THE THEORY OF COGNITIVE FIT - Human-Computer Interaction and Management Information Systems: Foundations - page 153

entity-relationship diagram-ERD). High and low congruence was operationalized via the complexity
involved in obtaining the required information from the ERD. In situations of high congruence, the
information requested could be obtained by simple extraction from the ERD. In situations of low con-
gruence, a number of transformations were required to obtain the required information. Such tasks
involved, for example, creating temporary views, developing new relationships, or performing outer
joins. Hence, the high congruency/low task complexity condition was matched to the presentation of
the data in the form of the ERD, while the low congruency/high task complexity condition was mis-
matched with the data presentation. These results support the theory of cognitive fit for close versus
far matches. In the authors' terms, congruency between the information request and the data repre-
sentation leads to better performance than when the data representation and the information request
are not congruent.
The second study, by Dunn and Grabski (2001), is also on database querying. The researchers
conducted an experiment to determine the effects of varying levels of match between the problem
(i.e., data) representation and the problem-solving task. The degree of match was operationalized
as the extent of localization of the data, as presented in each of the accounting models used as the
basis for data presentation. These models are the traditional DCA (debit-credit-account) approach,
in which data is presented in list format—that is, the data are symbolic in nature—and the REA
(resource-event-agent) accounting model, in which data are presented graphically. Supporting doc-
umentation for the DCA model adds details that cannot be represented within the model itself.
Individual REA models are presented “in parts based on business processes” in the area under con-
sideration. Use of this latter type of representation implies, therefore, that users may need to refer
to many such diagrams in order to respond to a particular information request.
From the perspective of cognitive fit, the objective of the study was to assess the efficacy of
varying degrees of match afforded by varying levels of localization of the required data in the
REA model compared with the unlocalized data presented in the DCA model. The researchers
proposed that the graphically oriented REA model would be more effective than the DCA model.
They conducted a study that varied accounting model and problem-solving task, as well as expe-
rience. The participants were requested to explain how they would obtain an answer to a particu-
lar task (information request) from the available representations (accounting models).
In terms of accuracy, the researchers found support for the notion of fit in both the strong and mod-
erate localization conditions (participants using the REA model outperformed those using the DCA
model), with no difference between the two models in the no localization condition. Interestingly,
REA model users took longer to complete the tasks than did DCA model users. The researchers sug-
gest that this finding could have been due to the fact that, because the REA model presented more
information, the participants may have suffered from information overload. Recall, however, that par-
ticipants using the DCA model were required to refer to further documentation used to support that
model, information that was not available directly from the chart of accounts; if participants did not
do this sufficiently well, then a time/accuracy trade-off would occur and participants using the DCA
model would be expected to be less accurate and to spend less time on solving the problems. Once
again, we see that accuracy and time should be assessed simultaneously to detect the possibility of
such effects.
Evaluation of Cognitive Fit Based on the Extent of Fit Between Problem-Solving
Task and Problem Representation
These two experiments address somewhat similar dimensions of cognitive fit, termed “close and
far” matches (Borthick et al., 2001), and “localization” of the data in the problem representation