DESIGNS THAT FIT - Human-Computer Interaction and Management Information Systems: Foundations - page 195

performance and machine performance. The intersection of both dimensions creates areas that
reflect poor or good performance by each component and areas in which both or neither performs
well. Tasks that fall into areas of distinct advantage to one component can be allocated directly. Tasks
that fall into areas of ambiguity will be left to the designer's judgment or allocated contingently
according to prevailing conditions. Such an allocation of tasks is also important in the design of flex-
ible systems that allow dynamic adaptation according to changing conditions. For example, the lay-
out of the screen may be adapted according to users' preferences that are learned gradually by the
system. Initially, layout decisions are performed by the human, and then, if some pattern of behavior
is detected, the task is performed automatically unless the user indicates otherwise.
Design for fit, therefore, can be seen to begin with an understanding of user and computer attri-
butes, which in turn determine relative performance capabilities and limitations. As suggested above,
the same line of thought could be applied for all dimensions of fit. To date, however, the “man is
better than machine” analyses have concentrated on cognitive aspects and to some extent on phys-
ical aspects, especially in the design of robots. One of the practical contributions implied by the
fit approach concerns, therefore, the starting point of HCI development; it suggest that the first
stages of development should include a comparative analysis of human and computer attributes
(Te'eni et al., 2006).
Affective Fit
Affect in HCI design is rapidly becoming an integral part of HCI research and teaching, in com-
parison to the past preoccupation with cognition in HCI (Picard and Klein, 2002). HCI researchers
are now adopting a more balanced and integrated view of HCI in which both affective and cogni-
tive aspects play a role in understanding user behavior and designing appropriately. A good exam-
ple is Donald Norman's new emphasis on emotions in computing (Norman, 2003). As with physical
and cognitive aspects of fit, I speculate (there is no evidence yet) that it is beneficial to fit the
human-computer interaction to affective characteristics of the user or even of the task. Examples 2
and 6 in Table 10.1 follow from findings in human-human communication, suggesting that the
computer system be designed to be sensitive to emotional aspects of the task (e.g., personal infor-
mation that is potentially embarrassing) and the user's mood (e.g., bright colors may clash with
an unhappy mood).
Affective fit may be seen as the design of the computer to fit the affective state that a user feels or
would like to feel, or to fit the affective state most appropriate for the action the user needs to take.
A consequence of poor affective fit is a negative feeling targeted at the computer, e.g., frustration
and annoyance. Letting the user choose an emotion that best fits her feelings is a trivial example
but it stresses the very common tendency to fit the icon or graphic in the message to match the
receiver's and sender's moods. Projecting again from the practice of effective human-human com-
munication to HCI, the system's appearance can be designed to exhibit empathy. For example, a
software agent or personalized assistant could be designed to display either a smiling face or a per-
plexed face, depending on the user's mood. Inappropriately projecting a smiling face provokes nega-
tive reactions from the user. Affective fit can therefore also have practical value in improving design.
The measurement of affective characteristics, of affective fit and of the impact of affective fit
is particularly difficult. Thus, measurement difficulties hinder both the study of affective fit and the
practical possibility of gauging the emotion to be able to adapt to it. In particular, it is difficult to
measure reactions to affective fit (or misfit), whether through psychological indicators, or obser-
vations of overt behavior. Some overt expressions of emotional reactions, such as swearing at the
computer or hitting the display, are clearly evident to an observer, but they usually reflect extreme