Information Technology Reference
In-Depth Information
1994). This notion is supported in two further studies. The experiment by Speier et al. (2003)
demonstrated that the increased complexity induced by interruptions in the work environment led
problem solvers to change their problem-solving strategy on symbolic tasks from analytical
processes (best supported by tables) to perceptual processes (best supported by graphs). In the
experiment by Wheeler and Jones (2003), the solution of a symbolic task (both accuracy and
time) was supported by a symbolic representation when there were few data points, and by a
spatial representation when there were numerous data points.
A number of studies addressed new tasks in new domains. These were characterized as multi-
attribute judgment tasks, in the accounting domain, and multi-criteria decision-making tasks in
map-related domains. In general, these are tasks that are best addressed using analytical processes
and holistic processes, respectively. In both of these sub-categories, the findings largely support
the theory. Dennis and Carte (1998) call into question the formulation of the time component for
cognitive fit, stating that graphs always result in quicker problem solving than tables. However,
their arguments are based on the simple form of cognitive fit, which does not apply to the types
of more complex tasks that they are addressing. In the intermediate stages of complexity (i.e.,
prior to the level at which decision makers can no longer handle the problem using analytical
processes), accuracy may still be better with symbolic representations while speed may be faster
with maps (i.e., spatial representations), as they found in their study. Hence it appears that the the-
ory of cognitive fit applies equally well to more complex tasks of this nature.
Finally, we examined studies that used new/different dimensions of fit in applying the theory.
The notion of fit occurring in studies addressed to date in this discussion is based on the match-
mismatch between the problem representation and the task. Three studies examined in this cate-
gory demonstrated the traditional match-mismatch form of fit, using new dimensions of fit
(Beckman, 2002; Hubona et al., 1998; Khatri et al., 2006). In the two other studies in this cate-
gory, fit is based on the degree of complexity in the relationship between the task and the problem
representation (Borthick et al., 2001; Dunn and Grabski, 2001). Although this could be construed
as task complexity, such a conceptualization would be somewhat misleading. In these tasks, the
choice of problem representation is contextual in nature in that it relates to the task in the context
of the problem representation used to support that task. The difference, therefore, is not absolute,
but is, instead, one of degree. From the evidence presented here, the findings with new dimen-
sions of fit appear to be essentially similar to those with traditional forms of fit.
Throughout all of these studies, the trade-off between accuracy and time remains undefined. It
is clear that a number of other variables influence the problem-solving outcomes, such as a require-
ment for accuracy or time. Certain studies that find an effect in time but not accuracy attribute the
findings to spending more time with a particular problem representation to achieve the same level
of accuracy, and vice versa. Certain studies investigate either accuracy or time alone. Further, we
have also seen that other environmental stresses such as interruptions influence the outcome. We
can engage in some prescriptions with regard to the conduct of such studies: (1) Do not forget that
performance instructions can bias the approach that participants take to solving the problem, and
hence influence the findings of any study (see Vessey 1994, and Figure 8.2); (2) both time and
accuracy should be measured so that if the effect does not appear in one of the variables it may be
manifested in the other; not measuring either accuracy or time in these types of studies may be
quite risky; (3) both time and accuracy should be measured simultaneously; (4) the effects of both
the independent and dependent variables should be addressed during analysis—this can be done
using the Messmer-Homans procedure (Vessey, 1987; Dennis and Carte, 1998).
Perhaps the major limitation of this analysis lies in the choice of studies for examination.
Because numerous studies have cited the original papers on cognitive fit, it was essential to devise a
Search WWH ::




Custom Search