Page 384 -

374 CHAPTER 13 Measuring the human

Of course, studies that use eye tracking, mouse movements, or other measurements
as proxies for attention run into all of the usual problems associated with indirect mea-
surements. Before undertaking such a study, consider triangulation approaches such
as postfact review of screen video with participants, asking them to describe what
they were thinking while they were interacting with the system. This “retrospective
think-aloud” approach (Bowers and Snyder, 1990) might be preferable to real-time
feedback, which might distract users from the task at hand. Interestingly, eye-tracking
analyses have been used to validate retrospective think-aloud (Guan et al., 2006).
Although eye tracking has been successfully used for both top-down, hypothesis-
driven experiments and bottom-up exploratory work (Jacob and Karn, 2003), appro-
priate experimental design may increase the odds of success. Exploratory analysis
offers the possibility of generating novel, unexpected insights, at the potential cost
of open-ended searching for illusive needles in haystacks of data. Hypothesis-driven
experiments constrain the analysis needed, helping avoid fruitless searches down
blind alleys.
A narrow focus can also help simplify exploratory work. A study of the effective-
ness of browser feedback for secure websites used eye tracking to study the use of
security indicators, including the secure web protocol indicator (“https://”), lock or
key icons, and security certificates (Whalen and Inkpen, 2005). Focusing on these
areas, researchers learned that users often looked at the lock icon on the browser
window before or after looking at the HTTPS header in the web location bar. Eye
tracking also identified potential confusion due to browser designs, as some users
looked at the lower left-hand corner of the browser (where the lock is on Netscape/
Mozilla browsers) rather than the lower right-hand corner (where it could be found
on the Internet Explorer browser used in the study) (Whalen and Inkpen, 2005).
Eye tracking can also be a vitally useful tool for understanding complex and cog-
nitively challenging workflows and tasks (see the “Measuring Workload” sidebar),
as demonstrated by explorations of the use of eye tracking in studying electronic
medical records (EMRs) and other clinical information tools. Examples include the
use of eye tracking to improve EMR design, through investigations of the detrimen-
tal impact of layout clutter on task performance (Moacdieh and Sarter, 2015) and
in conjunction with retrospective think-aloud, to understand information search and
access patterns during the use of EMRs (Wright et al., 2013). Other studies have in-
vestigated the use of eye tracking to identify the skill levels of EMR users (Kocejko
et al., 2015); to understand EMR workflow (Doberne et al., 2015; Mazur et al., 2016)
and visual search patterns (Fong et al., 2016); to explore how EMRs are used during
patient visits (Rick et al., 2015) and particularly the impact that they might have on
communication between patients and physicians (Montague and Asan, 2014); and
to examine how emergency physicians interpret test results (Nielsona et al., 2013).
Comparable studies with consumers of health information have examined under-
stand how audiences read and interpret health information, including antialcohol
messages (Brown and Richardson, 2012) and public safety guidelines (Bass et al.,
2016). An eye-tracking study of reading patterns for users of a health discussion
found that gaze patterns differ between user seeking information regarding their own
health, as compared to those seeking for information about someone else's symptoms