DESIGNS THAT FIT - Human-Computer Interaction and Management Information Systems: Foundations - page 194

rather than low, cognitive fit. The explanatory power, as well as the generality of this approach seems
higher than the approach taken in the empirical work on cognitive fit. For example, Payne et al. (1988)
used it to explain how decision behavior adapts to changes in information displays in order to mini-
mize effort.
A classification of representations relevant to HCI advanced by Norman (1991) helps position
cognitive fit in our discussion. First, there is the represented world of real objects and task demands.
The represented world may be documented in various forms of public knowledge such as topics,
organizational memory, and culture. Second, there is the user's mental model of the represented
world, that is, the representation of the task in its environment as the user sees it. This represen-
tation can be seen as part of the individual's cognitive characteristics. The third representation is
the internal representation of the real world stored in the computer and the fourth representation
is the representation displayed by the computer to the user. The last two representations would be
part of the Computer in Figure 10.1. Cognitive fit theory refers to the match between the charac-
teristics of the problem-solving strategy included in the user's mental model and the problem rep-
resentation as it appears on the system display, e.g., a spatial task with a graphical display.
In sum, the cognitive fit theory has theoretical and practical value. Theoretically it can explain
and predict performance, concentrating on the task level. Similar to physical fit, cognitive fit (nat-
uralness) strives to minimize the cognitive effort needed to transform representations. In this
sense, designs that fit are also those that impose minimal effort. Finally, cognitive fit rests on con-
vincing theoretical grounds. Unfortunately, its potential is currently limited by measurement prob-
lems. Practically, cognitive fit calls attention to specific factors that should be taken into account
when designing displays and can sometimes dictate the optimal designs. In particular, the idea of
characterizing the task as symbolic or spatial and providing the most effective display (e.g., graph-
ical or tabular) by fitting display to task is feasible and would appear to be important. This may
be especially true in the organizational context that relies heavily on appropriate presentations for
decision making. For cognitive fit to be of greater value it must, however, be generalized to enable
analysis of tasks that relate to other cognitive aspects beyond perceptual versus analytic processes.
Finally, as noted above, providing a sufficiently detailed analysis of the task is prerequisite to any
computation of fit, but it must be complemented with a corresponding characterization of cogni-
tive attributes such as memory, attention, and processing limitations.
SOME EMPHASES, EXTENSIONS, AND CONCERNS
Fit and the Allocation of Tasks
As noted above, Baecker et al. (1995) base their discussion of fit on the assumption that the computer
and the user perform the tasks that best suit their respective capabilities. Early work on man-machine
systems (during the 1960s and 1970s) adopted the systemic perspective in which a task would be
allocated to the subsystem that could perform the task in the most effective and efficient way. In
one of the earliest papers on task allocation, Fitts (1951) proposed a list of functions in which
humans outperformed machines and vice versa. This list—and similar ones that emerged over
time—e.g., Sanders and McCormick (1993)—could be used to allocate functions in the design of
man-machine systems. For example, machine is better than man in computing arithmetic prob-
lems but man is better than machine in comprehending text. Advances in technology of course affect
such lists; for example, computers may be able to analyze text in foreign languages better than
humans. However, the basic principle of allocating tasks is one that should be part of the design.
To facilitate this principle, Price (1985) built a decision matrix with two dimensions: human