Page 55 -

40 BERENTE AND LYYTINEN
shared by designers and other stakeholders during the design (see Table 4.1). These representa-
tions include instances of the executable code.
To understand cognitive iteration, it is important to explore how the minds of designers work.
This task is not unproblematic due to the intangibility and nonobservability of cognitive activity.
Representational artifacts, however, are tangible objects representing something about the design,
and can be identified, discussed, and tracked in a relatively straightforward manner. Below, we
analyze theoretical views of cognitive iteration in design, and then examine how cognitive pro-
cesses are reflected in changes in representational artifacts.

COGNITIVE ITERATION IN DESIGN

In a sense, all systems analysis and design depends on what goes on in the heads of designers. It
is a commonly held belief that this cognitive activity advances iteratively, where some forms of
mental looping take place to guide the design. A substantiation of this simple observation, beyond a
mere statement, demands, however, that we open ourselves to the vast cognitive science literature,
as well as to the wide array of treatments of cognitive phenomena in psychology, design, computer
science, and information systems research—complete with accompanying rival epistemologies
and ontological assumptions. Rather than attempting in this chapter to establish any distinct onto-
logical stance, we broadly review what we characterize as the “rationalistic” school of cognition.
We also address an alternative tradition as represented in critiques of artificial intelligence and
ethnographic analyses of design. We then offer examples of these two traditions in their treatment
of cognitive iteration in software design. The goal of this section is thus to illustrate the common
thread of cognitive iteration that permeates all perspectives on systems design, and to highlight the
importance of representational artifacts in iteration from an individual designer’s standpoint.

Views of Designer Cognition

The mainstream view of designer’s cognition falls squarely within what computer scientists refer
to as the “symbol system hypothesis” of cognition (Newell and Simon, 1976). This hypothesis
claims that cognitive activity is essentially comprised of “patterns and processes, the latter being
capable of producing, modifying, and destroying the former. The most important property of these
patterns is that they designate objects, processes, or other patterns, and that, when they designate
processes, they can be interpreted” (ibid., p. 125).
Two concepts that are associated with designer’s cognition in this view are: abductive reasoning
(Peirce, 1992) and mental models (Johnson-Laird, 1980). The reasoning process of a designer is
described as abductive (or retroductive) inference, which is different from and should be contrasted
with inductive and deductive inference, which are well-known modes of inference in scientific
study (Peirce, 1992). Abduction generates a design hypothesis (a mapping between a problem
space and a solutions space), often a “guess” by the designer in the face of an uncertain situation,
to a given problem and then works with this hypothesis until it is no longer deemed workable—at
which time another hypothesis is generated. Simon (1996) describes this form of cognitive activ-
ity as nested “generate-test cycles” and argues that they are fundamental to design. He conceives
of design as problem solving, where designers engage in a “heuristic search” for alternatives and
then choose a satisficing design to go forward. When the alternative is found not to be the proper
course, a new cycle of heuristic search begins. During design activity, designers engage in iterative
learning about both the problem space and the solution space (Cross, 1989; Simon, 1996).
Another critical aspect of a designer’s cognition involves the mental models that represent