Page 373 -

12.9 Summary 363

do not generate server requests, additional data recording measures may be
necessary. What other challenges or opportunities can you see in such dynamic
applications? Pick an online word processor, spreadsheet, or presentation tool:
how would you design a system to study its usability?
2. A hybrid system for automated computer data collection might involve a
combination of web server logs—ideally from a proxy that would track all of
a participant's interactions—and one or more software packages instrumented
to collect data of interest. What would be the pros and cons of such a system
relative to a full-scale activity logger that would track all user interactions?

3. Legitimate concerns about user privacy have led some researchers to be very
cautious about the data that they collect with keyboard or activity loggers. This
appropriate concern for user privacy does not come without a cost: in throwing
away details such as document titles, destination addresses for emails, and
specifics of visited web pages, researchers lose information that might have been
used to develop a more nuanced understanding of the underlying activity. For
example, was the user sending email to colleagues at work, or at home? Can
you think of ways to configure logging software to collect certain attributes of
document titles, email headers, and related information in a manner that might
prove useful for research purposes while still being respectful of user privacy?
What effect might greater notification—perhaps telling participants that you
might record sensitive information—have on your experiments?

4. The Fitts' Law, Children, and Mouse Control sidebar provides an example of
experiments that used mouse motion data to study how children differed from
adults in their use of mice. Although 3- to 5-year-old children do not make much
of use keyboards, slightly older children might begin to type. How would you
study differences between children and adults in terms of their use of keyboards?
What sort of data would you collect and how would you interpret it? How would
this differ if you were considering smaller keyboards such as those used on some
cell phones?

RESEARCH DESIGN EXERCISES

1. Experiment with web server logs and log analysis on your desktop.
(a) Start by getting or generating a web server log file. You might ask
computing support people in your school or company for some web log
data. Log files can be very large: you probably only want a small snapshot.
If your school or department gets a good deal of web traffic, you should
be able to get a few megabytes of log data. Be forewarned, some network
administrators may not like the idea of handing out this information. You
may have to convince them that you will use it responsibly. Alternatively,
you can install a web server and run it. If your computer does not have a