THE THEORY OF COGNITIVE FIT - Human-Computer Interaction and Management Information Systems: Foundations - page 126

are worse than others on both error and effort (Payne, Bettman, and Johnson, 1988). Context vari-
ables influence accuracy (Johnson and Payne, 1985).
Decision makers change strategy primarily in response to changes in task variables, which, as
we have seen above, reduce effort. Numerous studies report strategy changes as the number of
alternatives involved in choice decisions increases (see, for example, Payne, 1976; Olshavsky,
1979). Payne et al. (1988) have also demonstrated that strategy may change due to time pressure.
Note, however, that cognitively simpler strategies can be almost as effective as the optimal strate-
gies in many decision environments. Russo and Dosher (1983), for example, found that subjects
in one experiment traded off a slight loss of accuracy (20 percent) for a substantial increase in
speed (approximately seven times faster).
Studies reporting strategy changes due to context variables, and therefore a desire for accuracy,
are much rarer. As we have seen above, the relative accuracy of a strategy varies across contexts,
although the relative effort does not. Similar levels of accuracy can, however, be achieved across dif-
ferent contexts, although the effort required will vary. Accuracy considerations may therefore also
lead to strategy shifts. Nonetheless, Todd and Benhasat (1991), in a study of decision aids for choice
tasks, suggest that decision makers use aids only to reduce effort and not to improve accuracy.
Problem representation is one of the task characteristics that influences strategy shift. Cost-
benefit theory can therefore be used to examine the effect of graphs and tables on decision-making
performance. Certain other task effects, notably task complexity and time pressure, also appear to
be particularly relevant to decision making using graphs and tables; factors such as these are also
examined here.
Characteristics of Graph/Table Decision Making
To apply cost-benefit theory to decision making using graphs and tables, we first need to deter-
mine the characteristics of both the problem representations and the tasks they support. We view
problem representation and decision-making task as independent, for two reasons. First, data and
task can be presented independently. Second, solutions to the types of tasks examined here can be
reached when the data are represented in either graphical or tabular format.
Characteristics of the Problem Representations
Let us assume we are considering graphs and tables derived from equivalent data, so that all infor-
mation in one is inferable from the other, with a different type of information predominating in
each. Our intuition quickly lets us see a meaningful distinction between graphs and tables similar
to that characterized in the psychology literature as images and words (see Bettman and Zins,
1979; Paivio, 1971, 1978). We use the terms “spatial” and “symbolic” to characterize the differ-
ences between graphs and tables.
Graphs are spatial problem representations because they present spatially related information;
that is, they emphasize relationships in the data. They do not present discrete data values directly.
Spatial representations facilitate viewing the information they contain at a glance without address-
ing the elements separately or analytically; that is, the data in a graph are accessed using percep-
tual processes.
Tables are symbolic problem representations because they present symbolic information; that
is, they emphasize discrete data values. They do not present data relationships directly. Symbolic
representations, on the other hand, facilitate extracting specific data values; that is, the data in a
table are accessed using analytical processes.