Uncertainty Visualization: Framework Fight Night – 11/26/2012

Miriah’s argument

• Nested Model

  • design model by Tamara Munzer, Infovis 2009

  • Outermost box: domain problem characterization.
    • Next in: Data/task abstraction design.
      • Next box in: Encoding/interaction technique design.
        • Innermost box: algorithm design.

  • The designer moves inwards to the innermost box.

  • This is how Miriah is thinking about the problem
    • if abstraction is wrong, it doesn’t matter how good your encoding is
  • [Bill] what is the value added to existing nested model design?
    • specialized to the context of visualization
• Nested Model (Extension)

  • outcomes of design decisions are “blocks”

  • stacks blocks lead to a specific outcome

  • “guidelines” are statements about relationships between blocks
    • 2 kinds:
      • between level guidelines
      • within level guidelines (compare blocks within a single level)

  • don’t know much about the abstraction level (don’t have good task abstraction)

• Nested Model Extension –> Uncertainty Framework (outlined at all hands meeting)
  • “source" is a domain characterization problems
  • “PDF”, “characterization strategy”, “characterization” are all data/task abstraction problems
  • “visualization” is an interaction problem

Bill’s Argument

• any framework for visualizing uncertainty should explicitly and unambiguously describe 2 things:
  • uncertainty being represented
    • mathematical articulation
  • the visual encoding used
    • articulation much harder

• while distinctly different from a design standpoint, they are closely tied and cannot be reasoned about independently

• bill’s original framework is not design process, just data transfromation

Discussion

• Are we testing visual representation or are we testing the abstraction?
  • Do we need a cognitive/experimental separation between visual representation and statistical representation?
  • Can we even separate out the statistical property and the visualization? It’s a hard problem.
• Involves both a cognitive component and a perceptual component
  • How do we tease them apart?
• [Bill] You can’t tease apart the questions of what parameterization you are going to use to represent a category and what graphic(s) you will use to present that info
• [Sarah] Is uncertainty special or just another characteristic of the data?
• Fundemental issue: does one proceed the other? (characterization vs visualization)
  • Can we assert that considerations about uncertainty come prior to the visual encoding step?
• Even if testing not separable, framework for (reasoning) thinking about design process still helpful
  • Framework still needed in order to move to higher dimensions.
• Abstraction hierarchy is important
  • need to be saying something specific about visualization