Page 354 -

344 CHAPTER 12 Automated data collection methods

discussed earlier, will require additional programming and configuration, possibly
including changes to the sites under question. Although potentially labor-intensive,
these changes may provide access to otherwise unavailable data regarding user inter-
actions with your sites.

12.3.2 KEYSTROKE AND ACTIVITY LOGGERS
Modern GUI windowing systems that support multitasking and concurrent use of
multiple related tools present intriguing possibilities for HCI research. How often
do users change applications? How many tasks do people work on at any given
time? How long do users generally work within any given window before switch-
ing to another? What fraction of time is spent on overhead such as resizing win-
dows, moving windows, or adjusting system controls? Answering these and other
related questions requires data collected at the level of the operating environment.
Activity-logging software runs invisibly in the background, recording mouse
movements, keyboard input, and windowing systems’ interactions including win-
dow movement, resizing, opening, and closing. These tools act as proxies for user
interaction events, recording events as they happen, before they are passed along
to applications or the operating environment. Keyloggers are a special subclass of
activity-logging software, focusing only on keyboard input. Activity-logging soft-
ware has achieved a fair amount of notoriety in recent years, as these tools have been
used as “spyware,” to surreptitiously record user interactions in the hopes of stealing
passwords, finding evidence of criminal behavior, or collecting evidence of spousal
infidelity.
Commercial activity-logging products are often marketed as being tools for
employers and parents to track inappropriate computer use by employees and
children, respectively. Although some of these tools might be appropriate for
data collection for research purposes, some antispyware programs may defeat or
remove activity loggers. You may want to test the logging software on relevant
computers and disable antispyware measures before trying to use these tools to
collect data.
The disruption and recovery tracker (DART) (Iqbal and Horvitz, 2007)
logged window positions and sizes, window actions, user activities, and alerts
from various systems. DART's design presents an example of the responsible use
of these tools for conducting legitimate research while remaining sensitive to
privacy concerns. Keyboard logging was limited to a subset of possible choices,
including menu shortcuts and some punctuation, and only a portion of each win-
dow title was collected. The resulting data therefore did not include file names,
email addresses or subject lines, or web page titles. The analysis of more than
2200 hours of activity data collected from the main computers of 27 people over
a two-week period generated numerous insights into time lost due to email or
instant-messaging alerts, and how users respond to and recover from those inter-
ruptions (Iqbal and Horvitz, 2007). Logging studies can also be used to collect
data on effective use of devices, as in a study that captured mouse movements of