DESIGNS THAT FIT - Human-Computer Interaction and Management Information Systems: Foundations - page 200

Finally, fit is expected to enhance task performance (with respect to the goals set to or by the
user) and the user's well-being. Although extant research shows only piecemeal relationships
between different types of fit and their impact, future research will need to address the overall
impact of fit. Given this conceptualization (Figure 10.2), I now return to the opening questions
raised in the introduction.
Is fit useful for theory? Only if fit models can explain or predict performance and behavior will
they be useful for research. Research in IS, particularly on cognitive fit, indicates a positive link
between fit and task performance, yet more sophisticated and contingent models are needed to
explain broader and more realistic situations. These advances in theory are also needed to address
claims that fit theories are hardly falsifiable.
So what needs to be done? There is a need to substantiate some of the claims by more direct
observations that would open up the black box of the “fitting” process. The development of valid
and reliable measurement has been identified as a crucial step for further progress in all types of fit.
In addition, broadening the scope of fit is essential. However, generalizing from very specific cog-
nitive characteristics and processes to more general characteristics entails even more extensive
development of measures. Indeed, we need to develop measures of fit at different levels of granu-
larity according to the research question. Furthermore, while research on the link between cognitive
fit and task performance seems promising, research in the IS literature and in some of the HCI jour-
nals on the link between physical fit and well-being is limited. Finally, the link between affective fit
and both performance and well-being should be investigated.
Is there practical value in fit models? We have seen that “good fit” is not synonymous with
“good design.” In some cases it may lead to poor performance; thus, any conceptualization of fit
must state explicitly when designers should and should not strive for fit. Fit has value as a general
approach but also as a basis for more specific guidelines. The most basic contribution of fit is to
enumerate the relevant aspects of fit based on user characteristics as a first step towards achieving
fit. This step involves a task analysis in light of the relevant user's characteristics. It is then fol-
lowed by an informed allocation of tasks to the computer and the user. As noted above, these steps
are not usually included in common HCI methodologies.
A more specific contribution would be to map processes to tasks and displays (e.g., perceptual
processes and graphic displays). Only after we learn more about individual aspects of fit should we
also look at the interactions, for example, physical and cognitive. A third source of value to prac-
tice is the enumeration of types of cognitive fit between representations (see classifications by
Norman and by Davern above). Knowing what to look for is a necessary first step in any practice.
Moreover, understanding the sources of fit or misfit raises the question of whether to adapt the
computer or train the user (or enable learning). Again, an explicit consideration of training is not
usually included in common HCI methodologies. More generally, the view of fit in which adapta-
tion is a two-way process will hopefully promote the practice of training and learning in HCI.
The skepticism about the feasibility of using fit models to enable adaptive systems is reminis-
cent of the cognitive style arguments in the early 1980s (Huber, 1983). One of the basic arguments
for stopping all further research on cognitive style was that it is not feasible to design systems that
could cope with all the possible combinations and variations one could expect in real-life situa-
tions. (Another argument was about poor measurement.) Even if the argument against systems that
are adaptive to cognitive styles is correct, the research brings us closer to understanding what param-
eters should be incorporated into adaptable systems. Clearly, the last two examples in Table 10.1
on dynamic adaptation may be counterproductive in certain unpredicted conditions. Yet taking
them as a starting point for designing adaptive systems is an important practical contribution of fit
models.