Page 425 -

416 CHAPTER 14 Online and ubiquitous HCI research

A/B testing is also limited by the coarse-grain nature of the data. Knowing which
elements are clicked on which pages can be useful, but additional data might be
needed to know where and how those pages command user attention. Eye-tracking
techniques (Chapter 13) and additional software tools such as proxies and JavaScript
libraries (Chapter 12) can provide finer-grain detail when necessary.

14.2.3 ONLINE ACTIVITY

The rich stores of data created through our online lives provide tantalizing HCI re-
search possibilities. Exploration of online content and activity can provide deep insight
into how people communicate, create communities, learn, and interact online, includ-
ing how ideas develop and spread, and what we might learn from information dissemi-
nation patterns. Although the techniques are very similar to others discussed earlier
in this book—including both qualitative content analysis (Chapter 11) and statistical
review of automatically captured interaction data and human physiological signals
(Chapters 12 and 13)—the domain is qualitatively different, in that analysis of online
activity effectively involves the emergence of community and collective behavior.
14.2.3.1 Online communities
Computers have been used for online communities since the early 1980s, with the
early USENET discussion groups on the ARPANET (Leug and Fisher, 2003) leading
to online bulletin boards where home computer users with dial-up modems could
interact. The growth of the Internet in the 1990s led to the emergence of countless
bulletin boards for communities of interest, providing researchers with an opportu-
nity to study communication and patterns, community growth, and related dynamics.
Analyses of these communities often combine qualitative and quantitative meth-
ods. Qualitative methods might include thematic content analysis (see Chapter 11),
aimed at extracting common themes and types of interactions, perhaps guided by
some theory. These studies will typically involve reading through large numbers
of posts, coding contents for types of concerns, types of posts (questions, answers,
guidance, emotional support), and for conversational structure (introduction of
new members, arguments over controversial topics, resolutions of disputes, etc.).
Although time-consuming, these techniques offer the possibility of immersion in
the community under consideration, providing rich context that might enable deep
understanding.
As online community content is often, if not exclusively, found in the form of
online text, it is particularly well suited for automated analysis and quantitative in-
vestigation of patterns of interest, including how and when certain terms or types of
discourse are used. Forum content and posts can generally be downloaded, although
with varying levels of difficulty, depending on the underlying software platform.
Communities built on open platforms might provide programming libraries known
as Application Programming Interfaces (APIs) capable of extracting data. Using
these libraries, software developers might develop custom programs to gather and
collate data needed to address research questions of interest. Barring such facilities,