Professional Documents
Culture Documents
Sociotechnical Systems
Victoria L. Mitchell
and
Barrie R. Nault
November 2003
We thank the Social Science and Humanities Research Council of Canada, the Natural Sciences
and Engineering Research Council of Canada, and the Informatics Research Center at the
University of Calgary for generous support.
Copyright © 2003 by Victoria L. Mitchell and Barrie R. Nault. All rights reserved.
The Emergence of Functional Knowledge in Sociotechnical Systems
Abstract
British coal mines, holds that work processes consist of two separate dimensions, one social, the
other technical. The main tenant of STS is the principle of joint optimization – that is, the
dimensions are reciprocally interdependent and must be designed conjointly. Although it has
been intuitively appealing to conceptualize work processes along two dimensions, equivocal
outcomes in STS research suggests there may be additional dimensions. In the context of IT-
intensive reengineering projects, we examine the extent to which the technical dimension, the
social dimension, and their reciprocal interdependence explain variance in project performance.
Project performance is captured with an objective measure of IT-related project delay (IT-delay)
and a subjective measure of client satisfaction. We find that neither the main effects nor the
interaction effect significantly explains IT-delay or client satisfaction. Allowing for additional
that were embedded in the initial technical dimension emerge as a separate dimension. Although
the main effects of the technical, social and functional-knowledge dimensions do not
significantly explain IT-delay or client satisfaction, interactions among the three dimensions
explain a significant proportion of both measures of project performance. As a result, rather than
knowledge is a separate third dimension and the principle of joint optimization must incorporate
1
I. Introduction
Sociotechnical systems (STS) theory was originated by Trist and Bamforth (1951) in a paper on
the effects of mechanization in British coal mines. The authors argued that a work process could
not be seen either as a technical system - plant and machinery, or as a social system - social
relations and work organization, but rather had to be seen in terms of both of these dimensions.
Thus the term “sociotechnical system” was coined to describe a method of viewing organizations
that emphasizes the interrelatedness of the social and technological dimensions. The coal mining
studies suggest that most of that industry’s problems had resulted from the introduction of
significant changes in the technical aspects of production without adequate attention to their
impact on social structure and human requirements (Fox, 1995). The same argument applies
technology (IT) to reengineer their business processes without a clear understanding of the social
implications.
STS theory draws heavily from open systems theory. The term “system” suggests that all
parts of an organization are interrelated, so that the design of one necessarily affects the
operation of another. The “open” perspective implies that the social and technological
dimensions of work processes must be designed not only in relation to each other, but also with
reference to evolving environmental demands. STS reflects the goal of integrating the social
requirements of people doing work with the technical requirements needed to keep the work
processes viable with regard to their environments. These two requirements must be considered
interdependent, because arrangements that are optimal for one dimension may not be optimal for
2
In order to jointly optimize the technical and social dimensions, reengineering projects
require a thorough understanding of the knowledge that underlies both the IT platform and
related work processes. Technical knowledge of IT platform components and how those
components relate form the basis for technological change (Henderson and Clark, 1990).
Collective, objectified knowledge embodied in work activities and how those activities relate
provides a foundation for social change (Polanyi, 1974). Integrating the social and technical
dimensions creates reciprocal interdependencies (Thompson, 1967) that are often idiosyncratic,
and the ensuing functional-knowledge forms the basis of competitive advantage (Barney, 1991).
As functional-knowledge evolves over time, it becomes more extensive and more difficult to
investments in IT capital (the technical dimension) such as hardware and software at the
economy level (Oliner and Sichel, 2000) and at the firm level (Brynjolfson and Hitt, 1996).
Other research, however, found these positive returns only in highly automated sectors such as
IT hardware and other manufacturing that is a small portion – 12 percent - of the economy
(Gordon, 2000). Furthermore, research at the firm level found that possible inefficiency in IT
investment and management may have led to an under-realization of IT returns (Lee and Barua,
1999). These latter studies provide economy- and firm-level evidence that IT investments in the
technical dimension alone are not always successful. Indeed, previous research indicates that the
We test two main tenants of STS in the context of reengineering projects. The first tenant
is that work processes are best represented by two dimensions: a technical system and a social
3
system. The second tenant is the “principle of joint optimization” – because the technical and
social dimension are interdependent, performance can only be optimized by jointly considering
the two dimensions. A summary of our main results is as follows. First, we partition a set of
IT/reengineering issues into constructs representing social and technical dimensions, and find
that neither significantly explains project performance – either an objective measure of IT-
related project delay (IT-delay) or a subjective measure of client satisfaction. More importantly,
the interaction of the social and technical dimensions does not significantly add to the
between the social dimension and technical dimension fails to be significant in determining
dimensions, equivocal outcomes in STS research (Spender, 1996) suggests there may be
we find that a third dimension emerges from our set of IT/reengineering issues – primarily a
subset of those that made up our earlier technical dimension. We label this new construct
“functional-knowledge”, and redefine the separate technical and social dimensions. Although the
main effects of our social, technical and functional-knowledge dimensions do not significantly
explain IT-delays or client satisfaction, when interactions between the dimensions are included –
both two-way interactions and a three-way interaction – these three dimensions significantly
explain IT-delays and client satisfaction. Thus, our fundamental result is that in reengineering
projects STS theory should consider certain forms of knowledge as a separate dimension rather
than embedding all knowledge in the technical and social dimensions. And, joint optimization
4
The remainder of our paper proceeds as follows. The next section presents the theoretical
underpinnings and previous empirical studies of STS in high technology environments, and
the study’s research methodology. We then provide the details of our analysis and results. The
impact of these findings on the principle of joint optimization and the number of dimensions that
The key tenet of STS theory is the principle of joint optimization - an organization can perform
optimally only if the social and technical dimensions are designed to fit the demands of each
other and of the environment (Van De Ven and Joyce, 1981; Pasmore, Francis, Haldeman and
Shani, 1982). Attempts to optimize the technical or social dimension alone will result in the
suboptimization of the socio-technical whole (Trist, 1981). Thus, STS is a method of viewing
organizations that emphasizes the interrelatedness of the functioning of the social and technical
dimensions.
This main principle of STS theory has been extended to technologically intensive
failure of an implementation cannot be explained in terms of the technology alone as the same
organization can experience success with some information technologies and not others. Neither
can the success or failure of an information system be explained in terms of the social dimension
alone as the same technology will exhibit different properties in different organizations.
Therefore, both technical and social dimensions must be considered together (Burkhardt and
5
Brass, 1990). Moreover, a growing body of research suggests that critical problems companies
face are not technical, but managerial. Hence the STS perspective of people (social) along with
techniques and functional-knowledge keeps the focus on management (directing the interaction
Some case studies have encouraged the application of STS theory to IT-based contexts.
(1986) concluded that STS analysis was appropriate for knowledge work and would facilitate
integrated factory, Chase and Susman (1986) advocate the use of STS analysis to understand the
issues that may arise when planning and designing new processes. Bostrom and Heinen (1977)
promote the use of a STS approach as a MIS design methodology, suggesting that its utilization
solves many of the problems associated with new applications of IT, thus substantially reducing
STS theorists and practitioners alike, however, have been concerned with the
applicability of STS to various organizational settings. This issue first surfaced in Norway as
early as 1965, when STS applications were extended to include continuous-process industries
and manufacturing organizations (Thorsrud, Sorensen, and Gustavsen, 1976). Pasmore, Petee
and Bastian (1986) describe a longitudinal study involving two laboratories of a major health
care institution that employed STS analysis to facilitate the introduction of new technologies and
work practices - an intervention that did not achieve the results desired. Pava (1986) studied the
poorly structured problems with erratic inputs and outputs, and identified several conditions that
undermined the use of STS theory in that context. Numerous other researchers have pointed to
6
the omissions and ambiguities in the work of STS theorists (Brown, 1967; van der Zwaan,
1975; Maton, 1988). In his examination of the STS case studies, Kelly (1978) concludes that
the concept of joint optimization has little connection with the reality of STS practice. In their
review of 134 experiments using an STS approach to work redesign, Pasmore et al. (1982) point
out that few STS experiments actually involve technological change, rather, most concentrate on
rearranging the social systems around an existing technical system to approximate joint
optimization.
A common difficulty in studies of STS is that the meanings of “technical system” and
“social system” have not always been well-defined. One of the pioneers of STS, Emery,
suggested that the technical dimension was related to aspects of the natural sciences, whereas the
social dimension included occupational roles and structure, methods of payment, supervisory
relationships, and work culture (Emery, 1959). More recently, Pasmore et al. (1982) defined the
technical dimension as tools, techniques, procedures, skills, knowledge, and devices used by
members of the social dimension. In contrast, the social dimension was defined as people who
work in the organization and the relationships among them. Similarly, Shani and Sena (1994)
viewed the STS perspective as comprised of a social dimension – people, and a technical
dimension of techniques and knowledge. Thus, both Passmore et. al (1982), and Shani and Sena
(1994) consistently place knowledge in the technical system, although we recognize that there
are elements of knowledge, particularly collective and objectified knowledge in aspects of the
social dimension (Griffith, Sawyer and Neale, 2003). Stanfield (1976) described technology as a
assumption in STS. Indeed, he argued that STS studies varied greatly in where they draw the line
7
Knowledge and IT-Intensive Reengineering Projects
Kogut and Zander (1992) divide knowledge into information and know-how. Information is
components. Know-how is procedural knowledge, often informal, about the ways that various
parts of a total system fit together, and is embedded in the organizing principles of cooperation
the distinction between routines (components) and skills or how the routines are put together
particular technologies and in the organization capability by which the technologies are applied
humans in work (Orlikowski, 1992) suggests that any reengineering project is an interaction of
technology and the organization, and therefore informal knowledge of complex technologies is a
by-product of that interaction (Baba, 1990). Designers of formal technology typically cannot
predict exactly how a given piece of technology will perform (Tornatzky and Fleischer, 1990).
The designers may not know, for example, all the ways a technology can fail, or the techniques
to prevent or correct the failure. As a result of such gaps, the formal training, documentation and
8
policies that accompany technology often requires additional functional-knowledge that can only
be learnt through use (von Hippel, 1998; Orr 1996). For example, Hargadon and Fanelli’s (2002)
work shows that work groups both affect technology through social means of control, and
interact creatively with technology – discovering its hidden potential in coping with its emergent
processes, and as knowledge of the interactions between work processes and technology.
IT Project Performance
distributed across firms, leading to different patterns of IT use and effectiveness (Barney, 1991).
Using a resource-based view of the firm, Bharadwaj (2000) divides IT resources into physical IT
infrastructure, human IT resources and IT-enabled intangibles representing technical, social and
knowledge resources respectively. This latter knowledge resource draws from the intellectual
capital embedded in the skills and experience of employees, business processes, policies, and
information repositories, that determine not only individual project performance, but a firm’s
overall effectiveness (Clark, Cavanaugh, Brown, and Sambamurthy, 1997; Nelson and Winter,
1982; Mata, Fuerst and Barney, 1995; Clemons and Row 1991).
regarding cost, schedule, and technical specifications (Pinto and Mantel, 1990). It has also been
determined more subjectively by measuring client satisfaction (Baker, Murphy and Fisher,
1988). How a project’s clients view the manner in which a project was implemented
9
significantly affects their satisfaction with project outcomes (Melone, 1990). Two key measures
of client satisfaction are project validity and project effectiveness. In the context of
reengineering projects, project validity indicates the extent to which the reengineered process is
“right” for the intended clients – managers and employees that actually use the new technology.
Project effectiveness is concerned with determining whether the reengineered process has a
positive impact – for example, whether the business unit is more effective (Schultz and Slevin,
1975). Although a project may meet traditional performance measures such as being on time
and within budget, implementation success is questionable unless these aspects of client
problems must be identified in advance and action must be taken to resolve them. In order to
identify the technical and social dimensions of reengineering projects we conducted a survey of
the information systems literature that provided us with key implementation issues for IT-based
managers during this time period, were problems with software and database development,
policies (Brancheau, Janz and Wetherbe, 1996; Troyer, 2002). Organizational change issues
identified in the literature as critical to both IT projects and reengineering success were top
management support (Armstrong and Sambamurthy, 1999; Troyer, 2002), joint business unit and
IT planning (Madnick, 1991; King, 1997), shared understanding of the goals to be obtained
(Horner Reich and Benbasat, 1996, 2000) and user involvement (Noble and Newman, 1990;
Ravichandran, and Rai, 2000). These eleven issues, which we collectively call IT/reengineering
issues, capture the major social and technical obstacles confronting reengineering projects.
10
III. Methodology
Instrument Development
To measure the impact of the social and technical dimensions on project performance, we
constructed measurement scales for the IT/reengineering issues affecting project implementation
described immediately above. A survey item was developed for each of the eleven
IT/reengineering issues using a seven-point Likert scale. These items formed the basis of our
independent variables representing obstacles embedded in the social and technical dimensions.
Our dependent variable, project performance, was operationalized using two measures:
an objective measure being duration of IT-related project delay (IT-delay) and a subjective index
of client satisfaction. The objective measure, IT-delay, is indicative of the project’s technical
performance. Technical shortfalls delay project completion. For the subjective index we
adapted a validated instrument for project performance -Pinto and Slevin’s (1986) Project
Implementation Profile - that essentially measures perceived client satisfaction with the project
implementation process. In measuring client satisfaction we capture the extent to which social
A pilot study was conducted to evaluate the meaning, ordering and representativeness of
the survey items. The initial instruments were sent to four organizations (two in health care and
two in telecommunications) for review by project managers, IT managers and process users to
evaluate the degree to which questions and items represented the intended issues and concepts.
These participants were then interviewed by phone for their feedback. Their comments were
11
Data Collection
popular press, private agencies and government offices. The sampling frame consists of projects
implemented in the 1990s in the U.S. health care sector. One hundred and sixty-six project
managers and IT managers across 83 organizations were contacted by phone and informed of the
questionnaires across the two participant categories were obtained from 63 organizations about a
recent reengineering project. Recent refers to project completion with the previous 6 months.
The projects, which served as data collection points, varied across a number of dimensions. The
typical project was allocated a multi-million dollar budget, over a three-year time horizon to
change an institutionalized process involving multiple work flows. The majority of projects had
at least one team member with prior organizational change experience. Project duration ranged
from 4 months to 9 years, with an average completion time of 10 months. Of that, 6 months was
wide hospital information system (HIS) affecting the work patterns of physicians, nurses and
ancillary personnel. Several multi-hospital systems had implemented picture archive and
communication systems that eliminated hardcopy radiographs (x-ray film) and dramatically
altered the work patterns of radiologists. Physician order-entry systems altered the
communication flow between physicians and other members of the medical team, eliminating
ward clerks as intermediaries. Pharmacy information systems allowed for direct entry
prescriptions and digital updates to dispensing machines on each floor. This substantially
reduced foot traffic between the pharmacy and nursing units. Laboratory systems allowed
12
physicians to retrieve lab results in-house and remotely, in their office. These HIS applications
replaced massive paper trails with electronic data transfer and storage. In doing so, work
patterns were significantly altered for all members of the medical team.
We targeted our questions to the managers most knowledgeable about the subject of the
questions. IT managers were queried about the eleven IT/reengineering issues used to form our
independent variables. Using a seven-point Likert scale, IT managers were asked to identify the
extent to which each of the seven IT issues were a significant source of IT-related project delays,
and indicate the level of shared goals, joint planning, top management support and user
managers provided data concerning the dependent variables of IT-delays and client satisfaction.
Using separate respondents, one for our independent variables and one for our dependent
variables, reduces the threat to internal validity that could come from the same respondents
answering both independent and dependent variables (Campbell and Stanley, 1963).
Table 1 shows the descriptive statistics for the eleven IT/reengineering issues. Most of the means
for the individual issues are close to the center of the seven-point Likert scale, and most of the
standard deviations are close to 2.0. The exception is User Involvement that has both a low mean
and small standard deviation, indicating that there were few problems with User Involvement.
--------------------------------
Insert Table 1 About Here
--------------------------------
13
--------------------------------
Insert Table 2 About Here
--------------------------------
model restricted to two-factors – one representing the technical dimension and the other the
social dimension. Table 2 shows the variance explained by each component, and we note that
the eigenvalues suggest that the IT/reengineering issues fall into three factors rather than the two
specified – an issue we will return to later. The factor loadings in Table 1 show that Database
Software Development load on the same factor that we label the “technical” dimension. The four
remaining items, User Involvement, Top Management Support, Shared Goals, and Joint Planning
load on the second factor that we label the “social” dimension. In the context of the STS
Configuration, Training, and Software Development to load together as the technical dimension
since that dimension is defined to include tools, techniques, procedures, skills, knowledge, etc.
We would also expect User Involvement, Top Management Support, Shared Goals, and Joint
Planning to load together as the social dimension because that dimension is defined to include
people in the organization and the relationships among them. STS theory is unclear about where
Policies belong. If Policies relate to procedures, then Policies belong in the technical dimension
– indeed, that is where our PCA placed the item. However, Policies could govern relationships
between people, in which case the item would belong in the social dimension.
The factor loadings in the rotated factor matrix for the two-factor PCA (Table 1) indicate
an absence of correlation of items with more than one factor, suggesting both convergent and
discriminant validity for the technical - social dichotomy (Stone, 1978; Kerlinger, 1986). The
14
majority of primary factor loadings exceeded the 0.722 critical value for significance (Stevens,
1986), with the remaining greater than 0.60. This two-factor solution explains 60.7 percent of
the variance among the IT/reengineering issues. Cronbach’s Alpha for the seven items that
loaded on the technical dimension is a strong 0.83, and for the four items that loaded on the
The two-factor PCA loadings over 0.60 in Table 1 – those associating items with
dimensions – are all positive. The items are all coded in the “negative” so that, for example, the
item “Top Management Support” is increasing in the lack of support. Therefore, higher factor
scores from a given dimension represents increasing problems with that dimension.
Next, we use ordinary least squares (OLS) multiple regression models to test the
importance of joint optimization between the technical and social dimension in explaining IT-
delays and in explaining client satisfaction. The first regression model estimates the main effects
Y = ß0 + ß1 X1 + ß2 X2 + e,
where X1 and X2 represent the factor scores from the two-factor PCA analysis described above –
technical and social respectively, Y is either IT-delay or client satisfaction, the ßi s are parameter
estimates, and e is a random disturbance term. Because the factor scores are obtained from a
PCA with orthogonal (varimax) rotation, there are no concerns for multicollinearity in the first
regression model. Table 3 displays the results of six regression models, Models 1 and 2 focus on
--------------------------------
Insert Table 3 About Here
--------------------------------
15
Model 1 in Table 3 shows neither the technical nor social dimensions explain IT-delay –
together they explain less than 1 percent of the variance in IT-delays. The social dimension
explains significant variance in client satisfaction whereas the technical dimension does not. The
coefficient of the social dimension is negative, consistent with increases in the factor scores
representing the underlying items - a lack of User Involvement, Top Management Support,
Shared Goals, and Joint Planning. That is, the less User Involvement, Top Management Support,
Shared Goals, and Joint Planning (higher factor scores), the lesser is client satisfaction.
The second regression, Model 2, captures the reciprocal effect of joint optimization
through a multiplicative interaction between the factor scores representing the technical and
social dimensions:
Y = ß0+ ß1 X1 + ß2 X2 + ß3 X1 X2 + e,
where X1 X2 is the interaction effect between the two dimensions and the remaining terms are as
before. Although the multiplicative interaction is not the only manner in which reciprocal
interdependence between the technical and social dimensions could be formulated, the
multiplicative interaction would likely pick up any effects of interdependencies between the two
dimensions.
As Model 2 indicates, the interaction of the technical and social dimensions does not
significantly explain either IT-delays or client satisfaction. Variance inflation factors just over
1.0 indicate that the interaction is not collinear with the main effects, and the remaining
diagnostics are consistent with no multicollinarity. Indeed, apart from the social dimension
explaining 20 percent of client satisfaction, none of the other variables are significant in either
regression on either dependent variable. Therefore, joint optimization between the technical and
16
social dimensions – in the form of an interaction term - is insignificant in explaining IT-delays or
flexible form for modeling interdependence and joint optimization. The interaction form with
factor scores captures both a matching in levels of technical and social obstacles – that is
high/high and low/low versus a mixture, and a monotonic scale from a combination of low/low
up to high/high. Thus, the lack of a significant increase in explanatory power with interaction as
compared to only main effects suggests that if there is an effect of interdependence of technical
and social dimensions, then that interdependence is not defined by matching or a monotonic
As we noted earlier, the eigenvalues in Table 2 suggest three factors or dimensions underlie the
set of IT/reengineering issues: three of the eigenvalues are greater than unity (Kaiser rule), and
the third is roughly twice as large as the fourth and the fourth is only slightly larger than the next
three eigenvalues (scree analysis). These three dimensions explain over 73 percent of the
variance in IT/reengineering issues. Consequently we ran a three-factor PCA, and the rotated
It is clear from the loadings from the three-factor PCA in Table 1 that there are three
separate dimensions in the data. Comparing the two-factor PCA in Table 1 to the three-factor
PCA, we see that the items that loaded on the initial technical dimension divided into two
dimensions. In the three-factor PCA the first dimension consists of Database Development,
Hardware Linkages, Network Configuration and Software Development. The second dimension
17
is made up of Expertise, Policies, and Training. We continue to name the first “technical”, and
name the new dimension “functional-knowledge.” The inclusion of Policies with this new
dimension reinforces the interpretation of Policies as procedures we found in the two-factor PCA
rather than as rules that govern relationships between people. The four items that made up the
social dimension in the two-factor model, User Involvement, Top Management Support, Shared
Goals, and Joint Planning, remain loaded on a separate single dimension that we continue to
name “social.” As before, the absence of correlation between item groupings and the relatively
clean factor loadings, suggest both convergent and discriminant validity for the three
dimensions. All primary factor loadings except one (at 0.713) exceeded the 0.722 critical value
for significance (Stevens 1986). Cronbach’s Alpha for the new four-item technical dimension is
0.824, and for the functional-knowledge dimension is 0.81, both inside the acceptable range for
reliability. In both PCAs all the loadings over 0.7 are positive. Given the items are all coded in
the “negative,” higher factor scores from a given dimension again represents increasing
Of course, aspects of knowledge are embedded in all of our items – for example, Top
Management Support embeds knowledge of the relationship between top management and the
project. Our PCAs do not indicate that all aspects of knowledge separate into a third dimension,
but rather the items that are direct measures of functional know-how (Kogut and Zander, 1992) –
expertise, training and policies – are naturally separated from technical and social items in which
other aspects of knowledge are embedded. Based on this analysis we believe that in the context
of reengineering projects, the explanatory power of STS theory may be enhanced by including
18
Testing Joint Optimization with Revised STS Dimensions
Even with the addition of a third dimension, functional-knowledge, the validity of the original
principle of STS – joint optimization - still remains an important question. That is, do the
again use OLS regression model to test the importance of joint optimization of the social,
satisfaction. Model 3 estimates the main effects of the technical, social and functional-
knowledge dimensions,
Y = ß0 + ß1 X1 + ß2 X2 + ß3 X3 + e,
where X1 , X2 and X3 represent the factor scores from the three-factor PCA analysis described
satisfaction, the ßi s are parameter estimates, and e is a random disturbance term. Again, because
the factor scores are obtained from a PCA with orthogonal (varimax) rotation, there are no
Table 3 shows the results of the main effects model. After separating functional-
knowledge from the technical dimension, the latter is significant in explaining variance in IT-
delays. Neither of the other dimensions’ main effects significantly explain IT-delays, and Model
3 is not significant at the .05 level - together the three main effects explain less than 11 percent
of the variance in IT-delays. In the regression on client satisfaction the social dimension is the
only significant main effect and together the three dimensions explain 17.4 percent of the
variance in client satisfaction. As before, with the underlying items coded in the “negative”, the
19
To re-test the principle of joint optimization in the context of technical, functional-
knowledge and social dimensions, we formulate a second set of regression models that capture
the effect of joint optimization through multiplicative interactions between the factor scores.
With three dimensions there are three alternative specifications. The first is Model 4 that
includes the full set of interactions - that is, all two-way interactions between the technical,
Y = ß0+ ß1 X1 + ß2 X2 + ß3 X3 + ß4 X1 X2 + ß5 X1 X3 + ß6 X2 X3 + ß7 X1 X2 X3 + e,
where Xi Xj are the interaction effects between each two sets of dimensions and the X1 X2 X3 is
As we see from Table 3 (Model 4), the inclusion of the full set of interactions increases
the variance explained in IT-delay roughly fourfold to close to 38 percent as compared to the
main effects model, with the significance of the model and Adjusted R-Square also greatly
increased. In addition, the technical and functional-knowledge main effects, the two-way
interactions with functional-knowledge, and the three-way interaction are significant at the .02
level. We conclude that the full model – main effects and all two-way and three-way interactions
– is an excellent model for explaining IT-delays. In addition, all of the interactions with
regression on client satisfaction we find that the social dimension and the interaction between
social and technical dimensions are significant at better than the .01 level, and that the inclusion
of the interactions doubles the variance explained as compared to the model that includes only
the main effects. It is apparent that the interaction between social and technical dimensions,
which was not sufficiently refined in the previous two-factor model using a technical dimension
20
that included items from both current technical and functional-knowledge dimensions, is
Y = ß0+ ß1 X1 + ß2 X2 + ß3 X3 + ß4 X1 X2 + ß5 X1 X3 + ß6 X2 X3 + e,
or a three-way interaction as in
Y = ß0+ ß1 X1 + ß2 X2 + ß3 X3 + ß4 X1 X2 X3 + e,
where the variables and parameters are defined as in the full model. Model 5 shows the
regression results when only the two-way interactions are used, and Model 6 shows the
regression results when only the three-way interaction is used. For IT-delays, two-way
interactions between the technical and functional-knowledge dimensions, and between the social
and functional-knowledge dimensions remain significant at the .10 level, as do the main effects
of the technical and functional-knowledge dimensions. As compared to the full model, however,
the variance explained and the Adjusted R-Square falls by 9 percentage points. In the regression
model on IT-delays with only the three-way interaction, the three-way interaction along with the
technical dimension is significant at better than the .05 level. In this case, however, the Adjusted
For the regressions on client satisfaction with only the two-way interactions, the two-way
interactions of the technical and social dimensions as well as the main effect of the social
dimension is significant at the .01 level, as it was in the full model. The interaction between the
interaction or main effect is significant, and the Adjusted R-Square marginally decreases
compared to the full model. In the regression with only the three-way interaction, the three-way
interaction and the main effect of the social dimension are significant at better than the .025
21
level, but the Adjusted R-Square falls by roughly 30 percent relative to the other two regression
models. On balance, the effects of the interactions that include the social dimension consistently
In the two full models – regressions of the full set of interactions on IT-delays and client
satisfaction - there are some marginal concerns for multicollinearity. The variance inflation
factors are closer to 2.0 than 1.0 for the two-way interaction between technical and social and the
three-way interaction. Although not severe – variance inflation factors under 4.0 are considered
acceptable (Myers, 1986), we believe this explains why the three-way interaction in the model of
only the three-way interaction on client satisfaction is significant, when that interaction is not
More importantly, using Adjusted R-Square that controls for the number of variables in
the regression, each regression model with interactions is substantially better at explaining both
IT-delays and client satisfaction. From Model 3, the Adjusted R-Square of the main effects
regression on IT-delays is 0.06. From Models 4-6 the Adjusted R-Square values for the
regression models with interactions are between 0.127 and 0.300 for IT-delays – a substantial
and significant increase over the main effects regression. Similarly, from Model 3 the Adjusted
R-Square of the main effects regression on client satisfaction is 0.132. From Models 4-6 the
Adjusted R-Square values for the regression models with interactions are between 0.191 and
0.268 for client satisfaction - again a substantial increase over the main effects regression.
Indeed, this gain in explanatory power likely underestimates the true gain as some of the
interaction terms are not significant, and the results of models without those interaction terms
may yield even higher Adjusted R-Square values. Consequently, we conclude that separating the
original STS technical dimension into functional-knowledge and technical makes the principle of
22
joint optimization important in explaining IT-delays and client satisfaction in reengineering
projects.
V. Conclusion
In keeping with two underlying dimensions and the principle of joint optimization, STS theory
hypothesizes than the technical and social dimensions are separate from each other, yet
reciprocally interdependent (e.g., social problems affect technical problems and vice versa). The
factor loadings of the items in the two-factor PCA validates the separation into technical and
social dimensions. However, the regression models of the main effects, and main effects together
with interactions, refute the reciprocal interdependence aspect of STS, as evidenced by a lack of
significant interaction effects. Particularly for the models explaining IT-delays, regression
models with main effects of these two dimensions alone or with main effects plus interactions
that yield negative Adjusted R-Square values suggests a model specification error.
Although the two-factor model explains over 60 percent of the variance in the eleven-
item IT/reengineering issues data, relaxing the constraint of two underlying dimensions results in
a superior three-factor model that explains over 73 percent of the variance. The specification
error is revealed in the three-factor model when the items comprising the technical dimension in
the two-factor PCA solution separate into two distinct dimensions: technical problems and
respectively. The items and loadings for the social factor underlying the construct “social
After separating the technical dimension into technical and functional-knowledge, and
maintaining the social dimension, the models with the main effects of these three dimensions
23
alone explain little of the variance in IT-delays or in client satisfaction. However, once the
interactions between these different dimensions are included in the regression models the main
effects and interactions explain significant and substantial amounts of the variance in our
performance metrics. Our analysis concludes that the three dimensions – technical, functional-
knowledge, and social – are interdependent in their effects on project performance. Therefore,
once certain forms of knowledge are accounted for as a separate dimension in STS theory rather
than embedded in the technical and social dimensions, the main principle of the theory, joint
optimization, is confirmed.
separated as a dimension, regressing social and technical dimensions and their interaction on IT-
dimension, regressing all the main and interaction effects between social, technical and
effect of functional-knowledge and all of the interaction effects of functional-knowledge with the
other two dimensions are significant. In contrast, the social dimension is most important in
explaining our subjective measure of client satisfaction. Even here the functional-knowledge
items separated from the technical dimension indirectly play an important role as the social
dimension only has significant interaction with the refined version of the technical dimension.
It is worth noting that functional-knowledge was present in the two-factor PCA and the
subsequent regressions on IT-delays and client satisfaction – the IT issues that comprised the
Thus, our results are not simply that functional-knowledge is important, but that treating
24
functional-knowledge as a separate dimension is important for explaining outcomes in the STS
theory framework.
The effectiveness of STS theory has been questioned in contexts that are turbulent –
uncertain and rapidly changing – and in contexts where work is knowledge-intensive (see Nadler
and Tushman, 1997; Adler and Docherty, 1998). Indeed, these are contexts that characterize
reengineering projects, and where important knowledge resides in the individual and the
organization rather than with the technology. Direct application of STS theory to these contexts
requires that knowledge reside in both the technical and social dimensions as in Shani and Sena
(1994) where individual’s aptitudes and skills are part of the social dimension, and following
STS theory where knowledge is part of the technical dimension. However, we found that
empirically this direct application does not explain project performance, and nor does it follow
the main STS principle of joint optimization. Instead, separating functional-knowledge from the
social and especially technical dimensions is important in specifying the joint optimization
25
VI. References
Adler, N. and Docherty, P. 1998. Bringing Business into Sociotechnical Theory and Practice.
304-327.
Baker, B., D. Murphy and D. Fisher.1988. Factors Affecting Project Success. In Project
Management Handbook, Second Edition (Cleland and King, Ed). New York: Van Norstand
13(3) 274-290.
Bostrom, R. and Heinen, S. 1977. MIS Problems and Failures: A Sociotechnical Perspective.
Brancheau, J., Janz, B. and Wetherbe, J. 1996. Key Issues in Information Systems Management:
26
Brown, R. 1967. Research and Consultancy in Industrial Enterprises. Sociology, 1: 33-60.
Brynjolfsson, E., and L. Hitt 1996. Productivity Lost? Firm-Level Evidence on the Returns to
Burkhardt, M. and Brass, D. 1990. Changing Patterns of Change: The Effects of Change in
Technology on Social Network Structure and Power. Administrative Science Quarterly, 35:
104-127.
Campbell, D. and Stanley, J. 1963. Experimental and Quasi-Experimental Designs for Research.
Chase, R. and Susman, G. 1986. A Sociotechnical Analysis of the Integrated Factory. Journal of
Capabilities in the IS Organization: Insights From the Bell Atlantic Experience. MIS
Clemons, E. and Row, M. 1991. Sustaining IT Advantage: The Role of Structural Differences.
Fox, W. 1995. Sociotechnical System Principles and Guidelines: Past and Present. Journal of
Gordon, R.J. 2000. Does the 'New Economy' Measure Up to the Great Inventions of the Past?,
27
Griffith, T.; Sawyer, J and Neale, M. 2003. Virtualness and Knowledge in Teams: Managint the
27(2) 265-287.
Hargadon, A. and Fanelli, A. 2002. Action and Possibility: Rconciling Dual Perspectives of
Horner Reich, B. and Benbasat, I. 1996. Measuring the Linkage Between Business and
Horner Reich, B. and Benbasat, I. 2000. Factors That Influence the Social Dimension of
24(1) 81-113.
1069.
Kerlinger, F. 1986. Foundations of Behavioral Research. Fort Worth, TX: Holt, Rinehart and
Winston, Inc.
King, W. 1997. Integration between Business Planning and Information Systems Planning:
28
Kogut, B., and Kulatilaka, N. 1994. “What is Capability? Co-evolution, Complementarities, and
Kogut, B. and Zander, U. 1992. Knowledge of the Firm, Combinative Capabilities and the
Lee, B., and Barua, A. 1999. An Integrated Assessment of Productivity and Efficiency Impacts
of Information Technology Investments: Old Data, New Analysis and Evidence. Journal of
Madnick, S. 1991. The Information Technology Platform. In The Corporation of the 1990’s
Mata, F.; Fuerst, W. and Barney, J. 1995. Information Technology and Sustained Competitive
Mitchell, V.and Zmud, R. 1999. The Effects of Coupling IT and Work Process Strategies in
Mumford, E. 1996. “Planning for Human Systems.” Human Relations, 49(4) 527-532.
Myers, R. 1986. Classical and Modern Regression with Applications. Boston, MA: Duxbury
Press.
29
Nelson, R. and Winter, S. 1982. An Evolutionary Theory of Economic Change. Canbridge, MA:
Noble, F.and Newman, M. 1990. User Involvement as an Interactive Process: A Case Study.
Orlikowski, W.J. 1992. The Duality of Technology: Rethinking the Concept of Technology in
Oliner, S.D., and D.E. Sichel, 2000. The Resurgence of Growth in the Late 1990's: Is
Orr, J. 1996. Talking About Machines: An Ethnography of a Modern Job. Ithica, NY: ILR
Press.
Pasmore, W.; Francis, F,; Haldeman, J. and Shani, A. 1982. Sociotechnical Systems: A North
American Reflection on Empirical Studies of the Seventies. Human Relations, 35(12) 1179-
1204.
Pasmore, W.; Petee, J. and Batrian, R. 1986. Sociotechnical Systems in Health Care: A Field
Pava, C. 1986. Redesigning Sociotechnical Systems Design: Concepts and Methods for the
Pinto, J. and Mantel, S. 1990. The Causes of Project Failure. IEEE Transactions on Engineering
Pinto, J. and Slevin, D. 1986. The Project Implementation Profile: New Tool for Project
30
Ravichandran, T. and Rai, A. 2000. Quality Management in Systems Development: An
Shani, A.B., and Sena, J.A. 1994. Information Technology and the Integration of Change:
Spender, J-C. 1996. Making Knowledge the Basis of a Dynamic Theory of the Firm. Strategic
Stevens, J. 1986. Applied Multivariate Statistics for the Social Sciences. Hillsdale, NJ:
Stone, E. 1978. Research Methods in Organizational Behavior. Glenview, IL: Scott, Foresman
and Company.
31
Trist, E. 1981. The Sociotechnical Perspective. In Perspectives on Organization Design and
Behavior, Van de Ven and Joyce (Eds.), New York: John Wiley and Sons.
Trist, E. and Bamforth, K. 1951. Some Social and Psychological Consequences of the Long-
Corporation. http://www.csc.com/newsandevents/news/1303.shtml.
Van de Ven, A., and Joyce, W. 1981. Overview of Perspectives on Organization Design and
Behavior. In Perspectives on Organization Design and Behavior, Van de Ven and Joyce
van der Zwaan, A. 1975. The Sociotechnical Systems Approach: A Critical Evaluation.
Von Hippel, E. 1998. Economics of Product Development by Users: The Impact of Sticky Local
32
Appendix A - Survey Instruments: Project Manager Questionnaire
This survey is being conducted to gain a better understanding of the relationship between a
firm’s information technology and the redesign of its business processes. Your input is needed
to advance our knowledge in this area.
Background Information
What is the most recent redesign project you’ve implemented (project name)? ___________
When did this project start? __________________________________________________
What was the initial scheduled completion date? __________________________________
What was the actual completion date? __________________________________________
How many significant delays were experienced in implementing the project? ____________
What were the primary causes of delay? ________________________________________
What was the most critical IT-related delay experienced? ___________________________
How many members of your redesign team had previous redesign experience? __________
33
IT Manager Questionnaire
This survey is being conducted to gain a better understanding of the relationship between a
firm’s information technology and the redesign of its business processes. Your input is needed
to advance our knowledge in this area.
34
Table 1: Descriptive Statistics and Rotated Principle Component Matrices.
35
Table 2: Principle Component Analysis - Total Variance Explained.
36
Table 3: Multiple Regression Results
37