Table of Contents
Understanding the dynamics of information system use
Abstract
User behavior in information systems (IS) has traditionally been modeled based on well-
accepted behavioral theories like the theory of reasoned action (TRA), technology
acceptance model (TAM) and the theory of planned behavior (TPB). A major limitation
of this approach is its dependence on a snapshot view to capture the relationships
between the variables at any given time. Research has shown that such relationships
change over time (Szajna, 1996). The twin objectives of this study are to firstly, identify
the limitations of the current approaches to model IS use and, secondly, to demonstrate
the theoretical utility of understanding user behavior by incorporating feedback loops into
such models. This study captures the dynamics of user behavior by employing system
dynamics modeling. We have extended traditional hypotheses associated with user
behavior into a dynamic hypothesis by incorporating feedbacks. Specific feedbacks lead
from IS use to perceived ease of use and from productivity to IS-related work. We have
tested the model under different scenarios. The managerial implication of this work is a
better understanding of user behavior because these models have been able to
demonstrate archetypal IS use pattems. Using such models managers can analyze
different usage scenarios before making system changes or introducing new systems. The
theoretical contribution of this study is the identification of archetypal user behavior by
linking productivity with IS use.
Keywords: Information system (IS) use, system dynamics, archetypes
1. Introduction
Whenever a new or an updated or enhanced information system (IS) is introduced in an
organization, its success is dependent upon how well it is accepted and subsequently used
by individuals. The behavioral approach has traditionally allowed researchers to
conceptualize and model “IS use” as a dependent variable. Behavioral models posit that
user behavior is determined by a user’s attitude toward an information system. Intrinsic
and extrinsic motivators, in turn, determine the attitude. Since such models do not
explicitly link IS use to user productivity and tasks, this study forms a response to that
gap.
In this study “productivity” has been linked to IS use and we have employed system
dynamics as a research approach that complement the traditional research approach of
adding independent variables in order to seek a “better’ model. The approach taken in
this study preserves the principle of parsimony associated with the Theory of Reasoned
Action (TRA) (Fishbein and Ajzen, 1975) and its derivatives like the Technology
Acceptance Model (TAM) (Davis, 1989) and Theory of Planned Behavior (TPB) (Ajzen
1985, 1991). Appendix A contains a detailed explanation of these theories. The idea is to
recognize the direct and indirect influencers of IS use and, in doing so, incorporate
feedbacks between IS use and other variables (like IT-enabled productivity and computer
efficacy) in the research model.
This paper is organized as follows. We start by reviewing the literature on TAM. This
literature review allows us to outline the essential strengths and contributions of TAM.
We end this section by identifying major improvement opportunities while deploying
TAM as a research model. The next section describes how we built the system dynamic
model. We describe how we borrowed results from existing literature and incorporated
those results to develop the dynamic hypothesis which is used to draw out and test the
consequences of feedback loops between variables in the research model.. After
presenting the results, we discuss the results by presenting a comparative analysis of
results from four simulation runs. This section also compares and contrasts the simulation
results with “typical” results from traditional research. We conclude the paper by
highlighting the research contributions and future research directions.
2. Background
The most common and well-accepted frameworks for studying user behavior in
information systems (IS use) have been the TRA, TAM, TPB and their derivatives and
extensions. The common premise across these models is that IS use behavior is
determined by an intention to use an information system - which, in tum, is determined
by a user’s attitudes. Attitude is influenced by intrinsic and extrinsic influences
(motivators). Figure 1 shows the technology acceptance model.
Perceived
usefulness
Extemal
variables
Perceived
ease of use
Behavioral Is
Attitude >|
intention >) use
Figure 1. The technology acceptance model (Davis et al., 1989)
Over the years, many researchers have been attracted by the simplicity and explanatory
power of such models. Consequently, they have used such models as their conceptual
basis for studying IS use in different domains and contexts. In doing so, they have
extended the model and added more variables to account either for variation not
explained by the generic model or to attempt incremental improvements to the body of
knowledge associated with IS use.
A researcher typically looks for significant path coefficients when testing such models.
We present a sampling of recent research on IS use in Table 1 to summarize the diversity
of domains studied and the nature of knowledge that such studies generate.
Study Objective Use context Conclusions
The basic functionality
offered in the two Results confirm the basic structure of
Test whether individual environments (mainframe or | the model, including the mediating role
differences and IT minicomputer based and PC- | of beliefs. Results also identify several
Agarwal e based), for example, word individual difference variables
and Prasad. oeptance is mediated by processing, spreadsheets, (individual's role with respect to
e constructs of the ; - mars :
(1999) technology acceptance graphics, etc., was quite technology, an individual's level of
model similar. At the time of data education, and prior, similar
collection users had not experiences) that have significant
completely switched overto | effects on TAM's beliefs.
the new system
Examine the applicability TAM was able to provide a reasonable
of the Technology depiction of physicians’ intention to
Huetal. Acceptance Model in Physicians practicing at use telemedicine technology.
(1999) 3 explaining physicians! public tertiary hospitals in Perceived usefulness was found to be a
decisions to accept Hong Kong significant determinant of attitude and
telemedicine technology intention but perceived ease of use was
in the health-care context not.
Investigate the effect of a | Students using the digital Three system characteristics
set of individual library in the Open (relevance, terminology and screen
differences (computer University of Hong Kong design) have different effects on users’
self-efficacy and (that maintains 1,000 beliefs about the digital library. While
Hongetal. | knowledge of search electronic databases, various | relevance has a significant effect on
(2001/2002) | domain) and system e-journals, dictionaries, both perceived ease of use and
characteristics (relevance, | handbooks, and perceived usefulness of the digital
terminology, and screen encyclopedias, library library, the other two system
design) on intention to use | catalogues of local and characteristics, terminology and screen
digital libraries overseas higher education design, have significant effects only on
Study Objective Use context Conclusions
institutions, and special perceived ease of use.
indices linking to 40
distance-leaming
organizations throughout the
world.)
Test and compare sets of .
antecedent constructs A large-scale market trial of The PT se otlanteoedents Gepleiys)
Pl drawn from both TAM —_| asmart card-based electronic rl im
outs d the Perceived ayment system bein TAM while also providing managers
al. (2000) | Ore craiet Labia a with more detailed information
aracteristics of evaluated by a group of regarding the antecedents driving
ete (Pcl) retailers and merchants technology innovation adoption.
The fundamental relationships and
linkages among the TAM motivational
‘Al-Gahtani Bae empirical support for Students with one year of oined ketives nly aie d
e technology acceptance | experience in heterogeneous
(2001) model in the UK organizations across the UK ease of use) and the outcome construct
(IT acceptance) tested in this study
were in full agreement with prior
research
The study examined
volitional use of a bulletin
board system (BBS)
developed by the Institute for | Results support the extended TA M's
Test the extended TAM by | Management Accountants contention that perceived resources
Mathieson considering the (IMA). The IMA is the affect an individual's intention to use
etal. (2001) | individual's perceptions of | principal professional an information system. The main
resource availability organization for management | contribution of the extended model is
accountants, with that it expands TA M's range.
approximately 83,000
members world-wide in over
300 chapters.
An Executive Information
Test a TAM-Dased System (EIS), was used as The results supported the core TAM
Pijpers esearch model tp assess the IT tool under review. 87 | and found only a small number of
ae the factors that influence ‘ cme i
(2001) ithe ube OF TT. bY sailor senior executives drawn antecedent variables influencing actual
executives: y from 21 European-based use, either directly or indirectly.
multinationals were sampled.
Measure consumer
satisfaction with the
electronic commerce (EC)
channel through constructs | Subjects purchased similar
prescribed by 3 products through TAM components - perceived ease of
Devaraj et established frameworks, conventional as well as EC use and usefulness - are important in
al. (2002) namely Technology channels and reported their forming consumer attitudes and
Acceptance Model experiences in a survey after | stultification with the EC channel
(TAM), Transaction Cost | each transaction.
Analysis (TCA), and
Service Quality
(SERVQUAL)
Table 1. A sampling of recent TAM-based studies
Based on the Table 1, we can make the following observations about empirically
grounded studies that have employed statistical modeling.
1. These studies have been conducted in diverse domains (universities with student
respondents, senior executives, medical professionals, merchants, management
accountants etc.). The implicit assumption in all these studies is that IS use is
volitional. Hodgson and Aiken (1998) and Rawstorne et al. (2000) document the
difficulties and challenges associated with incorporating non-volitional use.
2. These studies are often aimed at comparing or extending theoretical models so as to
hypothesize a “new” situation. The “technology acceptance” component of the
models is often used as a surrogate for “technology use.” In their study (Plouffe et al.,
2000) observe that, “... once merchants have adopted the point-of-purchase
equipment needed for the new technology, consumers can use a smart-card as a
substitute for cash (p. 211).” Here is a case where a technology adoption decision by
merchants has been studied and the usage component of the consumers is assumed
away.
3. All such studies measure the perceptions of individuals at a given time by
administering a survey instrument. They often do not measure actual use. Either
intention to use is justified as a surrogate for actual use or other reasons are identified
for not measuring actual use. For instance, (Hu et al., 1999) write that, “actual
technology use was not used in the research model, a constraint resulting from the
early adoption stage of telemedicine technology (p. 97).”
4. While the implications of these studies are useful and add to the common body of
empirical knowledge, their applicability to a specific situation can be challenging. For
instance Agarwal and Prasad (1999) conclude that “it appears that there may be
nothing inherent in individual differences that strongly determines acceptance and,
because of the mediating role played by beliefs, it is possible to find alternative
means of facilitating technology acceptance and increasing individual productivity (p.
385).” Since such statements lend themselves to a broad range of interpretations, they
tend to dilute the value of useful and painstaking research.
Based on these observations, we feel that using “more of the same” approaches to
formalize and understand user behavior is fast approaching the stage of diminishing
utility for future research. While we agree with the generic TAM approach, the
managerial and theoretical utility of such approaches seems to be reaching the saturation
point. This is because, in spite of careful sampling and the sophisticated statistical
modeling, these approaches fall short of responding to, and accounting for, the real issue
at hand. Wexler (2001) captures this problem of the research and reality gap of IS use
studies when she observes that, “on the basis of his experience, Davenport concludes that
the single most important factor in user acceptance is the relevance of the system to an
individual's job success. He was surprised [however] that the Venkatesh (2000) study
found that factor to account for less than 30% of perceived ease of use (p. 1).” Such
criticism is understandable because none of the studies we sampled (of which we
presented a subset) incorporates an individual’s productivity vis-a-vis IS use.
Two studies have identified improvement areas for studying IS use. Szajna (1996)
showed that the when using TAM, the model paths change at different time periods while
studying 61 graduate students’ email use at different points in time. She measured actual
system use and concluded that once individuals have been using an IS, their subsequent
intentions are formed from their perception of its usefulness. She highlights the
importance of incorporating the temporal dimension in studying IS use by saying that “...
the difficulties in the intention-usage relationship from pre- and post-implementation
versions make an argument for the consideration of the experience component associated
with TAM (Szajna, 1996, p. 86, p. 91).” She further states that, “... the determinants of
the role of experience may be the key to understanding the belief-intention-acceptance
relationship (p. 92).”
Another study by Bajaj and Nidumolu (1998) addressed feedback from usage to
perceived ease of IS use when they studied 27 students using a debugger. They reported
strong statistical significance that past (lagged) usage significantly affects current ease of
use. They concluded that a longitudinal model is needed to better understand IS usage
(Bajaj and Nidumolu, 1998, p. 220). They reflected further that “merely convincing users
of the usefulness of the IS will not influence usage. A perception that the IS is easy to use
will lead to a more positive attitude toward using it, which will lead to greater usage
(Bajaj and Nidumolu, 1998, p. 220).” From a system dynamics view, Bajaj and Nidumolu
(1998) articulate a reinforcing or a positive feedback loop.
Based on the literature we have reviewed, it appears that is no delineable research
contour that could have emerged from the multiple applications of TAM to the study of
IS use. The large number of TAM-based studies that is matched by the diversity of users
and usage characteristics has resulted in a “fragmented adhocracy (Hirschhiem et al.,
1996)” in IS use research. By proposing to extend the TAM by incorporating productivity
and including feedbacks, we have applied the correspondence principles By
incorporating the findings of past research in the dynamic hypothesis we have been able
to provide an integrative perspective of IS use by addressing the time domain as well.
3. Model development
We have modeled a realistic work situation (shown in Figure 2) where a user of an
institutional IS (e.g. email system) is confronted by an institutional change in a new email
system. Our primary focus is to model what happens after a new information system has
been accepted (by an organization) and deployed. In a realistic scenario, volitional
aspects of information system use are diluted by the fact that work must go on regardless
of how a user perceives the system. The model can be identified by five. components: IS
use, 1S-related task, IS-use related stress, computer-related self-efficacy~and individual
productivity. Appendix B contains the reference modes for validating this study. Figure 2
can also be considered the dynamic hypothesis for this study. The two main loops
(“leaming” and “IT stress”) determine the logic of the simulation model.
IS use
IS use is conceptualized as the time spent using the IS and is measured in hours per day.
The use component is premised on the logic that the user calibrates her usage depending
on the IS-related task at hand. As the IS-related tasks increase or decrease, the user
adjusts her time spent on the IS depending on how much work can be backlogged.
| Any new theory, whatever its character--or details--should reduce to the well-established theory to which
it corresponds when the new theory is applied to the circumstances for which the less general theory is
known to hold (Weidner and Sells, 1968).
? Self-efficacy can be defined as one's personal beliefs about his or her ability to perform certain tasks (Bolt
et al., 2001)
IS-related task
IS related task is conceptualized as a stream of tasks coming in everyday. Typically, an
email user can expect a certain number of incoming mails that need to be responded to.
These inputs have been conceptualized in hours per day. As a user responds to emails,
her work backlog reduces. The rate at which a user responds to emails is a function of her
personal productivity with email use.
IS-use related stress
We assume that every individual has a normative limit on what she considers is a
reasonable time spent in reading and replying to emails in a given day. That determines
the [S-related stress for that individual depending on the extent to which actual use
exceeds that threshold. Stress is dissipated by work completion. An important variable in
this subsystem is the tolerable time for IS use. Since some individuals like to use
computers and may not mind spending time working on computers, there are others who
would like to minimize their interaction with a computer. Tolerable time for IS use
allows us to account for that aspect in the model.
IS-related zt
work Z IS-enabled
i productivity = +
“a
fo», raha
usefulness
IS work IS work IS-use related
+ backloa stress po
; DD tl
“
Intention to
IS self
efficacy + use the IS
Leaming +
Attitude toward <t
rine Sener aa =
Figure 2. The causal loop diagram showing causal influences.
Learning
Jawahar (2002) reported on the influence of dispositional factors on end-user
performance. He identified goal setting and computer self-efficacy as influencing end-
user performance. We have incorporated those results to formulate a learning loop that is
typically employed in systems dynamics and in psychology. Appendix C formalizes that
loop. Thatcher and Perrewe (2002) report a negative relationship between computer
anxiety to self-efficacy. We have incorporated that relationship by linking stress due to IS
use to learning (which influences IS self-efficacy).
Productivity
The productivity sector accounts for the productivity due to IS use. The level of stress
and the level of a user’s computer self-efficacy determine productivity. As self-efficacy
increases, productivity increases. However, as the stress due to overuse of the IS
increases, the productivity is suppressed. In order to aggregate the effects from both
stress and efficacy we use an additive term to determine aggregate productivity. This
productivity value varies between 0 and 1 and influences the rate at which backlog is
reduced.
TAM constructs
Previous research has shown that perceived usefulness and perceived ease of use
constitute the two dominant factors that affect an individual's intention to use a system.
Different studies (Adams et al., 1992; Hendrickson and Collins, 1996) of the causal
relationships between perceived usefulness, perceived ease of use, and system usage have
shown that Davis (1989) is correct in proposing that the indirect relationship between
perceived ease of use and intention to use, mediated by perceived usefulness, is an
important one.
Figure 2 shows that IS-related work is the only exogenous variable. It can be considered
analogous to extrinsic motivating factors employed in TAM and TRA. While there are
four reinforcing (positive feedback) loops, there are two counteracting (negative
feedback) loops. Appendix D shows the flow-rate diagram for the system dynamic model
that I constructed using the Stella software.
4. Simulation runs and scenarios
We simulated four scenarios and varied two variables: tolerable work backlog and
incoming work. Incoming work has been explained above and the tolerance for backlog
is an indicator of the need (intrinsic or normatively determined) to work at a certain level
of efficiency. Table 2 summarizes the four static scenarios with an initial self-efficacy
level of 20 ona scale of 100.
Run # | Tolerable work backlog Incoming work
1 Low Low
2 Low High
3 High Low
4 High High
Table 2. Four simulation scenarios with static assumptions (the low value = 1 and high
value 3 for incoming work; the low value = .5 and high value 3.5 for Tolerable
backlog)
We have chosen to show results for productivity changes over time. This is because
productivity is the outcome variable that informs management of the value of the
information system. While there are other variables whose behaviors over time are
equally important, we defer referring to those variables until we discuss these results.
Figure 3 shows the results of the simulations using static assumptions.
® productviy: 1-2-3-4-
7 08—
a: 05
0.00 36.00 72.00 108.00 144.00
Pagel Time ‘11:11PM Sat, Feb 08, 2003
3 = Fy 9 Productivity profile
Figure 3. Comparative results for productivity for the simulation runs.
We conducted four more simulation runs described in Table 3 primarily as an aid to
distill the implications of the results that I obtained in the main simulation runs. The idea
was to communicate to the reader that IS use, as has always been understood, is
determined by both intrinsic and extrinsic factors. Incoming work is the exogenous
variable and is considered the extrinsic factor.
Tolerable work | Incoming
backlog work
1 0.25 hours 1.0 hour per | Reference mode for
(low) day (low) intention to use IS Pa
is steep (varies
2 0.5 hours 2.0 hours per from .05 to 30 in
=
Run # Reference mode for intention to use
(higher) day (higher) 144 days)
3 0.25 hours 1.0 hour per | Reference mode for
(low) day (low) intention to use IS
is a slow increase
4 0.5 hours | 2.0 hours per | (aries from,05 to .
(higher) day (higher) 1 in 144 days)
Table 3. Four simulation scenarios based on tolerable backlog, incoming work and
intention to use.
Results for the next four simulation runs (shown in Table 3) reveal archetypal patterns of
behaviors as shown in Figure 4. Archetypal behavior patterns are the equivalent of a
theoretical relationship. This is typically what an IS manager would find useful to
determine operational and deployment strategies for new software introduction and
changes to existing information systems.
® productiviy: 1-2-3-4-
1 015m
at 0.55
0.00 36.00 72.00 108.00 144.00
Pagel Time 8:49AM Sat, Feb 08, 2003
a = Ta 9? Productivity profile
Figure 4. Comparative results for productivity for the simulation runs.
In the next section, we analyze these results and relate them to some of the other
variables like initial levels of self-efficacy, efficacy goals and attitude.
5. Analysis
The results in Figure 3 show that the productivity profile for simulation runs 1 and 3 are
almost identical. The increase in productivity is most marked in the initial month and
then levels off at approximately 0.67. The productivity profile for run 2 is similar to that
for run 4 in that there is there is a productivity dip before a sustained increase and
subsequent leveling out of the an IS user's productivity. It is important to note that lower
productivities are associated with lower work pressures (i.e. lower levels of incoming
work). It can be seen that a higher level of IS-related work leads to markedly higher
levels of productivity. The productivity levels for runs 2 and 4 tend to converge at around
the same level; although the productivity level for run 4 is marginally higher.
These results suggest that in high-performing organizations (where work pressures tend
to be higher) IS-productivity will tend to be higher. On the other hand, in organizations
and work situations where work loads are light, the level of productivity will taper off at
lower levels - in spite of a positive relationship between intention to use the IS and actual
IS use (i.e. a minimal level of volitional use that increases as the attitude toward using the
IS becomes more favorable). It is useful to note that the time needed to reach the same
level of improvement for the low workload case is identical (36 days or about a month).
However, what follows is instructive in that the productivity increases to respond to the
workload. In the dynamic hypothesis that we have employed, productivity is influenced
10
by IS self-efficacy. The empirical results appear to validate Bolt et al.’s (2001) findings
who show that, in the context of training, computer self-efficacy has a greater positive
effect on performance when task complexity is high than when task complexity is low. In
this case increased [S-related work is equivalent to higher work pressures and leads to
higher productivity - especially when we consider the on-the-job leaning component.
The other significant finding is that the productivity profile for simulation run 4 is higher
that that for simulation run 2. While high IS-related work is common to both these runs,
tun 4 assumes a higher level of tolerable backlog. This suggests that while positive work
pressure is desirable from the productivity standpoint, more meaningful IS effectiveness
can be obtained by reducing the pressures associated by IS use by being more tolerant of
IS-use related backlogs. While such an approach may result in short-term productivity
losses, the long-run productivity increases more that make up for the initial drop in
productivity.
Figure 4 shows the effect of changes in intention to use an IS on productivity. Runs 1 and
3 are similar and show gradual increases in productivity. The gradual increase in
productivity is understandable because of the low level of IS-related work. The pattern of
changes in runs 2 and 4 are those that we have come to expect with higher levels of IS-
related work. However, there is a difference between high levels of IS-related work that
are associated with different reference mode for intention to use IS. Run 4, describes the
productivity profile of a user with a higher level of intrinsic motivation to use an IS.
Therefore, it can be seen that some users attain higher IS-enabled productivity faster than
others depending on their attitude and, consequently, their superior intention to use the
IS. It is, however, instructive to note that in the long run, individual differences are
subsumed into the natural limits of productivity and performance.
Therefore, the primary lesson for IS researchers and practitioners is that, given enough
time, every IS user will get to the systemically determined productivity levels. However,
some users can reach higher levels of productivity in a shorter timeframe - depending on
the workload and expected productivity.
6. Conclusions
We have shown in this study that by employing the system dynamics (SD) approach, we
can complement existing research and analytical methods for understanding and studying
IS use. At the same time, this approach allows both researchers and practitioners to
perform meaningful “experiments.” These experiments can generate and convey insights
that can be generalized and be managerially relevant and applicable.
An important contribution of this study is the recognition and modeling of non-volitional
use (exemplified by ERP implementations, institutional software changes and an
individual encountering a different IS after joining another organization). The theoretical
contribution of this study lies in the identification of the IS use archetype shown in Figure
5. Identification of such archetypes will allow researchers and managers to build and test
predictive models. Archetype A behavior is exhibited by individuals who are faced with a
higher workload and a higher need for productivity. Archetype B is the productivity
11
profile of users who face a low workload levels and productivity requirement were not
stringent. It can be seen that IS-enabled productivity is a function of task requirements.
IS-enabled
productivity
A
B
Time
Figure 5. The IS use archetype for an individual
From a theoretical standpoint, archetypes contribute meaningfully to the research
discourse on IS use which is dominated by static models that lend themselves to
statistical analysis. This opens up increasingly creative research opportunities wherein
extended dynamic hypotheses can be tested for diverse environments. If the IS use
archetype identified in this study is found to he resilient, it can become a guiding
framework to develop customized approaches to introduce and institutionalize new
information systems.
It is important to understand from the results of this study that people do not use
information systems just because they like to or because they have a positive attitude. In
the real world, the mandatory use component overshadows the volitional component;
however, the volitional component is crucial in determining the rate at which a specific
user can ramp up to higher levels of productivity. This finding has important implications
for analyzing and designing policies for IS use. Policy analysis of IS use implies an
analysis of the various management decisions that are made during and after the
introduction of new or updated information systems. The model described in this study
has abstracted out many details of the individual IS use process per se, and concentrates
on the dynamics of the IS use process, including those factors that affect the performance
and intentions of individuals using the IS.
Since this study has been able to demonstrate how IS use can be modeled over time, we
have planned future studies that will cover include additional scenarios for IS use
(different applications and different implementation schedules) and other variables like
the level and nature of user support, initial and ongoing training, and the nature and
quality of work processes.
12
References
Adams, D.A.; Nelson, R.R. and Todd, P.A. (1992). Perceived usefulness, ease of use, and
usage of information technology: a replication, MIS Quarterly,16(2), 227-247.
Agarwal, R. and Prasad, J. (1999). Are individual differences germane to the acceptance
of new information technologies?, Decision Sciences, 30(2), 361-391.
Ajzen, I. (1985). The theory of planned behavior, Organizational Behavior and Human
Decision Processes, 50, 179-211.
Al-Gahtani, S. (2001) The applicability of TAM outside North America: An empirical
test in the United Kingdom, Information Resources Management Journal, 14(3),
37-46.
Bajaj, A. and S.R. Nidumolu (1998), A feedback model to understand information
system usage, Information & Management, 33(4), 213-224.
Bolt, M. A; Killough, L. N. and Koh, H. C. (2001). Testing the interaction effects of task
complexity in computer training using the social cognitive model, Decision
Sciences, 32(1), 1-20.
Davis, F. D.; Bagozzi, R. P.; Warshaw, P. R. (1989) User acceptance of computer
technology: A comparison of two theoretical models, Management Science,
35(80), 982-1001.
Davis, F. D (1989). Perceived Usefulness, Perceived Ease Of Use, And User Acceptance
of Information Technology, MIS Quarterly, 13(3), 319-340.
Devaraj, S.; Fan, M. and Kohli, R. (2002) Antecedents of b2C channel satisfaction and
preference: Validation e-Commerce metrics, Information Systems Research,
13(3), 316-333.
Fishbein, M., and Ajzen, I. (1975). Belief, attitude, intention, and behavior: An
introduction to theory and research. Reading, MA: Addison-Wesley.
Hendrickson, A. R. and Collins, M. R. (1996). An assessment of structure and causation
of IS usage, ACM SIGMIS Database, 27(2), 61-67.
Hirschheim, R.; Klein, H.K. and Lyytinen, K (1996). Exploring the intellectual structures
of information systems development: A social action theoretic analysis,
Accounting, Management and Information Technology, 6)(1/2), 1-64.
Hodgson, L. and Aiken, P. (1998). Organizational change enabled by the mandated
implementation of new information systems technology: A modified technology
acceptance model, Proceedings of the Conference on Computer personnel
Research, 205-213.
Hong, W.; Thong, J. Y. L.; Wong, M. and Tam, K. Y. (2001/2002). Determinants of user
acceptance of digital libraries: An empirical examination of individual differences
and systems characteristics, Journal of Management Information Systems, 18(2),
97-124.
13
Hu, P. J; Chau, P. Y. K; Sheng. O, R. L. and Tam, K. Y. (1999). Examining the
technology acceptance model using physician acceptance of telemedicine
technology, Journal of Management Information Systems, 16(2), 91-112.
Jawahar, I. M. (2002). The influence of dispositional factors and situational constraints
on end user performance: A replication and extension, Journal of End User
Computing, 14(4), 17-36.
Mathieson, K.; Peacock, E. and Chin, W. W. (2001). Extending the technology
acceptance model: The influence of perceived user resources, Database for
Advances in Information Systems, 32(3), 86-112.
Pijpers, G. G. M; Bemelmans, T. M. A.; Heemstra, F. J. and van Montfort, K. A. G. M.
(2001). Senior executives' use of information technology, Information and
Software Technology, 43(45), 959-971.
Plouffe, C. R.; Hulland, J. S. and Vanderbosch, M. (2001) Research report: Richness
versus parsimony in modeling technology adoption decisions - Understanding
merchant adoption of a smart card-based payment system, Information Systems
Research, 12(2), 208-222.
Rawstome, P.; Jayasuriya, R. and Caputi, P. (2000). Issues in predicting and explaining
IS usage behaviors with the technology acceptance model and the theory of
planned behavior when usage is mandatory, Proceedings of the Twenty-First
International Conference on Information Systems, 35-44.
Szajna, B. (1996), Empirical evaluation of the revised Technology Acceptance Model,
Management Science, 42(1), 85-92.
Thatcher, J. B. and Perrewe, P. L. (2002). An empirical examination of individual traits
as antecedents to computer anxiety and computer self-efficacy, MIS Quarterly,
26(4), 381-397.
Weidner, R.T. and Sells, R.L. (1968). Elementary Modem Physics. Boston: Allyn and
Bacon.
Wexler, J. (2001). Why computer users accept new systems, Sloan Management Review,
42(3), 17.
Viswanath, V. (2000). Determinants of perceived ease of use: Integrating control,
intrinsic motivation, and emotion into the technology acceptance model,
Information Systems Research, 11(4), 342-365.
14
Appendix A: Behavioral theories used to study IS use
The theory of reasoned action (TRA) defines relationships among beliefs, attitudes,
noms, intentions, and behavior. An individual’s behavior (e.g., use or rejection of
technology) is determined by the person’s intention to perform the behavior, and this
intention is influenced jointly by the individual's attitude and subjective norm, defined as
the person’s perception that most people who are important to him think he should or
should not perform the behavior in question. According to the theory of reasoned action,
attitudes toward a behavior are determined by beliefs about the consequences of that
behavior.
The technology acceptance model (TAM) is a management information system-specific
model derived from theory of reasoned action (TRA). The technology acceptance model
predicts that user acceptance of any technology is determined by two factors: Perceived
usefulness (U) and Perceived ease of use (EOU). Perceived usefulness is defined as the
degree to which a person believes that use of the modeling approach will enhance his or
her performance. Perceived ease of use is defined as the degree to which a person
believes that the modeling approach will be free of effort. Since TAM is based on Ajzen
and Fishbein's (1980) Theory of Reasoned Action, which recognizes the importance of
social norms in influencing individual behavior, the more a modeler perceives that others
(in the project or organization) who are important to him think he should perform a
behavior (use any or a specific modeling framework), the more s/he will intend to do so
(Ajzen & Fishbein, 1980, p. 57).
Theory of planned behavior holds that attitudes, subjective norms, and perceived
behavioral control are direct determinants of intentions, which in tum influence behavior.
Taylor and Todd (1996) state that the influence of peers and the influence of superiors are
antecedents to the subjective norm. Taylor and Todd also view self-efficacy, resource-
facilitating conditions, and technology-facilitating conditions as determinants of
perceived behavioral control. Tools supporting different modeling frameworks do not
exist in a vacuum and neither do software practitioners. The attitudes of clients and peers
can positively or negatively influence the attitudes and behavior of individuals. Software
development organizations are cultures with many different factors contributing toward
their growth and development, and unfortunately at times, their dogmatic beliefs about
modeling approaches and tools. Perceived behavioral control refers to "people's
perception of the ease or difficulty of performing the behavior of interest" (Ajzen, 1991).
If behavior is not under complete volitional control, the performers need to have the
requisite resources and opportunities in order to perform the behavior. The perception of
whether they have the resources will affect their intention to perform the behavior, as
well as the successful performance of the behavior. There is significant evidence in the
literature that process-oriented modeling approaches are considered easier to use. It is
quite possible that based on past experience or because of role requirements, individuals
tend to perceive more or less control respectively over use of modeling approach.
15
Appendix B: Reference modes
In order to obtain reference modes, I interviewed a diverse group of email users. These
included students as well as faculty members in a university as well as a group of email
users in a corporate setting. A change in the email system was common to both sets of
users. Common to both populations was a change in the email system - from ‘pine’ to
‘Webmail’ for the university and from pine to Lotus Notes for the corporate site. For both
sites we found that end-user reported user productivity increased over time after the new
email system was implemented (shown in Figure A-1). However, some use users did not
report significant increases in their IS-related productivity. Others reported significant
improvements. Almost all users reported high satisfaction with the email system with
some complaining about occasional problems with email outage.
IS-enabled . .
productivity A - This was a typical
A productivity profile reported by
users who faced a high workload
and stringent productivity
A requirement (low backlogs).
B - This was a typical
productivity profile reported by
users who faced a low to moderate
workload levels and productivity
requirement were not stringent
(high backlogs were tolerable).
Figure A-1. User reported reference mode for IS-related productivity
The other relevant information for modeling purposes was that users who did not report
very high productivity scores did not use the email intensively. Those reporting high
productivity scores tended to be “power users” or those who received and sent a high
volume of email. Users reported that they typically spent about one to two hours on the
email and generally checked emails. Heavy users spent anywhere between 2-3 hours per
day using the email system. Most of the users time in the initial stages of the new email
deployment was spent in learning new functionalities and shortcuts and developing
expertise in using those features. Another time-consuming activity included (re)-building
and managing address-books.
16
Appendix C: Assumption of the learning curve
Psychologists interested in learning theory study learning curves. A learning curve is the
graph of a function L(t) , the performance of someone learning a skill as a function of the
training time t. The derivative dL/dt represents the rate at which self-efficacy improves. If
Lomax is the maximum level of performance of which the learner is capable, it is
reasonable to assume that dL/dt is proportional to La-L(t). (At first, learning is rapid.
Then, as performance increases and approaches its maximal value, the rate of learning
decreases.) Thus
dL _
ab =k(Ly,, —L(t)
where k is a positive constant. We can solve this linear differential equation to sketch the
learning curve.
dL
at FAL =kLyox
Let
I(t) =e!" =e"
If we multiple the differential equation by I(t), we get
e* 7 +kLe" =kL,,,e"
(e*LY =KLye"
L(t) =e™ (fy. @"dt +C) =Lyy —Ce™
where the constant C has to be positive since it's reasonable to assume L(0) < Lax.
17
Appendix D. The system dynamics model
E]
Task
of
23
DQ rraiy A
productuty
EF]
Suess
Tolerable tme
for I$ use
Fraction change
‘ange
1S usetime
ae
A stress related
In IS use time productivity
Tolerable effect
work backlog
Perceived
usefulness
ease atude
ofuse
dissipation gee
EF] Eticacy ae
set
efficacy
Leaming
per hour
Tei eicacy
goal
Leaming
productivity
18
Back to the
