DESIGNS THAT FIT - Human-Computer Interaction and Management Information Systems: Foundations - page 197

For example, a user who models a given task as a spatial task interacts with a graphical display
(good informational fit). However, the user has incorrectly modeled the task as a spatial task,
while in fact it is symbolic (poor reality fit). This situation may result in poor performance,
despite high satisfaction with the technology. As Davern notes, when users misperceive the situ-
ation to have a good fit with the task (overlooking, for example, the reality fit), they will have no
incentive to learn and adapt, and as a result, their performance will decline rather than improve
over time. Hence, a comprehensive treatment of fit clarifies why effective HCI needs to incorpo-
rate learning and training.
Practically, designers should be aware of all types of fit and their impact on performance
because different designs are required to establish different types of fit. In particular, the distinc-
tion between informational fit and tool fit is a subtle one that is ignored in cognitive fit theory, in
which the information and the (computerized) tool that manipulates the information are inter-
twined. I return to this point in the conclusion.
Davern's analysis indirectly raises another concern with the theories that build on mental mod-
els. Explaining the user's interaction with computers by relying on the central role of mental mod-
els is only one option, which is appropriate for controlled behavior found in complex or novel
tasks. Another is to regard users' behavior in known situations as primarily perceptual reactions
based on experience (Kahneman and Tversky, 1979). Obviously, a significant portion of human-
computer interaction can be characterized as rote behavior. Don Norman argues that in such
cases, mental models play no role (Parush, 2004). Fit would need to be conceptualized in relation
to habits rather than mental models. Practically, this requires the designer to be sensitive to dif-
ferent modes of behavior (controlled and rote).
Fitting the design to the user's mental model is therefore not always good. In some cases, it may
be better to train the user, that is, change the user's understanding or habits. In complex tasks that
involve controlled behavior such as the interaction with decision support systems, the designer may
attempt to restrict and guide decision behavior (Silver, 1991). In automatic or rote behavior, it is
sometimes necessary to unlearn certain habits in order to retrain users on new devices rather than
fitting the old device. For example, in the past, when experienced users of typewriters were asked
to convert to computer keyboards, they first had to overcome some of their old typing habits (such
as backspacing to underline or using the carriage return to advance a line), which interfered with
working efficiently with computers. Fitting the keyboard to old habits was infeasible.
Davern's extensions to cognitive fit underscore the need to refine the concept of fit. First, striv-
ing for an optimal design that creates a good fit should be regarded only as the beginning of a good
design. HCI design should also include the user's adaptive behavior as circumstances change and
fit declines or rises. Second, the need to consider different types of fit and the interactions between
them further expands the scope of research into fit. Both directions warrant further research.
Dynamic Fit and Personalization
The last two examples in Table 10.1 highlight the possibility of fitting the human-computer interac-
tion to the way the dialog progresses. As in human-human communication, changes occur during a
human-computer dialog, e.g., an unpleasant or incomprehensible message comes up, and both the
computer and the user can react to maintain effective interaction: the computer can display the
material at a slower rate or with more detail, or it can change the tone of voice to comply with a
changing mood. This type of fit requires not only knowledge about the alternative designs and pos-
sible reactions but also a real-time detection and feedback mechanism that provides the necessary
signals to direct and control adaptation. Achieving physical fit in advanced environments will