visualization. San Francisco, Morgan Kaufmann Publishers.  dos Santos, S. and K. Brodlie (2004). Gaining understanding of multivariate and multidimensional data through visualization. Computers & Graphics 28(3): 311-325. Chiricota, Y. and G. Melançon (2007). Visually mining relational data. International Review on Computers and Software, May 2007.
/net/html/perso/melancon/Visual_Analytics_Course/data/pages/introduction.txt · Last modified: 2016/05/19 22:36 by melancon