Levels of Task Abstraction - Clarifying Terms
Task abstraction is an important concept - but a tricky one to define. Abstraction can mean many things when it comes to task. I realize that in class I have given multiple definitions - and here I want to clarify a bit. I have been using the term “abstraction” when discussing tasks in two different ways. Here I’ll review those two different senses of abstraction that I described in class (they come from the Task Cubes paper that was mentioned in lecture and was optional reading).
Level of Abstraction can refer to either concreteness (vs. generic) or composition (small step vs. big multi-step project). I sometimes used abstraction to refer to both, but the Task Cubes paper uses abstraction to refer to concreteness. For you (student), understanding these two concepts is important. For me (professor/writer) I should use my terms consistently. This came out in the Content Survey where people had different notions of what abstraction was.
The Concepts
When we describe tasks, we can make our descriptions vary along two axes:
Concrete vs. Generic - How does specific is the description about the meaning of the data. For example, something might just generically refer to a number (e.g. “interval sequential variable”) or a concrete use of the number (e.g., a grade for this class on the 0-100 scale).
Composed (or complex or high-level) vs. basic (or simple or low level) - How “big” is the task. Is it a simple “atomic operation” or a complex multi-step process. Reading a number is a basic/simple operation whereas determining which forecast is more believable exemplifies a more complex/composed task. Sometimes this is refered to as low-level vs. high-level, but that terminology can be ambiguous.
Note that these describe two different ways that a task description can be more or less abstract. A task description can be very abstract in one way, and very not-abstract in the other.
Here’s an example:
Concrete, Composed: Decide whether to bring my blue rain jacket for a trip to Chicago tomorrow (Oct 16th as I write this) given the weather forecast map.
Generic, Composed: Decide about what to bring on a trip by interpreting a weather forecast map.
Concrete, Basic: Read the weather map to get the predicted temperature in the Oak Park neighborhood.
Generic, Basic: Read a specific value from a color-encoded (chloropleth) map.
This idea that there are multiple ways to think about task abstraction, and that they are orthogonal, was a key insight of a paper called “Task Cube” that I really like.
- (optional - but recommended) Alexander Rind, Wolfgang Aigner, Markus Wagner, Silvia Miksch, and Tim Lammarsch. Task Cube: A three-dimensional conceptual space of user tasks in visualization design and evaluation. Information Visualization 15(4) (October 2016), 288–300. (doi) (web pdf)
In the paper, they use the term “high-level” and “low-level” to refer to the level of composition. They use term “abstract” as I have used the term “generic”. So, “level of abstraction” for them means “amount of concreteness” - an abstract task is a generic one, a low-level of abstraction is a concrete one.
Of course, a key point of the paper is that across the visualization literature we don’t use the term “abstraction” for task abstraction consistently.
I am certainly guilty of using the term “level of abstraction” to refer to either the amount of genericness or amount of composition. And sometimes I use it to refer to both.
In the future, I will try to be more clear and use “level of composition” and “level of genericness” (since level of abstraction is ambiguous). Unless I am intentionally trying to bring both concepts together, or believe that which one I mean can be deduced by context.
Why Task Cube?
The “Task Cube” is a cube, because they define a third axis perspective that is orthogonal to the other two (genericness and composition). Perspective distinguishes whether a task description is about “why” and “how” - it is a little less clean than the other axes, so I won’t describe it here. Read the paper if you want to know about it.