Affording Design, Affording Redesign - Principia Designae: Pre-Design, Design, and Post-Design - page 92

7.2.7 Sketches for Design: Promoting Inference
and Discovery
So far, we have examined how sketches represent thought and how they convey
thought. Each is a step in creative thinking. But creativity also depends on going
beyond what is in a sketch, in making inferences, seeing implications, altering and
recombining elements, and more. Going beyond the information in a sketch is key,
and expertise helps. There are numerous examples. The addition of arrows to
diagrams of mechanical systems, such as a car brake or bicycle pump, turns
interpretations from structural to functional. Without an arrow, people describe
the spatial relations among the parts. With arrows, people describe the causal action
of the system from start to finish (Heiser and Tversky 2006 ). Data presented as bars
induces inferences of discrete comparisons whereas data presented as lines induces
inferences of trends (Zacks and Tversky 1999 ) . The students in the class in design
of information systems who drew correct sketches also made correct inferences
from their sketches (Nickerson et al. 2008 ) . Getting the sketch right gets the
thinking right.
This especially holds in design, where designers use successive sketches to hold
design conversations (Goldschmidt 1991 , 1994 ; Schon 1983 ). Designers put skel-
etal ideas on paper, and then reflect on what they've drawn. Sometimes they see
new things in their diagrams, things they might not have intended. Visual patterns
repeated, or overall configuration; these are perceptual inferences. Designers also
see, make inferences about, functional aspects of what they are designing. They
may see the flow of traffic or the light changing throughout the day or seasons. They
may see bottlenecks or unused open spaces. These processes can be studied using
protocol analyses. As they design, designers report their thoughts out loud; their
thoughts are categorized and coded, and the temporal progression and interlinking
of thoughts are analyzed (Goldschmidt 1991 , 1994 ) .
In our own work, we asked expert and novice architects to design a museum
under certain constraints (Suwa and Tversky 1997 ). They worked (with pleasure,
for the most part) for about an hour, usually producing a series of sketches,
gradually refining their designs. Afterwards, they were shown a videotape of their
sessions and asked to report why they made every pencil stroke. This retrospective
protocol analysis allows the designers to focus completely on the design task,
without interrupting themselves, a more natural situation. However, it runs the
risk of forgetting and reinterpretation on the part of the designers. Nevertheless,
the protocols have indeed provided valuable data.
In an early analysis, the information in the protocols was categorized by topic:
emergent properties such as shape and size; spatial relations such as “connected” or
“near” or “configuration;” functional properties such as views or circulation; and
conceptual knowledge. Typically, there were longs strings of comments on one of
these topics, and then a shift of focus. The architecture students were not consistent
in what attributes drove the shifts of focus or the continuing strings. In contrast, for
the architects more than the students, focus shifts were characterized by both spatial