Page 380 -

370 CHAPTER 13 Measuring the human

This chapter discusses a variety of options, with an eye toward cost-benefit trade-
offs: as some tools are clearly more difficult and expensive than others, we strive to
use the simplest and least expensive tools suitable for a given job.

13.2 EYE TRACKING
13.2.1 BACKGROUND
Countless traditional HCI studies used—and continue to use—measurements of
mouse or keyboard interactions in an attempt to see how users control computers.
This approach can be very useful, but it paints a necessarily incomplete picture, as
simply knowing which keys were pressed and where the mouse was moved does not
help us understand what's going on—where were they looking? Which aspects of the
system drew their attention?
Eye-tracking systems can help us begin to answer these questions. Using
cameras or other sensors, these systems continuously track the orientation of the
fovea—the center of the field of vision. This information can be used to identify
where the user is looking, which is in turn assumed to be the center of their atten-
tion. Although perhaps overly simplistic, this simplified model provides the basis
for all eye-tracking work (Duchowski, 2007). Generally, eye-tracking systems will
use transform raw data regarding gaze direction into a series of coordinates map-
ping direction into (x, y) coordinates on the display being viewed. These coordi-
nates can then be further transformed into trails identifying where the user looked
and when (Figure 13.1), providing information that can help us understand how
user attention relates to task completion, and possibly how aspects of the interface
command attention and influence whether or not tasks are completed successfully
and how long they take.
Technologies and applications have progressed significantly since the first use
of eye tracking in the early 20th century (Jacob and Karn, 2003). Modern systems
use sensors based on the desktop or on head-mounted devices to track the reflection
of infrared light from the cornea or retina (Jacob and Karn, 2003; Kumar, 2006).
Eye-tracking devices have become increasingly inexpensive, with highly functional
commercial systems now available for less than $200. Open-source university-
developed systems costing less than $100 have shown performance comparable to
more expensive commercial systems (Agustin et al., 2010; Johansen et al., 2011).
The advent of low-cost cameras and other inexpensive hardware have reduced the
costs of eye trackers (Kumar, 2006), although inexpensive systems may lack collect
data at a lower frequency than higher-end alternatives. Systems are often hard to
use, requiring calibration for each user and inconvenience such as head-mounted
devices or restrictions on the range of movement allowed to the user (Jacob and
Karn, 2003).
Interpretation of eye movements is a nontrivial challenge, due to the constant
motion of our eyes. Rapid motions known as saccades last anywhere from 10 to
100 ms (Duchowski, 2007). These movements are used to reposition the eyes to a