DESIGNS THAT FIT - Human-Computer Interaction and Management Information Systems: Foundations - page 196

frustrations and are rare. In general, physiological indicators can gauge less extreme emotional
reactions to affective fit. Moreover, in comparison with psychological indicators (e.g., question-
naire or observations), physiological indicators are less susceptible to misinterpretation (Wilson
and Sasse, 2004). Examples of physiological measures include pupil size variation, electric brain
potentials, heart rate and blood volume pulse, and facial muscles (Partala et al., 2003). Nevertheless,
the measurement of affect, especially non-intrusive measurement, as a basis for studying and achiev-
ing fit by adaptive designs remains a major challenge (Cockton, 2004).
In sum, affective fit is identified as an unexplored territory, but one that may gain priority as
research in the area of affect in HCI matures. Although there is no theory, or evidence at hand, it
would appear that affective fit is a topic worthy of research. Drawing on the discussions of the
impact of physical and cognitive fit, one can speculate that affective fit will impact the user's well-
being and attitudes, and perhaps performance, too. The notion of negative energy generated in sit-
uations of frustration or the effort required to overcome negative emotions may lead to analyses
of design that minimize such emotional effort. Such analyses would parallel the notions of mini-
mizing physical and cognitive effort. Affective fit may therefore be an important complement to
physical and cognitive fit in order to provide a holistic perspective of HCI.
When Good Fit Is Bad
Davern (1996) provides two important extensions to the study of cognitive fit. First, he shifts the
emphasis from the state of fit (or misfit) resulting from the interaction between representations to
the process of fitting the system and its actual use to the task demands. In other words, fit is not
only the engineered solution offered to match the task demands but fit should also include the
user's unintended (by the designer) appropriations of the system (DeSanctis and Poole, 1994).
Returning to communication metaphors, consider the following scenario that includes both an
initial design as well as the user's adaptive behavior. Suppose there are theoretical grounds to sug-
gest that rich media (such as video conferencing) best fit equivocal tasks, whilst lean media (such
as e-mail) best fit unequivocal and routine tasks. Research has shown that fit results in higher task
performance than situations of no fit. The dynamic treatment of fit requires the researcher (or
designer) to further consider the users' adaptive behavior when they are faced with poor fit. For
instance, users faced with lean media nevertheless adapt and find ways to overcome these limita-
tions, e.g., by requiring more feedback (Te'eni, 2001). In this case, while task accuracy may not
differ between rich and lean media, the user may invest more time and effort when using e-mail
(poor fit) than when using video conferencing (good fit). Moreover, in situations of poor fit where
the user faced with equivocal tasks is obliged to use lean media, designs that support feedback
will be more effective. Thus, incorporating the user's reactions into the fit concept provides a
more dynamic and more complete explanation of the user's task behavior and performance. This
shift can therefore lead to a more valuable research framework of fit.
Davern's second contribution is to disentangle fit into four types (the names in parentheses are
his): a) the correspondence between the computer representation and the real-world task demands
(“representational fit”), b) the correspondence between the computer representation and the user's
characteristics and mental model (“informational fit” 1 ), c) the correspondence between the user's
mental model and the real world (“reality fit”), and d) the correspondence between the user's men-
tal model and the computer tools for manipulating representations (“tool fit”). 2
This classification of types of fit strengthens the value of the fit concept both theoretically and
practically. Theoretically, even though one or two types of fit may be high, performance may decrease
because reality fit or representational fit is poor (Goodhue, in this volume, makes a similar case).