Page 73 -

58 CHAPTER 3 Experimental design

FACTORIAL DESIGN IN HCI RESEARCH
Factorial design has been commonly adopted in user studies in the HCI field.
For example, Warr et al. (2016) used a 3 × 3 factorial design to investigate the
differences between three window switching methods in a desktop environment.
The between-group factor of the study was the window switching method:
the Cards interface, the Exposé interface, and the Mosaic interface. Three
groups of participants took part in the study, each completing tasks under one of
the assigned window switching conditions. The within-group factor of the study
was the number of open windows on the screen (3, 6, and 9). Under a specific
window switching condition, each participant completed the same number of
trials with 3 open windows, 6 open windows, and 9 open windows, respectively.
Learning and fatigue might occur during the experiment. In order to address
these two factors, participants were given time to practice selecting windows
until they were comfortable with the procedure. The order of the 3, 6, and 9
window conditions was counterbalanced through a Latin Square Design.

3.4.3 INTERACTION EFFECTS

One advantage of a factorial design is that it allows us to study the interaction effects
between two or more independent variables. According to Cozby (1997), an interac-
tion effect can be described as “the differing effect of one independent variable on
the dependent variable, depending on the particular level of another independent
variable.” When a significant interaction exists between independent variables X and
Y, the means of the dependent variable Z would be determined jointly by X and Y.
Let us explain interaction effect through an example. Suppose we are conducting
an experiment that investigates how types of device (mouse and touchscreen) and
experience impact the effectiveness of target selection tasks. Two types of user are
studied: novice users and experienced users. Based on the data collected, we draw a
diagram as shown in Figure 3.6. As you can see, novice users can select targets faster
with a touchscreen than with a mouse. Experienced users can select targets faster
with a mouse than with a touchscreen. The target selection speeds for both the mouse
and the touchscreen increase as the user gains more experience with the device.
However, the increase in speed is much larger for the mouse than for the touchscreen.
It is critical to study interaction effects in HCI studies since performance may be
affected by multiple factors jointly. There are numerous studies that did not identify
any significant effect in individual independent variables but found significant results
in interaction effects.
Interaction effects may have important implications for design. For example, the
interaction effect in Figure 3.6 would suggest that the touchscreen performs better
than the mouse during the initial interaction. But users can make greater progress in
learning the mouse than the touchscreen and eventually achieve higher efficiency
with the mouse. This result may imply that a touchscreen is a more appropriate input