Interfaces

GUI
Studying Context: A Comparison of Activity Theory, Situated
Action Models, and Distributed Cognition
Bonnie A. Nard i
It has been recognized that system design will b enefit from explicit study of th e con text in wh ich users
work. Th e unaided individual divorced from a social g roup and from supporting artifacts is no longer
the mod el user. But with th is realization about theimportance of context come many d ifficult
questions. What exactly is context? If the ind ividual is no lon ger central, what is the correct un it of
analysis? What are the relations between artifacts, individua ls, and the social groups to which th ey
belong? This chapter compares th ree approaches to the study of context: a ctivity theory, situated
action models, and d istribu ted cognition. Iconsider th e basic concepts each approach promulgates
and evalua te the usefuln ess o f each for the design of technology.1
A broad range of work in psychology ( Leont'ev 1978; Vygotsky 1978; Luria 1979; Scr ibner 1984;
Newman, Gr iffin, and Co le 1989; Norman 1991; Salomon 1993), anthropology (Lave 1988; Suchman
1987; Flor and Hutchins 1991; Hutchins 1991a; Nardi and Miller 1990, 1991; G anttand Nardi 1992;
Chaiklin and Lav e 1993), and compu ter scien ce ( Clement 19 90; Mackay 1990; MacLean et al. 1990) has
shown that it is not possib le to fu lly understand how people learn or work if the un it of study is th e unaided
individual w ith no access to other people or to ar tif acts for accomplishing the task at hand. Thus w e are
motiv ated to study context to understand relationsamong individuals, ar tif acts, and social groups. But as
human-computer inter action research ers, how can w e conduct studies of contex t that w ill h ave v alu e to
designers who seek our expertise?
Brooks (1991) argues that H CI specialists w ill b e mo st v alu able to designers when we can provide
(1) a broad background of compar ative understanding over many domains, (2) high-levelanalyses usefu l
for evalu ating the imp act of major design decisions, and (3) information that suggests actual designs r ath er
than simp ly gen eral design guid elines or metr ics for ev alu ation. To b e ab le to provide su ch exp ertise, w e
must d evelop an appropriate analytical abstraction that ``discards irr elevant details while isola ting and
emphasizing those properties of artifa ctsand situations that are most significant for d esign'' (Brooks, 1991,
emphasis added). It is esp ecially difficult to isolate and emphasize cr itical properties of ar tif acts and
situ ations in studies that consider a fu ll context because the scope of analysis has been wid ened to
acco mmodate such holistic bread th. Tak ing context seriously means f inding oneself in the th ick of th ecomplexities of p articu lar situations at par ticular times with particular individuals. Finding commonalities
across situations is diff icu lt b ecause studies may go off in so many diff erent dir ections, making it
problematic to provide th e compar ative under standing across domains that Brooks (1991) advocates. How
can w e confront th e b looming, buzzing confusion th at is ``co ntext'' andstill produce gener alizable r esearch
results?
This chap ter looks at three approaches to the study of contex t—activity theory, situated action
models, and th e distr ibuted cognition approach —to see wh at tools each offers to help manage th e study of
contex t. In particu lar we look at the unit of analysis proposed by each approach, th e categories offer ed to
support a d escr iption ofcon tex t, the ex ten t to which each tr eats action as structured prior to or during
activ ity, and the stance toward the conceptual equivalence of people and things.
Activity theory, situated action models, and d istr ibuted cognition are evolv ing frameworks and
will change and grow as each is exercised w ith emp irical stu dy. In th is chapter I ask where each approach
seems to be headed...