WHO IS THE USER? INDIVIDUALS, GROUPS, COMMUNITIES - Human-Computer Interaction and Management Information Systems: Foundations - page 35

structural properties such as size of membership and demographic makeup. Collectives differ along
other important dimensions that may affect their HCI needs, such as their participation patterns,
extent of shared history, degree of social identity, and commonality in mental models. These kinds of
social and cognitive dimensions should be identified by the researcher and carefully assessed since
they are known to be critical to group processes such as information search, knowledge exchange, and
learning (Jones, 1997; Reagans and Zuckerman, 2001). For example, we might anticipate the HCI
needs of looser collectives (those with less frequent interaction) to differ from collectives with
stronger ties (those with more frequent interaction among members). Lack of familiarity among indi-
viduals, unfamiliar language, status differences, physical differences, distinctive thought words, and
emotional disparities can all make information sharing difficult (see Okhuysen and Eisenhardt, 2002;
Sun and Zhang, this volume; Weisband, Schneider, and Connolly, 1995). Accounting for these types
of variation across collectives will help researchers to design interventions and explain differences in
system effects. To date, substantial research attention has been given to the dimensions of size and
member diversity, but many more dimensions that distinguish among collectives are in need of study.
Cross-Level User Behavior
Individual, group, organization, and community are not distinct levels of analysis. Instead, they
are inherently intertwined spheres in which people interact with one another, and distinctions in
the levels can be arbitrary. As examples, the individual user may operate in and out of multiple
group contexts. The groups making up an organization may include members both inside and out-
side of that organization. Communities, such as communities of experts, may be subsets of organi-
zations or operate outside of organizational boundaries altogether. The researcher has the
challenge not only of incorporating multilevel data in studies but also of specifying or bounding
those levels in a meaningful way.
Traditionally, boundary specification was done by researchers based on formal membership,
such as assignment to teams by an employer, or registration of membership by individuals in a pro-
fessional association. Though still relevant, such boundary setting is limiting in a world where
collectives are defined by many attributes other than formal structure and where research concerns
include cognitive and emotional phenomena such as knowledge creation or information acquisition
and transfer—phenomena that can take place across fuzzy boundaries rather than clear-cut ones
(Foreman and Whetten, 2002). Researchers are advised to set levels of analysis using indicators
other than formal structure and to establish boundaries that are meaningful to the phenomena being
studied, even if such boundaries are fuzzy.
Hackman (2003) suggests “bracketing” when studying multiple levels of analysis; this means
including in conceptual and empirical analyses constructs that exist one level lower and one level
higher than those targeted for study. The reductionistic tendency of science may make it easier to
move down rather than up a level; but, as Hackman points out, more insight often lies in higher-
level factors. This is especially true in HCI settings where social context can have powerful impacts
on cognition, affect, and behavior (Benbasat and Lim, 1993; Nardi, 1996; Orlikowski et al., 1995).
Spiraling up to a higher level of analysis may be more important than drilling down since infor-
mation that lies with individual users or groups can take on major impact as it moves up or out to
other organizations and communities where it may be exploited for learning or goal achievement
(see Okhuysen and Eisenhardt, 2002). Once levels of analysis are bounded, the researcher can fol-
low “informed induction” (Hackman, 2003), which is the process of bootstrapping to ever better
explanations of a phenomenon by drawing upon all the information one can capture—qualitative
and archival data as well as quantitative measures.