Sketches, and Icons, and Features, Oh My!
Rapporteur's Summary: Session 2

Initial Comments

``Somewhere over the rainbow, way up high,
There's a land that I've heard of, once in a lullaby,
Somewhere over the rainbow, skies are blue,
And the dreams that you dare to dream really do come true.''

Many aspects of visual and spatial reasoning in design are just starting to be investigated, and much is still at the ``dare to dream'' stage of development.

Session 2 included dreams by Forbus, Park & Gero, Leclerq, Li and McFadzean & Cross. Their papers covered the topics of qualitative spatial reasoning, a representation for qualitative reasoning about similarity of shape, a tool for interpreting conceptual architectural sketches, properties of shape schema grammars, and the relationship between design events and graphical events while sketching.

Themes

The discussion following the presentation of papers centered mostly around the role of sketching, sketching conventions, icons, and the definition, detection and use of features in spatial reasoning. In the brief summary of the main points of the discussion given below, bullets are used to provide the main points, with text annotations as needed. While the bullets were issues raised during the discussion, the points made in the annotations may not have been.

Sketches:

• What are the roles of a sketch? of sketching?
  How to interpret a sketch depending on role?
  { It's clear that sketches can be used as a memory aid, to evaluate a proposed solution, to deliberately attempt to trigger associations with known previous designs or known problem decompositions, to refine requirements, etc. What other roles are there? Can one interpret a sketch without previously knowing its role? }

• Is there a competence/performance issue wrt sketching?
  { Is there an underlying language of sketching? Can personal variations in sketching, and sketching errors be filtered out to determine the underlying meaning?

• If marks form "words" does personal sketching skill form a dialect?
  { Do different sketchers use different marks? Do different sketchers have different but consistently used mark vocabularies?

• How can we take advantage of sketching conventions?
  { Can conventional sketching `marks' act as islands of certainty when trying to understand a sketch?

• How can we avoid making interpretations too early when representing sketches?
  { Representations of sketches in the computer need to be composed from primitives. These primitives need to have meaning if the representation is to be useful. The mere act of representing a sketch makes a commitment to a particular interpretation of the marks. More interpretation leads to a smaller representation and more direct use of the representation. However, once the sketch is represented in a certain way -- i.e., once a particular interpretation is imposed -- it is hard to interpret it in another way. }

• How does sketching structure relate to the structure of design reasoning?
  { Can the reasoning being carried out during designing be detected by looking at the sketching process? at the sketch produced? For example, do the breaks in the sketching process correspond to consideration of design alternatives, or does it indicate design decomposition? Does the precision of the marks indicate the stage of the design process? (e.g., conceptual) }

• How do sketches affect thinking?
  How does sketching affect thinking?

  • What is the role of the tool?
    e.g., how much is it suited to `providing ambiguity'?
    { Sketching can both help and hinder designing. For example, if the sketching tool isn't capable of producing the kind of ambiguous and suggestive marks that the user wishes to make, then this impedes natural and successful communication. }

  • What is the role of the user's skill?
    • i.e., with the tool.
    • i.e., with the use of sketching conventions.
    { The tool may have all the capabilities required, but the user may be unaware of them, or may not be skilled in their use, leading to stilted communication. }

  • What is the role/status of sketch?
    Is precision required? Is it a `throw-away' sketch?
    { Sketching places demands on the user to supply information that can then be recorded in the sketch. If the sketch is intended to be a precise record of a detailed portion of the design then the user is required to include all the details. Thus sketching may in places `drive' the detailed design process. }

  • How do expert and novice sketchers differ?
    { Do they sketch differently? Can the user's status as an expert or a novice be easily determined from the sketch? Does the act of sketching affect their design thinking in different ways? Do they reason using sketches in different ways? Do they tend to assign different roles to sketches? }

Features:

• What are the features?
  i.e., what regularities are worth noting for use?
  e.g.,
  • patterns of pen pressure over time;
  • indentations in line sequences;
  • "shared space";
  • holes punched in the paper with the pen; etc.
  { Depending on one's point of view, there are many possible things that might be considered a feature. If the goal is to try to detect user frustration, or a poor quality of support surface under the pen, holes punched into the paper (instead of marks) might be appropriate features. }

• As features represent points of view, and have reasoning utility:
  
  • what are the appropriate points of view?
  • what are the corresponding "sketch features"?
  { What points of view are there? Why do we need the features? What features will allow us to reason about a user's goal, level of expertise, design reasoning, or emotional state? }

• Features in drawings and sketches:
  • what to look for (i.e., level of detail)
  • effect of `purpose' of extraction?
  { Features exist at many different levels of detail, and in many different `dimensions'. For example, in time, one might be interested in a pause, in a change of speed during the drawing of a line, or the number of marks per minute. What can different dimensions tell us? What can different levels of detail tell us? }

Icons:

• Continuum of "iconness"
  • abstraction, mapping, ...
  • geometry, topology, color, texture, ...
  • whether drawn to scale or not
  { Icons are intended act as a reminder of the thing they represent, so they have to have the power to remind us of the original. Icons are often abstracted graphical representations of the original. As abstraction is a process, and there are usually many ways to abstract something, and many possible degrees of abstraction, this suggests that there is a continuum of "iconness". How then can one determine whether a graphical entity is an icon or not? }