Page 383 -

13.2 Eye tracking 373

in virtual environments were looking (Steptoe et al., 2008). Other studies have used
eye-gaze history data to help users monitor semiautonomous agents, using visual cues
from prior gaze information to highlight where users should look for a monitoring
task (Taylor, 2015). Eye-tracking systems have also been developed for GUI interface
control including pointing and clicking (Kumar et al., 2007), window selection (Fono
and Vertegaal, 2005), multimodal interfaces (Stellmach and Dachselt, 2013; Pfeuffer
et al., 2016), and for remote collaboration (Higuch et al., 2016).
Researchers have used eye tracking to study user behavior with a wide range
of computer interfaces. Web browsing and navigation have been particularly well-
studied in this regard. In a pair of studies, researchers at Microsoft used an eye-
tracking system to examine the impact of factors such as the placement of a target
link in a list of results and the length of the contextual text snippet that accompanies
the results (Cutrell and Guan, 2007; Guan and Cutrell, 2007). In study of placement,
users were observed to be more likely to look at links early in a list than later and to
spend more time looking at the earlier links (Guan and Cutrell, 2007). Consideration
of the length of text summaries led to interesting results: when looking for a specific
link, users tended to focus on more search results as the summaries got longer. This
effect was less notable for open-ended “informational” tasks that were not focused
on a specific goal. The researcher speculated that this difference was due to the rele-
vance of the summaries in each case: summaries that were useful in the informational
task were distractions that obscured the specific link name in the other tasks (Cutrell
and Guan, 2007). Other studies have examined patterns in eye movements as us-
ers interact with websites, moving both within individual pages and across multiple
pages (Card et al., 2001; Goldberg et al., 2002; Buscher et al., 2009).
Other experiments have used eye tracking to understand the progression of eye
focus during menu selection tasks. One study found that eye-focus patterns in tasks
involving reading menu items differed significantly from selecting items. Although
users fixated on each item when reading menus, they tended to use sequences of eye
movements in a given direction—known as “sweeps”—when performing selection
tasks (Aaltonen et al., 1998). Eye tracking has also been used to study differences
in how user attention differs for alternative visualizations of hierarchical structures
(Pirolli et al., 2000), and to build document summaries based on eye-gaze data de-
scribing areas that were the focus of user attention (Xu et al., 2009).
Given the complexity of eye tracking, some researchers might be tempted to look
for other measurements that might provide hints as to where a user's attention is
focused. For GUI-based systems, mouse position and movement might be seen as a
proxy for eye gaze, as we might tend to look where the pointer goes as we move the
mouse. A strong correlation between mouse movement and gaze might completely
eliminate the need for eye tracking in some GUI-based contexts. Alas, the reality is
somewhat more complicated. A number of studies have attempted to track the rela-
tionship between gaze and mouse movement, developing algorithms for using mouse
position to predict gaze (Chen et al., 2001; Bieg et al., 2010; Huang et al., 2012; Diaz
et al., 2013; Navalpakkam et al., 2013), although the nature of the relationship might
be somewhat task dependent (Liebling and Dumais, 2014).