You are on page 1of 24

Decision Support Systems 43 (2007) 990 – 1013

= www.elsevier.com/locate/dss

A DSS Design Model for complex problems: Lessons from mission


critical infrastructure
Robb Klashner a,*, Sameh Sabet a,b
a
Information Systems Department, New Jersey Institute of Technology, Newark, NJ 07102-1982, United States
b
Tyco Telecommunications, 250 Industrial Way West, Eatontown, NJ 07724, United States
Available online 27 June 2006

Abstract

This paper presents a new DSS Design Model for complex, mission critical decision-making situations and its
technical, conceptual, and partial empirical evaluation. The new model was derived from conceptual design research
and through a deep qualitative field research study at an electric power utility control center—a typical arena for bwickedQ
problems. The model suggests an iterative process of theory, simulation, and decision-making interactions. The DSS
Design Model is validated using a tool instantiation, microgrid mini-case, and current research of a KMDSS for
telecommunications. Main findings suggest that broader and more integrated approaches are necessary to design DSS
for complex domains.
D 2005 Elsevier B.V. All rights reserved.

Keywords: Decision support systems; Design; Wicked decisions; Electricity; Infrastructure; Complex adaptive systems; Knowledge
management

1. Introduction However, with the explosion of software utilization


in every facet of society, the degree of difficulty and
There are numerous bDecisionsQ associated with complexity inherent in software design has presented
the development of Decision Support Systems problems for software engineers. They are now
(DSS). DSS developers are encumbered by the attacking bwickedQ [79] real-world problems within
same domain constraints and complexities that dis- Software Architecture research, which uses the ar-
rupt other software development life cycles (SDLC). chitectural metaphor from systems theorists [85],
Software engineering has increasingly brought more hardware development, and information systems
rigor to the SDLC with methodologies formalized as (IS) research [107].
programming languages and built software systems. DSS research can benefit from the progress in
Software Architecture research. During the early
* Corresponding author. Tel.: +1 973 596 5481. years of DSS research, design choices were intui-
E-mail address: klashner@njit.edu (R. Klashner). tively understood in most cases because of the
0167-9236/$ - see front matter D 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.dss.2005.05.027
R. Klashner, S. Sabet / Decision Support Systems 43 (2007) 990–1013 991

straightforward nature of the stakeholder criteria development. A great deal of the model’s complexity
(e.g., requirements) and the decision in question. comes from the necessity to incorporate a multidisci-
However, stakeholder decisions have become highly plinary set of theories and methodological approaches
subjective due to increased problem complexity in order to address all the stakeholders and constraints
[66] and convoluted contextual situations arising involved in a bwickedQ problem. The logic behind the
from rapidly evolving domain constraints. These model will primarily be presented in the context of
domain changes have altered the nature of designing systems that are used to support electric
bSupportQ because the interaction for which the power grid dispatch and control functions in a large
DSS is created is no longer a simple relationship, US utility.
but a multifaceted one because decision-makers The paper also presents an overview of how the
must engage in bsolving various semi- to ill-struc- model is currently being applied to another mission
tured problems involving multiple attributes, objec- critical infrastructure design effort to show its gene-
tives and goalsQ [66]. One can infer from these ralizability to other domains where DSS plays a key
developments that DSS design is no longer intuitive role in daily operations. The remainder of the paper is
or deterministic. organized into the following sections. In Section 2, the
These developments may indicate that the historic theoretical background for this work is presented
DSS design predisposition for deterministic SDLC including limitations of classic DSS design app-
approaches are frustrating the use of nondeterministic roaches; DSS design issues for complex domains
design activities. Therefore, in order for DSS such as the electric power industry; SDLC con-
designers to meet the bwickedQ [79] problem chal- straints; and theoretical integration using a Complex
lenge, a more comprehensive bSystemsQ design ap- Adaptive Systems (CAS) theoretical framework. In
proach is needed to address the nondeterministic Section 3, we discuss the research methodology that
complexities arising from today’s real-world deci- is a blending of various research streams with
sion-making requirements. Although both software respect to decision-making used in the new DSS
engineers and DSS designers can make progress on Design Model. In Section 4, we present the DSS
this class of problems by hypothesizing about the Design Model with its primary relationships and
solutions, it is also logical to first extract intricate interactive phases. In Section 5, we discuss how
details about the problems from specific domains we have validated the new DSS Design Model
where we know they exist. This strategy is a logical through a software architecture tool development,
step for DSS designers. a mini-case from a second electric power field site,
This paper presents a new DSS Design Model to and an empirical study of a Knowledge Manage-
address domain complexities leading to bwickedQ pro- ment DSS development for global telecommunica-
blems. We contend prescriptive life cycle approaches tion network fault analysis research, which is
must be augmented with descriptive approaches to currently underway. In Section 6, we provide our
utilize the model for process adaptation to domain- conclusions.
specific constraints. These constraints were derived
from data acquired from a field study in the electric
power industry – a mission critical infrastructure with 2. Theoretical background
real-time domain constraints – and during the subse-
quent software architecture tool development. Decisions have become more complex because the
Although this new DSS Design Model appears solution sets are dynamic and reflect the changing
simple and straightforward, it can exhibit both com- nature of domains. In addition, the bprocess of elab-
binatorial and inherent complexity due in part to our oration and refinement of the issues, alternatives, and
theoretical framework that incorporates morphogenic decision criteria itself is an important component of
principles, described later. Thus, based on this theo- the problemQ [95]. This emerging paradigm negated
retical foundation, the model reflectively and concur- the historical DSS design approaches that depended
rently informs its own evolution, thereby directly upon deterministic problems with quantifiable con-
impacting the design of the proposed DSS under straint sets [8,20].
992 R. Klashner, S. Sabet / Decision Support Systems 43 (2007) 990–1013

2.1. Limitations of classic DSS design approaches integrated into the business plans and operations –
i.e., the daily decision-making processes – effectively
Context and process largely determine resulting making ICT a necessary aspect of these processes that
software and decisions [54,95]. The software engi- DSS analysts must now factor into their designs. As
neering (SE) community widely concurs design deci- markets evolved, strategists used ICT to facilitate
sion-making controls the resulting product, or agility. Short-term mergers, extended supply chains,
bsoftware processes are software tooQ [71]. Thus, as and rapid integration efforts were all facilitated by
decision facilitating software, next generation DSS ICT. Ecommerce markets emerged based entirely on
can play a significant role in decision-making if ICT functionality. These market developments
their designers factor in real-world complexities with changed the conventional nature of brick-and-mortar
a broader view of process interactions [21]. The DSS businesses including their boundaries and standard
of the past addressed narrowly defined problems operating procedures.
reflecting well-defined technological borders, organi- Integrating DSS with ICT capabilities in light of
zational roles, and less dynamic markets. Power sum- information-rich domain dynamics [95] creates a
marizes Little’s bcriteria for designing models and richer design and will provide needed capabilities
systems to support. . .Q from 1970, which include for agility in decision-making. These capabilities
bease of control, simplicity, and completeness of re- will better address the bwickedQ aspects and problems
levant detailQ [78]. Thus, in most cases, those early [20,79] associated with complex decision-making due
DSS designs could be intuitively understood; i.e., they to ever changing domain constraints. The utilization
were easy, simple, and complete. The primary reason of new DSS designs should be done in the field where
for the lack of complexity was the straightforward the constraints of the domain are salient, e.g., real-
nature of the stakeholder criteria and the actual deci- world strategic planning [20] or recommending indi-
sion in question. The organizations were almost as viduals possessing the appropriate expertise [62].
complex then as they are now, but the expectations on Real-world field problems facilitate requirements eli-
technology were not as high in the past. Today, how- citation, analysis, and design [21,33] activities that
ever, clear delineation of boundaries is disappearing have tangible correlation with emerging bwickedQ
resulting in convoluted problems [20]. These deve- problems [79].
lopments are exacerbated by a growing dependence
on Information and Communication Technologies 2.2. DSS design issues for complex domains: the case
(ICT). Specifically, the individualistic nature of soft- of electric power industry
ware and information systems designs has been
replaced by designs that tightly couple resources The electric power industry has spent over 100
through network capabilities. years learning about electric power faults [41]. Yet,
ICT have historically been used to extend control we still have rolling blackouts [91] because the do-
over technological resources beyond typical human main is so complex that experimental tests cannot ever
capabilities (i.e., in space, time, or scale). DSS sup- be devised to find and prevent all faults. The electric
ported this extended capability to conceptualize, ana- power industry in North America has created an in-
lyze, and adjust resources in order to solve problems. frastructure spanning thousands of miles [13,41]. This
Often, only a small number of individuals control infrastructure is tightly coupled with a bcomplexQ of
organizational resources using technology (e.g., elec- ICT that forms their integrated information infrastruc-
tric power grid dispatch centers). Tightly coupled ture (I3) [52,53]. They communicate over a private
strategic approaches overlap functionality and the telecommunications network that is typically triply
control of resources (including DSS) with integrated redundant (e.g., fiber, microwave, and radio). In the
ICT designs. Resources not collocated with the con- USA, there are approximately 9000 electric power
trolling individuals must be managed using ICT. Their busses. Data retrieved from the domain through Su-
efforts and accomplishments are instrumental to over- pervisory Control and Data Acquisition (SCADA)
all organizational strategy and specific tactical cases. [6,52,102] are transmitted to command and control
Thus, ICT provided strategic advantages that were locations every 2–4 s from ubiquitous locations
R. Klashner, S. Sabet / Decision Support Systems 43 (2007) 990–1013 993

throughout the geographical area, which can span be made based on large quantities of data (e.g.,
50,000 square miles with millions of residents. SCADA). Generally speaking, raw data are gathered
These data are used by Automatic Generation Control in order to generate usable information [6], which is
(AGC) and other software systems to dynamically slightly different than knowledge [66,95]. IS can ob-
adjust the generation–transmission–distribution pro- viously process the huge volume of domain data, but
cess for reliability and optimal economic dispatch bonly a fraction of the needed information exists on
during periods of stability. But, nearly any significant computers; the vast majority of a firm’s intellectual
perturbation will render these automatic tools nearly assets exist as knowledge in the minds of its employ-
useless (or worse if faulty information is not detected) eesQ [66]. Thus, expert operators must still interpret
[52,91]. the information. For example, electric power grid
Grid dispatch control centers (GDCC) have opera- operators often abandon their DSS during grid state
tors familiar with the domain-specific constraints. changes to intuitively understand the nuances of the
They have intuitively guided utilities for decades infrastructure because of their years of experience.
past faulty conditions in both the physical electric These nuances cannot easily be captured in lines of
power grid and their information infrastructure. software code.
These human operators cannot (and should not) be Compounding their problems are doubt and risk
automated out of the electric power production pro- arising from the ICT and DSS currently in place.
cess because of the life and death nature of this Individual mechanistic or rudimentary computational
infrastructural service, because this industry is the components within the electric power grid act in a
largest one in the US [13,42]; it is literally the deterministic manner so that DSS can be designed to
lifeblood of the country. Mission critical operations leverage their predictability [6]. However, even when
such as infrastructure control should always have rudimentary components are configured in a large
humans in the loop to solve the uncommon, nonde- system, they produce undesired or unexpected beha-
terministic, and unexpected problems. But they need vior such as harmonics, which oftentimes cannot be
integrated support when making time-sensitive deci- attributed to the root cause. Field data indicate that
sions that have overloaded them (and their DSS) prior to the large West Coast blackout on August 10,
with information. 1996, operators observed unusual harmonic readings
Historically, IS analysts creating a variety of DSS on the grid, but they were unable to pinpoint the cause
in the electric power domain needed to be concerned [52]. The type of complexity associated with large
with a limited set of software requirements arising structures of components is very common (but exac-
from various institutional (e.g., Federal Energy Regu- erbated) in large software systems [74] built from
latory Commission (FERC)), organizational, and ope- thousands of individual modules, subprograms, and
rational constraints or very narrow economic objects.
principles associated with their form of a vertically Researchers have been unsuccessful in finding a
integrated monopoly. During the last 15 years in the foolproof method for testing these large software
US, deregulation of the electric power industry has systems. Unfortunately, this lack of assurance
necessitated the use of other economic or strategic increases the complexity further for electric power
theories and concepts such as market power in order grid command and control because of their dependen-
to create advanced DSS [23,72,82,94]. These electric cy on ICT and DSS. Nevertheless, the trend in the
power domain changes have increased the crosscut- electric power industry is to continue advancing ICT
ting effects (e.g., contradictory Federal and State laws to address new domain constraints [6] created by
regulating ICT design in a suboptimal manner) deregulation. However, these problems may be a re-
[52,53] necessitating new frameworks for the design sult of both social and technical (i.e., sociotechnical)
of DSS to mitigate these bwickedQ problems [79]. By design decisions [53]. For example, the recent US
definition, bwickedQ problems do not have solutions, Northeast 2003 blackout was to a large degree prop-
only best possible resolutions [79]. agated by incorrect decision-making arising from a
The electric power industry fosters the creation of dependency on ICT. The following quotes are from
wicked problems. GDCC operational decisions must transcriptions of the relevant grid dispatch control
994 R. Klashner, S. Sabet / Decision Support Systems 43 (2007) 990–1013

centers’ audio files obtained by and for the House these real-world complexities associated with design.
Committee’s investigation of the blackout [91]: Prior requirements engineering research has indicated
the importance of maintaining real-world research
[Approximately 15:07 August 14, 2003]
activities [33,34,43]. Adaptation also infers nondeter-
First Energy/Jerry Snickey: We have no clue. Our
ministic design processes because deterministic pro-
computer is giving us fits too. We don’t even know
cesses are less likely and capable of recognizing and
the status of some of the stuff around us. A0822dua-
responding to domain state changes. We assert that the
nEJ1.pdf pp. 33, lines 3–5
lack of adaptation to nondeterministic domain variety
is a limitation in classic DSS design approaches that is
MISO/Don Hunter: I called you guys like 10 min-
reflected in the Systems Development Life Cycles
utes ago, and I thought you were figuring out what
(SDLC) used to conceptualize the design process.
was going on there. A0822duanEJ1.pdf pp. 33, lines
We examine the SDLC to draw out these conflicts
23–25
in order to further substantiate the existence of these
limitations.
First Energy/Jerry Snickey: Well, we’re trying to. Our
computer is not happy. It’s not cooperating either.
2.3. Systems Development Life Cycles (SDLC)
A0822duanEJ1.pdf pp. 34, lines 1–3.
bSo the rational model of the design process, that is
These operators must make life and death split-
generally taught in our engineering textbooks (such as
second decisions based on their intimate knowledge of
Pahl and Beitz, 1984) is dead wrong and seriously
the DSS and intuition as to which information source
misleading. . .In software, we must exorcise the wa-
is correct and reliable as seen in the transcripts [91].
terfall model in favor of co-evolutionary modelsQ.
Generally speaking, as with any emergency response
[emphasis added] F. P. Brooks [14].
system, a DSS that is bnot used on a regular basis
before an emergency will never be of use in an actual An SDLC that prescribes behaviors will predeter-
emergencyQ [100], indicating a coevolutionary ap- mine the evolution of the design process. These
proach to design is needed. If electric power systems approaches do not match the nondeterministic nature
are designed to evolve and grow in parallel with the of the bwickedQ problems we seek to address. Also,
operator’s knowledge capital, intuitive capabilities the process leading to design is where many of the
DSS features can facilitate the dynamic parsing and complexities of the artifact are teased apart. Later life
reintegration of large volumes of knowledge [40,66]. cycle steps, such as testing, implementation, deploy-
Another primary constraint on the design of DSS is ment, are natural extensions of the earlier decisions.
the current grid disposition. Our field investigations The SDLC must be understood in order to appreciate
determined that the electric power operations have the emphasis on software process [71], because
bstatesQ [52]. To a large degree, electric power grid bGood decisions result from sound process. . .Q [54,
states (e.g., blackout state) predetermine the actions of pp, 392]. Decision-makers typically define the prob-
individuals in all circumstances. For example, when lem both in terms of a solution within the process
the state of the infrastructure system changes, opera- design and in terms of a certain SDLC model [11,12].
tors are beset with contradicting information as seen Therefore, a strong correlation exists between what
in the empirical data just quoted [91], their limited software process is chosen for the design of DSS and
human capacity [54] to process the volume may over- the resulting decision support quality. In other words,
whelm them [38]. the decision of how to conceptualize the system’s
A new generation of DSS can ultimately counteract evolution largely determines the success of the DSS.
a great deal of the complexity associated with this sort An SDLC can be modeled prescriptively or de-
of wicked problem. DSS designed to be adaptive will scriptively [21,81]. Typical software or IS develop-
facilitate knowledge capture and reuse for over- ment life cycles are prescriptive, such as the popular
whelmed operators. To be adaptive, there must exist Rational Unified Process (RUP) [46]. DSS Kleindor-
a direct connection to the domain bvarietyQ causing fer et al. classify (i.e., in their (Figure 1.1) DSS as one
the confusion. Field research is necessary to capture of the deterministic or prescriptive decision science
R. Klashner, S. Sabet / Decision Support Systems 43 (2007) 990–1013 995

approaches [54]. Hevner et al. assert that DSS design is workers, and existing software/information systems
generally prescriptive, b. . .within the context of dIS preclude broad prescriptive models. Custom theoreti-
design theories.T Such theories prescribe. . .meth- cal frameworks that are derived from the field assist
ods. . .instantiation. . .models. Such prescriptive theo- DSS research by guiding decision-makers to a variety
ries must be evaluated. . .Q [37]. of applicable theories and methods. The following
Choosing a prescriptive SDLC model, however, section presents the theoretical underpinning that
often produces negative results. It is a well-known emerged during the descriptive analysis of the
and documented fact [86] that critical decisions made GDCC SDLC.
in the prescriptive requirements phase have serious
and ongoing repercussions throughout the SDLC. 2.4. The integration of theoretical perspectives
Osterweil asserts, bthere should be no presumption
that process code must be overly prescriptive, author- This research is essentially premised on concepts
itarian, or intolerable either Q [71]. Therefore, we con- from design research, various qualitative theoretical
cur with Scacchi, who said, bThis, of course, should and methodological approaches, and Complex Adap-
raise concern for the relative validity and robustness tive Systems (CAS) theory. We will very briefly
of such [prescriptive] life cycle modelsQ [emphasis examine key aspects of each to highlight their rele-
added] [81]. Besides, bsound prescription should be vance. The IS community has a clarification of what
based on careful descriptionQ [54, pp. 388]. represents design research (http://www.isworld.org/
Descriptive models (i.e., not generally as in [98]) Researchdesign/drisISworld.htm) [101], with which
are utilized in an ongoing process improvement ap- we philosophically concur. We shall, however, scru-
proach that is designed to first capture, and then tinize some design principles, conjectures, and beliefs
articulate the historical software development in a in order to re-examine these precepts so that some
domain [10,49,87]. This knowledge is leveraged to degree of freedom is introduced into this discussion.
improve upon the SDLC. Design activities for DSS Instead of making the argument in our own words, we
are intrinsically tied to nondeterministic elements ari- will make some key points by quoting some notable
sing from domain constraints and humans [21] in the scholars who recently attended a NSF sponsored Sci-
decision-making process. Thus, future approaches that ence of Design workshop:
consider nondeterministic human aspects of the de-
sign and decision-making will strengthen the resulting 1. bIt is a fact that uncertain and open-ended processes
domain analysis. require radically different institutions, organiza-
The electric power industry has standard operating tions and incentives from deterministic processes:
procedures, but their SDLC processes were neverthe- the factory approach will not work for designsQ.
less undergoing dramatic evolution due to the nonde- Baldwin [1].
terministic social factors associated with the rapidly
changing deregulatory environment [67]. Prescriptive 2. bA science of software design, then, must be about
models did not adapt well at the GDCC field site radical, not about normal, designQ. Jackson [45].
because their internal organizational processes were
dynamically evolving in response to domain changes 3. bBy its very nature, design involves the integration
arising from Federal and State deregulation of the of information from a heterogeneous collection of
electric power industry [67,94]. Thus, as with [9], knowledge sources each provided by a particular
political maneuvers outweighed sound design deci- stakeholder with a particular interest in the out-
sions during this turbulent transition. This observation comeQ. Easterbrook [27].
from the field demonstrates that qualitative data must
be used to limit the combinatorial possibilities be- 4. bLarge, complex systems are hard to evolve without
cause certain aspects of the existing legacy processes undermining their dependability. Often change is
are not quickly, economically, or easily replaced with disproportionately costly,. . . I believe that system
bidealQ prescriptive approaches. Symbiotic relation- architectures are pivotal in meeting the above
ships between the built infrastructure, knowledge challenge. . . First, dependability properties tend
996 R. Klashner, S. Sabet / Decision Support Systems 43 (2007) 990–1013

to be emergent, and are much more readily mo- chitecture research [46,52,75,93,97] and IS architec-
deled and controlled at an architectural levelQ. ture research [107] abstract away lower level details of
McDermid [61]. software code to focus on major barriers to success or
the reuse of component–connector configurations at a
We have already partially addressed point #1 above high level. These high-level abstractions reduce some,
in our discussion of the SDLC, and will introduce but not all forms of complexity.
nondeterministic aspects of our DSS Design Model in The aforementioned challenges drew us to a theo-
a later section. In point #2, Jackson is referring to retically grounded Complex Adaptive Systems frame-
design activities that extend beyond the boundaries of work [15] developed as an bintegrationistQ research
the known approaches and existing knowledge in paradigm [17]. Buckley’s framework was premised on
order to meet the new demands of the user or domain. distinct concepts he gleaned from General Systems
Earlier, McDermid made a similar point by encourag- research and other disciplines. At times, the concepts
ing requirements engineers to go against their are synonymous with the current software architecture
borthodoxQ beliefs [60]. Generally speaking, many meaning. The bmodern systems theoristsQ from the
still consider qualitative social research to be radical 1950s and 1960s had tightly coupled the concepts of
or unorthodox. organization, information, and communication. The
Qualitative theory and methods, such as Grounded environment can be viewed as a bsetQ or bensembleQ
theory or ethnography, generate large amounts of non- of elements, states, or events that are to some degree
standard data. Grounded theoretical analysis [30] can distinguishably different and are generally regarded as
be a way of handling these problems. Grounded the- bvarietyQ [15]. Buckley extended those specific orga-
ory facilitates the understanding of previously unno- nizational concepts to assert that CAS elements were
ticed sociotechnical relations surrounding a group task almost entirely linked by the intercommunication of
or activity (e.g., dispatching of electricity). Grounded information.
theory can play a significant role in mission critical All known domain entities exist in a state of
domains (such as medical diagnostic) using DSS. For borganized complexityQ [15, pp, 38] bby a complex
instance, researchers have noted difficulties due to the net of relationsQ that conceptually lies between two
lack of theoretical guidance, bUnfortunately, there is opposites represented by two ideal organizing con-
no theory available to guide the selection of the best structs: borganized simplicity. . .is a complex of rela-
modelQ. [58]. We discuss the relevance of ethnography tively unchanging componentsQ; and bchaotic
later in this paper. complexityQ refers to ba vast number of components
Point #3 draws out two key aspects relevant to this that do not have to be specifically identified and
research. Qualitative approaches are more apt to dis- whose interactions can be describedQ. Buckley also
cover stakeholder agendas. Specifically, bGrounded asserted that there are relatively stable spatial, causal,
theories take concepts built covertly into any descrip- and/or temporal relations between belements or
tive account, make them visible, examine them, and eventsQ considered to be bconstraints.Q Chaotic com-
produce adequate definitions for themQ [99]. Their plexity is the complete lack of constraint and orga-
agendas can make normal development processes nized simplicity is the presence of maximum
meaningless because their requirements are not au- constraint. Buckley [16] connects these organizational
thentic, which is demonstrated later in a mini-case. concepts to a distinct type of information theory.
Easterbrook also emphasizes the need for integration, The conversion of theoretical variety into infor-
a key cause of systemic complexity. mation includes recognizing and selecting a subset
McDermid summarizes several other interwoven of the variety, then bmappingQ the environmental
concepts in Point #4 – size, complexity, evolution, variety and constraints into its own organized struc-
architecture, and emergent properties – that we have ture and/or information. This process is basically
focused our research on up to this point. Our unor- communication of the original variety and its asso-
thodox domain has necessitated design research uti- ciated constraints in a manner that remains some-
lizing the aforementioned qualitative approaches in what invariant between transmitting and receiving
conjunction with software architectures. Software ar- elements. However, Buckley also introduces an ab-
R. Klashner, S. Sabet / Decision Support Systems 43 (2007) 990–1013 997

stract model of morphogenesis that indicates that keholders with varying agendas. These constraints
two isomorphic systems will not deterministically merge to create pressure on DSS designers. We
evolve to the same future state because the occur- needed representational sufficiency when discussing
rence of any particular event within a set of con- the domain and the variety of constraints. Typically,
straints is governed by probabilities. Constraints electric power planning approaches [89] merely ab-
interact within the context of sensitivity/tension, stract the domain details away by focusing on special
organization, and contingency to morphogenically set of circumstances to arrive at requirements. How-
evolve the system to another state or blevelQ that ever, to properly understand crosscutting effects, data
is determined by the successful mapping of variety collection and analysis methods that map domain
to information or knowledge. variety into useful information must be used. Qua-
Reuse of Buckley’s research is possible because litative methods often seem less rigorous than quan-
CAS utilizes concepts such as components, informa- titative methods (e.g., semi-structured interviews vs.
tion, events, constraints, and relations that are consis- surveys), but when applied systematically are more
tent with many software and IS architecture methods. formal and effective than ad hoc or informal analy-
The fundamental ability to evolve the DSS Design sis, which is often used by practicing system analysts
Model based on the domain variety as interpreted by [50,105].
the current systemic model state reflects one of the Generally speaking, several qualitative methodolo-
tenets of the CAS theoretical framework. Thus, the gies have been explicitly developed to determine a
mapping of CAS concepts into DSS design enables more accurate picture beneath the visible organiza-
designers to modify existing multidisciplinary theories tional processes. The combined emerging and dyna-
within the integrationist paradigm as Buckley intended. mic picture can often not conform to brationalQ actor
expectations [14,84]. This indicates a need for new
research approaches. For example, empirical studies
3. Research methodology have examined usage of Computer-aided Software
Engineering (CASE) tools with the Grounded Theory
We have sought to gain a thorough understand of research approach to study the organizations’ experi-
ICT use in mission critical infrastructure by utilizing ences in terms of processes of incremental or radical
deep qualitative methods in conjunction with design organizational change [70]. Their findings were then
research principles. This research strategy was not used to develop a theoretical framework for concep-
chosen a priori but was a morphogenic and evolution- tualizing the organizational issues around the adoption
ary process as the data were collected and analyzed. A and use of these tools. Our research synthesized a
qualitative, field-based study enabled us to explore the methodological framework that was necessary to ad-
effects of institutional forces such as the deregulation of dress emergent phenomenon present in the field data.
the vertically integrated electric power industry. We The resulting synthesized DSS Design Model (pre-
were especially interested in observing what role ICT sented later) is a derivative of the framework and the
played in the inevitable emergent processes. To our field data [52]. It utilizes approaches from various
knowledge, no other prior studies have concurrently research disciplines such as Software Engineering
explored the wide spectrum of technical and social (SE), Information Systems (IS), and sociology for
factors in a context of command and control of a data collection.
major utility. These factors have shaped our research Requirements capture is an integral part of most
design. The literature guided us in our consideration of SDLC. As shown earlier, decisions about prescrip-
potentially important aspects as we observed emergent tive SDLC approaches map directly and in a deter-
behavior at our field site and throughout their industry. ministic fashion to design decisions. Therefore, we
have incorporated descriptive methods into our DSS
3.1. Methods for DSS design Design Model starting with the requirements compo-
nent of the SDLC. Requirement engineering (RE)
The preceding sections have all pointed to a has been drawn to empirical data [35,36]. Under-
confluence of constraints arising from multiple sta- standing the real-world has been a priority in RE for
998 R. Klashner, S. Sabet / Decision Support Systems 43 (2007) 990–1013

over 20 years [33,34,43,47]. The interest is due in terorganizational negotiations with the Independent
part to the irreplaceable domain expert in system System Operator established by California state law
design [2]. Realistic scenarios, goal-oriented require- AB1890.
ments, and use-cases are methods used by engineers
to formalize the domain data [2,24,34,47,50,55, 3.3. Data collection
90,103] that all factor into this research design.
Empirical data are the only way to accurately under- The electric power grid system (including its ICT)
stand the bwhyQ [35,36,105,106] behind irrational is constantly undergoing a morphogenic process.
stakeholder behavior in mission critical infrastructure Since legacy IS and software are common in most
domains because of the widely varying activities mission critical infrastructure, a data-driven approach
leading to software requirements. In addition, every using the methods detailed below will provide a view
domain – especially mission critical domains – has a of system variety, component behaviors, and interac-
body of bknowledgeQ used to binterpret and under- tions. Data were collected at different locations and
stand that worldQ [33,44]. This bshared under- levels through a variety of methods: semi-structured
standingQ [73] become bdomain models, or and unstructured interviews; reviews of company and
ontologiesQ [92] that regularize the terminology, a industry documentation; observation of operations
pragmatic extension to most ontological research and organizational activities; participation in IEEE
today. standards making activities associated with the indus-
The research began with ethnographic methods try. This triangulation through several theoretical per-
that have been used for requirements gathering spectives of various techniques of data collection
[31,48,49,87] because they effectively elicit the nec- provided multiple perspectives to identify emerging
essary domain knowledge from experts. The data phenomena.
captured from ethnographic methods are useful for The first author started with ethnographic methods
scenario, goal-oriented, and use-case research. These such as observation in order to carefully describe the
qualitative research approaches were synthesized [52] domain without contamination from informant opi-
in order to capture as much of the domain intricacies nions. Thus, the observations were taken from an
[21] as possible in conjunction with other relevant uninformed perspective. This approach facilitates the
concepts. The original research extended that synthe- discovery of information that informants often take for
sized utility to software architectures [4,28,52,55, granted or have integrated within their subconscious
68,75,83,92,93,96,97]. We propose the synthesis of specific activities associated with daily work habits.
this research is applicable for DSS design research Also, when planned software upgrades or unexpected
with modifications. events occurred (e.g., disturbance in the grid system),
the observer was allowed to remain in the GDCC. All
3.2. Site selection grid dispatch personnel on both 12-h shifts were
observed randomly throughout the field study, but
The field site for this research effort was initially a this was later augmented with impromptu and semi-
Grid Dispatch Control Center (GDCC) of a major formal extended interviews.
West Coast utility. The system analysts at GDCC These initial semi-structured, then unstructured,
were required to design information systems based and later impromptu interviews were conducted
on numerous constraints arising from different stake- with a cross-section of the grid dispatch informants
holder groups, e.g., FERC, State Public Utility Com- to gather data about the impacts from deregulation;
mission, and other organizations. Their constraints their duties; the technology they had used and were
were often contradictory, which put the system analyst now being forced to utilize; and the organization as
in the position of having to interpret many ambiguous a whole. These interviews provided information
or contradictory policy documents in order to arrive at about the general operation of the GDCC, an ope-
an equitable solution to be embedded in the software. rator’s role as a member of the GDCC team, and
The GDCC management was interested in gaining a important insights these skilled individuals had
clearer picture of the design process to facilitate in- about the industry in general. Support staff outside
R. Klashner, S. Sabet / Decision Support Systems 43 (2007) 990–1013 999

the main GDCC was also interviewed to get other This research is premised on the knowledge that the
perspectives. bsystemQ under design suffers from numerous cross-
Informants included all supervisors, upper man- cutting effects [52]. Field data regarding the activities
agement, IS analysts who designed and developed of individuals from the GDCC and general electric
GDCC software to support the GDCC operators; power industry will play a vital role in DSS design in
electric power marketers and forecasters who inter- this domain. However, the appropriate research me-
acted with the GDCC operators in planning process- thods must be applied to extract the relevant informa-
es; management; and accountants who monitored the tion for humans intuitively grappling with the wicked
result of the operations for discrepancies. At least infrastructural problems. Without these data, the de-
one individual from each control center duty station sign of DSS will lack real-world relevance and will not
was semi-formally interviewed, which totaled 30 h of seize the opportunity to move onto the next generation
taped interviews (see [53] for actual data such as of problems–solutions fostered by agile markets, mer-
recordings of interviews). The field data collected gers, and integrated technological solutions.
were analyzed to determine constraint–resource asso-
ciations and determine critical relationships. In addi- 3.4. General decision-making approaches
tion, sporadic observations, conversations, and
interviews were conducted with operators acting in Based on these data, a model was developed and
the marketing function and support staff such as applied to a software architecture design and deve-
accountants or forecasters. lopment tool suite [52]. The emerging model evolved
Pointers to additional materials such as Standard based on the data and repeatedly applying some basic
Operating Procedures, Federal and State Law, organi- tenets of SE and IS analysis and design methods. The
zational tariffs, etc. were given to the researcher du- model needed to be iterative and incremental, i.e.,
ring the interactions with the informants. These concurrently modifying model, domain analysis, and
additional materials provided a quantifiable baseline tool design based on new data. Iteration allows for
in this domain but did not represent how things were analysis either with prior results based on the same
necessarily implemented or run. These ideal structures theoretical baseline or differing theoretical perspec-
and models inform us of the industry’s bshared under- tives. The incremental nature of the model is nothing
standingQ [52,95] in the same way Weber’s ideal new to decision-making since [54] noted it as both
organization [104] can be leveraged to understand bincrementalismQ [57] amounting to successive limi-
how real organizations work. The ideal structure ted comparisons, and blocal searchQ [22] that is sim-
must be compared with field data for a clearer view plistic reactionist managing. The model should reflect
of the critical relationships. the decision elements summarized by [54] (presented

Table 1
Elements highlight areas and issues in design decision-making
From Kleindorfer [54] From SE and IS
Decision elements Decision sciences Historical analysis and design
Problem finding What is the real problem? What is the actual problem domain?
Institutional Stakeholder analysis to Stakeholder and user analysis to determine
arrangements measure their goals, views, objectives, the real requirements
constraints, and agendas
Information Both objective and Circumscribe application and
gathering subjective facts, assumptions, and values problem domain
Choice process What approaches to choices should be Are the requirements valid and verifiable?
considered?
Implementation Specific approach using feedback, Traceability and model checking between
legitimization, and control through process stages
accounting
1000 R. Klashner, S. Sabet / Decision Support Systems 43 (2007) 990–1013

in the Table 1) and juxtaposed to traditional SE and IS


analysis and design to draw out their similarities. The Theory 1
System
five elements indicated in Table 1 are integrated and
Domain
throughout the DSS Design Model (presented later). Analysis

Fig. 2. Multiple data feeds into theory and analysis with feedback to
4. DSS Design Model the domain. Synchronous data or information exchange (Box #1)
between entities.

Designer researchers and practitioners can utilize


this new DSS Design Model framework to develop a that exist between the system domain component and
sense of systemic btrajectoryQ [41,42] associated with the information infrastructure component, which must
these implementations that evolve over decades. This be considered separately, but simultaneously during
understanding is necessary to address wicked pro- data collection.
blems [79] arising from domain constraints not typi- The research model has three primary interactions
cally a concern of designers [52,53]. The research with auxiliary feedback loops that are necessary based
process we followed was based on the theoretical on the underlying CAS theoretical framework. The
and methodological framework presented earlier. applied theoretical constructs (e.g., descriptive
The systemic considerations when developing this approaches such as ethnography or Grounded theory)
model are a reflection of these approaches. We have determine initial incremental steps through the model,
generalized the research process as a model while but later increments are guided by the inherent feed-
simultaneously validating and evolving it (see Discus- back loops. The incremental relationships in the
sion). The five decision elements (Table 1) are inte- model components allow testing of theoretical con-
grated throughout the DSS Design Model, but are jectures without jeopardizing the actual operation of
most heavily concentrated around and within the the- the infrastructure, thus satisfying a fundamental con-
ory and analysis component (shown later). A unified straint from the domain. Also, the increments create
graphical representation (Fig. 4) of the DSS Design temporal milestones. This process necessitates the
Model is shown at the end of this section. utilization of a nondeterministic and descriptive
Conventional infrastructure (e.g., electric power, SDLC. The combination of these conceptual con-
telecommunications) generates data from the physical structs provides a great deal of intellectual leverage
urban infrastructure and the ICT that is utilized in the over the process, but also greatly increases the com-
operations and accounting of the built infrastructure plexity of the seemingly simple DSS Design Model.
(Figs. 1 and 4). These two legacy systems should be The remainder of the model consists of the follow-
modeled as separate but tightly coupled components ing components: theory and analysis, simulation, de-
(i.e., represented in the image by them touching in a cision/design. Two-headed arrows between model
stacked formation) to demonstrate the strong depen- components (Figs. 2–4) indicate an interaction leading
dency between them. A great deal of the morphogenic to information exchange in both directions. The dotted
evolution is a direct result of symbiotic relationships line indicates an interaction that can be a real-time

Theory
System and
Analysis
Domain

Information Simulation
Infrastructure Capability

Fig. 3. Simulation: the added component interacts with theory and


Fig. 1. The real-world; ICT supporting urban infrastructure. analysis (Box #2).
R. Klashner, S. Sabet / Decision Support Systems 43 (2007) 990–1013 1001

1
Theory System
and Domain
Analysis
Information
2 Decision
Infrastructure

Simulation 3 4
Capability

Fig. 4. DSS Design Model. This visual representation infers both iteration and incremental activities, but no predetermined pattern is inferred—
only a general morphogenic process contained within the CAS theoretical framework.

synchronous exchange of information. We will use the standingQ, to achieve ba solid foundation for new
interactions between the DSS Design Model compo- knowledge generationQ [66]. In our DSS Design
nents to discuss these components and model speci- Model, if the introspective analysis indicates the ne-
fics, i.e., data–theory interaction; simulation–theory cessity to use a different theory, it is integrated within
interaction; decision/design interaction. the CAS framework and data collection resumes.
The relation between the theory component and
4.1. Phase I: data–theory interaction domain (Fig. 2) is maintained to continuously inte-
grate emerging data and update the theory (e.g., if a
The research is driven by the available data. Grounded theoretic approach is utilized [70]). The
Qualitative techniques (e.g., ethnographic methods) real-time synchronous exchange of information
were used because of the lack of experience of the (represented by the dotted line and Box #1) can result
researcher and the deep domain knowledge of the in an immediate evolution of ideas, concepts, and
informants. This dichotomy facilitates good data viewpoints is possible, e.g., analysts/theoreticians or
collection using ethnographic methods. Based on practitioners meeting with domain experts to ex-
field data in the electric power industry, it became change ideas. Two other data feeds influence the
obvious that a conceptual framework (with the choice of theory.
associated technical mechanisms) was needed to A direct feedback arrow (curved arrow on top of
utilize numerous theoretical constructs due to the Figs. 2 and 4) from the information infrastructure
variety of data in the domain. Also, many qualita- indicates a purely technical data feed from the specific
tive methods (e.g., ethnomethodology [32]) do not DSS or general IS (e.g., integration testing). Based on
scale well, so there must be an underlying theoret- the CAS theoretical framework, we assert SDLC ac-
ical motivation for different data collection and tivities that are normally after design (e.g., testing,
analysis approaches during subsequent iterations deployment, maintenance) are not independent, but a
through the model. natural morphogenic outcome of the design activity.
Our DSS Design Model incorporates an introspec- Thus, the single feedback arrow from the information
tive analysis that maps well into the future DSS architecture component captures the necessary data
research agenda proposed by Nemati et al. [66]. Spe- for design from these concurrent or later SDLC activ-
cifically, they said, bone research area of DSS ities. More recent agile SDLC configurations (e.g.,
becomes the development of a set of theoretical foun- eXtreme Programming [5]) have radically rearranged
dations upon which to build future development and the order of activities such that testing is before
applicationsQ. They went on to further specify the design, which means the testing becomes an aspect
combining of theoretical perspectives, such as of the theoretical analysis from a software engineering
bGestalt theory of insight [39 [sic]] combined with perspective (e.g., software engineering) and the sim-
Newell and Simon’s [46 [sic]] theoryQ or bCognitive ulation of component-based techniques are applied.
dissonance and Perkin’s [51 [sic]] theory of under- The second data feed into the theory and analysis are
1002 R. Klashner, S. Sabet / Decision Support Systems 43 (2007) 990–1013

from the simulation component described in the next head to theory/analysis component). The utilization of
subsection. domain-specific knowledge in [7] has similarities with
this DSS Design Model. Typically, the GDCC simu-
4.2. Phase II: simulation–theory interaction lation resulted in an evolutionary prototype [86] sim-
ilar to that used in [64] for software architecture
The integrated theories (i.e., in theory component) modeling, which demonstrates another aspect of the
at any particular increment during an iteration of the technical feasibility of the theoretical synthesis under-
DSS Design Model dictates many aspects of the lying the DSS Design Model.
overall model execution. Obviously, the choice of The simulation and theoretical components lever-
theories should be predicated on the research goals, age the flexibility provided by modern software en-
existence of prior collected data, and results from gineering and software architectures techniques to
previous theoretical analysis, all of which may or integrate theoretically grounded assertions with the
may not be consistent with the current theoretical data feed. Emerging technologies are used to dynam-
component. Since the type of data collection and ically replace and integrate computational compo-
analysis techniques are mandated by the theory, it is nents within systems. Thus, from a fundamental
only logical these same constraints will guide the inception of the DSS, the design incrementally
simulation design. The basic CAS tenets of integra- becomes more like the real-world artifact. To make
tion and morphogenesis through interpretation of do- the transition theoretically sound, we have leveraged
main variety captured in feedback mechanisms factor a holistic simulation theory developed by Zeigler
heavily into the DSS Design Model capability to during the General Systems Theory era, but evolved
oscillate somewhere between the conceptual opposites to its current capability [108]. One reason for the
of organized complexity and chaotic complexity, e.g., choice of Zeigler theory was its natural overlap
to a lesser degree, but somewhat analogous to pre- with the CAS theoretical framework such as the
scriptive and descriptive SDLC. shared morphogenic properties and component-
The morphogenetic approach utilized in this re- based approach to modeling. However, the transition
search resulted in the modification of the design from simulation theory (e.g., Zeigler) to component-
model itself through the addition of the simulation based prototype creates a need to change theoretical
component to the model (Figs. 3 and 4). Decision- perspectives. In addition, some design recommenda-
making processes often employ some sort of critical tions from the simulation or software component
abstraction akin to simulation to reduce the informa- architecture may be self-evident for use in the deci-
tion overload most humans suffer with in mission sion process, i.e., not requiring theoretically based
critical domains. For example, the California Energy analysis) resulting in a decision/design feedback for
Commission has a large contingent of analysts that the human actors described next.
use data in mathematical models and simulations in
order to advise policy decision-makers. Although sim- 4.3. Phase III: decision/design interaction
ulation is one of the common three modeling para-
digms (i.e., linear programming, simulation, and The cumulative effect of the theoretical analysis
spreadsheet models) [66], the field data indicated and/or simulation can combine to dynamically influ-
that it needed to take an even more prominent role ence the design decision-making process (Fig. 4, Box
due to the mission critical nature of build infrastruc- #3). Decision-makers such as IS analysts, software
ture and their dependence on simulation. engineers, project managers, and other primary stake-
Simulation is still used for information infrastruc- holders interpret these inputs based on their bshared
ture modeling, but with a more expanded role than understandingQ [52,95] of the relevant domain re-
typically ascribed to it. The simulation referred to in source constraints. Thus, the combination of theory
this case can be of the traditional type [56] or a type and simulation will inevitably impact the decisions,
modified to hide simulation complexity from a par- but only to the degree that it does not violate the
ticular stakeholder group [7], wherein the simulation decision-makers’ shared understanding of the design
output is analyzed (i.e., Fig. 3, Box #2 input arrow- goals.
R. Klashner, S. Sabet / Decision Support Systems 43 (2007) 990–1013 1003

These decisions may be made in real-time as The DSS Design Model takes a step toward
the various stakeholders observe the prototype as it correcting the mismatch between theory and design
is time-spliced into the working infrastructure as practice. We discuss how the model was validated
observed at the GDCC. By choosing different time through three separate activities: a software archi-
increments, their subjective impression of the de- tecture tool development; a microgrid design case;
sign will be altered in a nondeterministic manner and our current field work in the mission critical
because the data are live and random based on the telecommunications domain where we are designing
electric power customer base. For example, if a a Knowledge Management DSS (KMDSS) for net-
high-power transmission line goes offline due to work operators to use in alarm correlation.
an airplane accident while the prototype is running,
the software may crash bringing down the entire 5.1. Reduction of mismatches
system leaving a lasting impression that may kill
the development project. As noted earlier in the The DSS Design Model presented here seeks to
theoretical background, this systemic behavior is reduce the theory–design mismatch by tightly cou-
the nature of real-world design and, more general- pling the theoretical aspects of DSS design into the
ly, of wicked problems. The design decisions im- SDLC and implemented process. Our motivation for
pact the built information infrastructure (i.e., Fig. 4, this approach was domain instances. The FERC made
Box #4) that needs to be factored into a systemic deregulatory decisions regarding the electric power
model. industry based on several different groups of powerful
Newly integrated design decisions immediately lobbyists, committees, and political bodies. For exam-
impact the symbiotic relation between the infor- ple, the FERC made many of the specific design
mation infrastructure and the domain thereby com- decisions [67] based on bWhatQ and bHowQ panels
pleting the first full iteration since the general of industry experts. It is interesting to note the strong
iterative flow of data shown in Fig. 4 is counter- correlation with the traditional requirements engineer-
clockwise. Note, however, depending on the data, ing what–how rule of thumb [86]. Primary economic
the theory or theories utilized, and the type(s) of arguments came from the Energy Power Research
simulation/prototypes developed there may be an Institute (EPRI), the public at large, or economists
undetermined number of incremental steps in any [26,69,82].
particular iteration through the entire DSS Design Economists utilized various theoretical frame-
Model. works, such as Game Theory and Cournot–Nash
Equilibrium, to arrive at suggestions that deregula-
tors largely followed. Often, the economic argu-
5. Discussion ments were combined with other research
disciplines to adapt their theories to empirical
bIndeed, it is precisely in the exploration of Twicked
data. For instance, after observing certain market
problemsd for which conflicting or sparse theoretical
behavior, EPRI researchers [19] presented an eco-
bases exist that design research excels (March and
nomic market making argument leveraging Kirch-
Smith, 1995; Carroll and Kellogg, 1989)Q. V. Vaish-
off’s laws of electric power flow by developing an
navi and W. Kuechler [101].
engineering–economic model of the electric power
A basic disconnect exists between the theoretical network [94]. Local grid operations personnel, in
framework used by the electric power industry archi- contrast to these economic arguments, utilized ICT
tects and the IS designers who actually built the grid developed by IS analysts based more on a shared
information infrastructure. The mismatch between understanding of the electric power industry and a
their theories and reality demonstrates the need for heuristic approach that was more in line with how
DSS designers to pursue more comprehensive their people made decisions. Thus, the electric
approaches in order to correctly match or address power deregulation debacle in California resulted
the domain constraints that greatly contribute to the in part from this paradigm mismatch in DSS design
evolution of the system. that was directly tied to theoretical assumptions and
1004 R. Klashner, S. Sabet / Decision Support Systems 43 (2007) 990–1013

expectations not matching the ICT designs and face enumerate constraints that are then transformed
implementations in the field. into strongly typed objects within the space. Require-
ments engineers, system analysts, and/or software
5.2. The ArckBuilder case architects act as the mediators to guide the constraint
negotiations. Their expertise emerges through this
The DSS Design Model described earlier was process.
utilized as the basis from which to concurrently The CASA component is the heart of the Arck-
evolve a specific instance of a software architecture Builder suite because it uses the Harvester and
tool suite (ArckBuilder) [52]. The ArckBuilder im- CARESA components in the design process. An
plementation encourages a particular architectural up-to-date catalog is downloaded from the Harvester
process. ArckBuilder is presented here as an example server whenever the software architect needs a revi-
of how design research concepts and the results from sion. The service proxies are used in the canvas area
qualitative methods can be integrated into tool envir- of the tool to construct system architectures that
onments in the future within the CAS theoretical combine all types of resources. After computational-
framework. ly combining and instantiating the new component, it
Each domain resource has a set of behavioral con- appears in the canvas. Different components are
straints (e.g., specifications) that are captured in the combined using first-class architectural connectors,
resource service proxy, i.e., in a service-oriented archi- which themselves are derived from services (e.g.,
tectural style similar to [3]. As noted by [63], compo- telecommunication infrastructure for a distributed
nents can be evolved by modifying a subset of the system). The entire architecture can then be exported
bcomponent’s properties, e.g., interface, behavior, or as an executable file that dynamically simulates the
implementationQ. ArckBuilder understands and classi- aggregated service behavior.
fies those features of an architectural description that ArckBuilder makes use of a strongly typed
are placed in property lists. Thus, architecture con- programming language to maintain bshared under-
structed using service proxies must adapt to resource standingQ consistent, which is necessary as indicated
constraints using the properties associated with these by the data. The tool has been presented in a cursory
proxies. The existence of these proxies within the manner to demonstrate that it is possible to complete
domain guides and, where necessary, restrains the the process from qualitative data acquisition to design,
crosscutting influence by various stakeholder and then to a finished instantiated software architec-
demands. ture environment. Also, designing in the architectural
There are three ArckBuilder components that have context with trusted components greatly obfuscates or
the following functionality: Harvester gathers eradicates traditional SDLC stages. Once the essential
resources from the web; CARESA (Computer-aided design decisions arising from the theory and analysis
Requirements Engineering with Software Architec- are complete, the other activities such as testing,
tures) component addresses evolution, requirements implementation, and deployment occur in a nondeter-
negotiation, and stylistic constraints; and CASA (Com- ministic manner.
puter-aided Software Architecture) is a canvas type of DSS projects utilizing the DSS Design Model to
interface for design, technical collaboration, and sty- achieve similar functionality as the ArckBuilder de-
listic views. sign would provide the capability to mix and match
The Harvester component gathers resource pro- computational components derived from varying the-
xies from the dispersed domain (e.g., electric power oretical perspectives in order to juxtapose the results
industry). These proxies are connected to live or and guide decision-making. Utilizing the varying
simulated resources. Since the proxies are strongly apertures of differing theoretical lenses, architectural
typed interfaces, they can be categorized within the component/connector replacement and reuse provides
overall domain classification called a bcatalogQ [52]. something new for decision-makers—the ability to
The CARESA component is a use-case based inter- dynamically juxtapose service offerings concurrently
face that is the front end for a tuple space (e.g., with real-world observations of infrastructural beha-
Linda [76]). Stakeholders using the CARESA inter- vior in a design context.
R. Klashner, S. Sabet / Decision Support Systems 43 (2007) 990–1013 1005

5.3. Microgrid EMS-DSS analysis mini-case arrangements essentially redefine historical control
relationships creating service economies.
The DSS Design Model was conceptually vali- The bmicrogridQ (i.e., microcosm of the normal
dated in a project described by the following mini- grid) is a fundamental component in an experimental
case. We were investigating the representational suf- commercial park planned as a 200-acre, 2.4 million
ficiency of the model with analytical and field data square foot joint venture. The park design utilized a
entirely separate from the original field site location. microgrid that consisted of Distributed Energy
The following subsections correlate with the various Resources (DER), i.e., combined heat, cooling and
DSS Design Model components that were incremen- power (CHP), fuel cells, micro-turbine generators,
tally, concurrently, and iteratively put into operation in and photovoltaics, power storage facilities, and intel-
the field. An advanced energy management system ligent network devices. The design combined the
(EMS) DSS was needed to support a commercial microgrid with ICT to create niche services that
park’s coordinated electricity marketplace. The would result in an internal electric power marketplace.
EMS-DSS project was not completed due to killer
issues associated with high-level requirements [8,9] 5.3.2. Theory/analysis component: stakeholder data
(explained later), but the DSS Design Model was The CAS theoretical premises used in this case
introspectively evolved and its use during this process were the same as before. Initially, we conjectured
will demonstrate the flexibility in the model. Some most of the integration work would be technical and
specific constraints will be discussed below to elabo- would be relying on mechanical, civil, and other
rate the domain. engineering approaches. However, the inclusion of
economic theory was inevitable, but later iterations
5.3.1. Domain-information infrastructure component: of the model revealed an electricity auction market-
commercial park niche market place would be necessary with its associated theory
Regulators were trying to decrease the cost of [18,29]. As the project progressed, it also became
electricity for customers. Organizations were respond- apparent that the utilities utilized the interconnection
ing to pressures and incentives created by deregula- standards to maintain control over what DER could be
tion. Skyrocketing prices in California dramatized the used—even equipment entirely on private property
emerging competitive market. California’s high-tech needed their approval. The study of standards theories
industry is reflected in its load configuration that [25] was necessary because of the prominent role the
demands high-quality electricity. Evolving the com- interconnection issue assumed. Trust theories were
mon centralized distribution paradigm into a market under consideration because of the inevitable situation
by utilizing an ICT augmented microgrid is a logical, where marketplace competitors should change to col-
but controversial, outcome of deregulation. There laborators due to a disruptive microgrid state change
were numerous policies and laws that contradicted event. Also, requirements regarding interconnection
each other. In addition, contradicting hypothesis and dynamic architecture configuration management
could be deduced using economic policies associated capabilities were determined from stakeholders. The
with deregulation, microgrid economics [39], and in- theories guided data collection and requirements
terconnection standards (i.e., IEEE P1547 at that analysis, but did not completely alleviate conflicting
time). Policy makers (e.g., California and FERC reg- requirements, which fostered extensive complexity.
ulators; Department of Energy) were DER supporters. The data collection methods in this case were
Commercial parks specializing in niche markets similar to those used earlier, but with no observation
providing power quality and/or special pricing arrange- because the project was primarily conceptual. Data
ments to high-tech tenants could meet these new mar- collection began with the academic engineers to dis-
ket demands. Parks have several potential market cern the technical constraints and requirements asso-
offerings because they can act as value added reseller ciated with DER. Other park stakeholders were
of bancillary servicesQ [51,72,94] necessary in the included as we iterated through the model. The polit-
deregulated market. Tenants can outsource power pro- icalization of the microgrid project necessitated data
blems to park owners or their consulting firms. These collection from the California Energy Commission
1006 R. Klashner, S. Sabet / Decision Support Systems 43 (2007) 990–1013

and Department of Energy. Policy makers suggested use Game theory or Trust theories to determine how
that processes controlled by utilities (i.e., the DER much data to collect or provide through the service
interconnection) created market barriers by artificially proxies, which were entirely different considerations
inflating integration costs. Policy theory also created from software or mechanical engineering.
contradictions with electric power engineering meth-
ods. Thus, both streams of data were relevant. Utility 5.3.4. Decision/design component: multiple stake-
representatives were interviewed since it represented a holder agendas
major constraint in the project. Data were gathered EMS-DSS designers must make the decision as to
from DER vendor specifications and personnel who what theoretical perspectives to base future prototypes
were involved in their software engineering projects. upon—a wicked problem given the contradictions in
The first author participated in the IEEE interconnec- theory and data. The flexibility of component services
tion standardization process, which included docu- architectures partially offsets these decision comple-
mentation and meetings with all stakeholders xities. The design of the experimental park facility
present, to collect data. The engineers provided that was suggested to the stakeholders and agreed
DER prototype operational data and we collaborative- upon enabled the leveraging of service architectures.
ly developed simulations discussed next. The EMS-DSS design provided the capability to con-
nect the components into microgrid systems. This
5.3.3. Simulation component: triangulating the approach provides maximum flexibility for architec-
behaviors tural variety of microgrids in order to test various
We utilized three general types of simulations in market scenarios.
order to determine the behavior of the domain-specific The natural gas, electrical microgrid, and compu-
resources. The academic engineers and their graduate tational infrastructures had to accommodate DER in
students were currently utilizing hand-coded simula- cooperation with local gas and electric utilities. The
tions to model the various DER devices. The simu- eventual park design provided:
lation data were compared to the DER prototypes in
the laboratory and verified to be accurate. The second 1. Various alternative distribution options such as a
type of simulations was generated from commercial premium power circuit.
packages (e.g., Simulink by MathWorks) and were 2. A context dependent bridge between networks with
utilized to model common electric power and me- interlocking capabilities.
chanical engineering devices, which were assembled 3. Unified constant–frequency integration control
into larger components (e.g., as part of the intercon- devices [109].
nection). The third type of simulation was a service- 4. Multiple electric power network designs (radial or
oriented one developed as part of ArckBuilder (see network).
prior subsection) to model the entire commercial park 5. Multi-point power quality monitoring such as
and microgrid. Each resource owner (e.g., DER) power flow, or harmonics.
would have to develop a service proxy that could
then be assembled into an ArckBuilder model. The 5.3.5. Postmortem of the EMS-DSS: need to develop
service proxy to represent one of the other simula- foresight with theory
tions or a real-world artifact. Since the proxies were The microgrid project was more narrowly defined
written in Java and represented live services, they than the initial case because the academic engineers
could all be instantiated as Java objects and executed and policy makers wanted DER to succeed in the
simultaneously. park. However, it was no less bwickedQ due to un-
The DSS Design Model guided our decisions as to revealed agendas. System failures are often linked to
which services to create and how to design them undiscovered bkiller requirementsQ [8,9] that are in-
based on theories pertaining to domain constraints herently obfuscated because they are entangled in the
such as markets or auctions. Each theory was instruc- domain details. bHigh-level requirementsQ [8,9], on
tive about how to represent a resource as a service. the other hand, focus on other aspects of the process
Future work was discussed where we were going to to determine killer issues, such as political drivers.
R. Klashner, S. Sabet / Decision Support Systems 43 (2007) 990–1013 1007

We were unable to solve these high-level require- from command and controls centers, but with narrow
ments problems as described next. applicability and limited success. ACT must be inte-
The utility was a mandatory stakeholder due to the grated with current ICT deployed in the global net-
legacy policies. Their behavior was inconsistent with work to provide bclearerQ solutions to network faults
their own SOP, which indicated they were treating the than is currently available. The tool must be
park in a special manner, possibly due to the market designed using a global, sociotechnical, and geopo-
threat an independent park running their own micro- litical awareness [80]. Due to the application of the
grid represented. One instance of an artificial barrier is DSS Design Model, we have determined that the
electric power network data collection. They insisted ACT must be designed as a KM augmented DSS
the utility would install the data collection devices, [80].
but refused to put in enough devices to facilitate the
accurate architectural reconfiguration of the microgrid 5.4.1. Domain-information infrastructure component:
by the EMS-DSS. Also, we discovered late in the tracking faults around the globe
SDLC that the DER vendors were competing with A Network Management System (NMS) moni-
each other to develop proprietary EMS-DSS of their tors the global optical network domain we are
own. Thus, they would not share interface information examining. The field environment forces significant
necessary to build the appropriate service proxies. The constraints on the design of the system. The global
inevitable result was the project stalled and the park optical networks in question are mission critical and
owner, although already putting in additional built provide vital aspects of national security. Node-to-
infrastructure for the experiment, lost incentive to node communications used by multiple instances of
continue because of the economic downturn changed the ACT are facilitated via the network manage-
their agenda too. ment Data Communications Network (DCN). Final-
These types of deviant behavior can be addressed ly, the ACT must be designed to work at the
with the appropriate sociological theory but is not various levels of hierarchical reporting schemas
easily teased out a priori even when killer require- available in the network. This hierarchy is based
ments are suspected. Although the CAS theoretical on the principle that a central Network Management
framework and the DSS Design Model enable multi- Center (NMC) has overall coordination responsibil-
ple theory utilization, it will take further research to ity for the entire global network, i.e., the operation
determine how to combine and compare theoretical of every node. Thus, inter-node activities are mon-
perspectives, especially when the results are contra- itored and coordinated through the NMC. This
dictory. We discuss current research next, which is constraint implies that the ACT must function at
intended to further validate our past work in the a local jurisdiction within a node while still sup-
electric power industry. porting overall network end-to-end fault diagnosis
at the NMC level: i.e., a requirement not met by
5.4. A KMDSS for telecomm network operators prior industry tools.
The operators of these networks who will use ACT
Strategic planning promises to become even more
are comprised of loosely coupled, inter-organization-
complex in the future, as the Internet and telecommu-
al, virtual teams distributed in various geopolitical
nications technology will allow more organizations to
domains. These users must coordinate their efforts
become global in nature, and suppliers, producers and
and share knowledge premised on the emerging tech-
customers will become more closely connected
nical and social state of affairs within the context of
throughout the world. J. Courtney [20; Section 2.2,
each nation’s security. Also, members of the teams
pp. 20 (i.e., pp. 4 of 22)].
vary greatly in experience, prior knowledge, and the
We are currently using the DSS Design Model for diagnosis process they have adopted.
a semi-automated fault diagnosis system [65]: the
Alarm Correlation Tool (ACT). These types of sys- 5.4.2. Theory/analysis component
tems have traditionally assisted operators in main- An extensive review of the various correlation
taining global optical telecommunications networks algorithms available in the literature is provided by
1008 R. Klashner, S. Sabet / Decision Support Systems 43 (2007) 990–1013

[65]. The initial iterations of the DSS Design Model 5.4.3. Simulation component: merging prototypes with
were largely influenced by traditional DSS design existing applications
theory [77,78,88,40] because of the existence of a It is imperative to simulate the behavior of the
well-defined application domain. However, after fol- correlation algorithms and specific fault scenarios to
lowing the model’s guidelines, a different set of verify that the tool is operating properly. Simulation is
problems was identified. The data collection and necessary because the ACT operates on real-time
analysis from domain feedback showed that there alarm data retrieved from the NMS that exhibit cha-
was a need to address: the geopolitical aspects in racteristic, uncharacteristic, and false positive beha-
the domain; knowledge capture; knowledge transfer vior. Telecommunications suppliers in industry have
between command and control centers; a wide vari- progressively developed complex transmission and
ety of users. The existing DSS theories we were equipment simulators that provide a basis for testing
using did not address these challenging demands. theories and hypotheses as necessary. These commer-
During a subsequent iteration, we have explored cial simulators can be interchanged as complete com-
aspects of the Emergent Knowledge Processes theo- ponents per the DSS Design Model. The simulators
retical perspective [59] because it has several princi- can be exchanged in parallel with the various algo-
ples we are finding useful, such as the emergent rithms in the ACT prototypes to better understand the
nature of knowledge sharing and dealing with optimal configuration.
bnaı̈veQ users given the unpredictability of the global
user base for the ACT. A rigorous, multiple field site 5.4.4. Decision/design component: extending their
evaluation of the ACT prototypes will generate em- ability to share knowledge
pirical data in future iterations of the DSS Design The first iteration of the design model resulted in
Model. a simple correlator system to assist in local diagno-
Data were collected for use with rule-based cor- sis; i.e., it did not include knowledge management
relation algorithms that were the first attempt at an techniques in the design. The next iteration included
ACT design solution. Further analysis by domain the use of rule-based correlation only. Through simu-
experts showed that determining all rules covering lations and domain expert feedback, it was deter-
all possible fault scenarios would prove prohibitive. mined that a model-based addition would allow for
Subsequent DSS Design model iterations demon- capturing more scenarios with less user effort or
strated a model-based approach appeared to capture cognitive load. However, field data indicated man-
most common scenarios, while rules could be inte- agement of knowledge to be an important factor in
grated into the design in order to capture the more daily operations. Therefore, next iteration introduced
extraneous scenarios. During future iterations, we a KM design aspect that allowed for sharing of
will juxtapose theories regarding correlation algo- acquired knowledge. The current design iteration is
rithms with other relevant theories to perform data a modular, hierarchical alarm correlation system. We
analysis. are also working on the new requirement to provide
The ACT is intended to integrate with the currently knowledge sharing across the entire network. This
deployed NMS. Therefore, the ACT must interface to latest design activity requires the exploration of the-
the local NMS to retrieve both fault data and network ories of knowledge to inform our existing kernel
topology models. Every network node runs an in- theories. The CAS theoretical framework facilitates
stance of the ACT. As local engineers gain experi- this integration because it separates the concepts of
ence, they will theoretically be able to capture this data and information, thereby lending itself to
knowledge in the form of updated models and/or rules knowledge theories.
in the system. An instance of ACT will distribute By utilizing this new DSS Design Model, we have
these data and knowledge to other nodes, where evolved the ACT design while investigating the ap-
other ACT instances are running. Finally, at the propriate algorithms and architectural combinations to
NMC, a bhigher-levelQ ACT must be able to integrate deploy in the system. We are able to analyze various
root causes determined by each node’s ACT to create algorithms and use simulation techniques to verify
end-to-end root cause analysis. their proper behavior. Furthermore, by taking into
R. Klashner, S. Sabet / Decision Support Systems 43 (2007) 990–1013 1009

account the domain constraints as well as the existing with certain basic similarities to the approach sug-
ICT infrastructure, the design was augmented to allow gested by Nemati et al. [66].
for more successful integration in the field. In addi- The new DSS Design Model was devised during a
tion, we are able to progressively integrate more period of great upheaval associated with electric
appropriate theories for design of the ACT KMDSS. power industry deregulation spurred on by increasing
sociotechnical dependencies on Internet technologies
[53]. To further validate the model and test its general
6. Conclusion applicability to other mission critical infrastructure,
we are conducting a more narrowly focused develop-
The new DSS Design Model presented in this ment of a KMDSS for use in global telecommunica-
paper has evolved over time to address the bwickedQ tions fault correlation.
design problem we became aware of during our initial
field study in the complex, mission critical electric
power industry. The model reflects a long process of Acknowledgements
data collection and validation within a Complex
Adaptive Systems (CAS) theoretical framework that The authors wish to especially thank John L. King,
itself proposes the integrated use of numerous theories Dean of the School of Information, University of
and approaches. Utilizing the CAS framework has Michigan, for his support and many insights includ-
resulted in a multidimensional approach to design ing, but not limited too, the very essence of the design
research that leverages theoretical concepts, ranging model presented in this paper. We would like to
from deep qualitative methods to software engineer- acknowledge and thank Professor Jerry Fjermestad
ing practices, in various complimenting ways. This for his insights and recommendations. Also, we are
paper has explained the model using following con- very grateful to the anonymous reviewers for their
texts: conceptual rigor, technical instantiation, mini- guidance.
case study, and ongoing empirical evaluation. Using
this model in an adaptive manner should experiential-
ly lead researchers to a new understanding of what References
DSS design must become.
The new DSS Design Model suggests future de- [1] C.Y. Baldwin, What can the social sciences gain from the
sign processes will be evolutionary and iteratively science of design? in: K. Sullivan (Ed.), NSF Workshop on
the Science of Design: Software and Software-Intensive Sys-
encompass dynamic theory–simulation–and decision- tems, NSF, Airlie Center, VA, 2003, p. 2.
making interactions. Although simple in appearance, [2] R. Banach, M. Poppleton, Retrenching partial requirements
when the DSS Design Model is understood within into system definitions: a simple feature interaction case
the CAS framework, it assumes a dynamic complex- study, Requirements Engineering 8 (4) (2003 November)
ity that no prescriptive software development life 266 – 288.
[3] L. Baresi, R. Heckel, S. Thöne, D. Varró, Modeling and
cycle (SDLC) or process is capable of mastering. validation of service-oriented architectures: application vs.
The new DSS Design Model is conceptually a multi- style, Proceedings of the 9th European Software Engi-
threaded configuration with dynamic runtime behav- neering Conference and 10th ACM SIGSOFT Founda-
ior playing out in concurrent operations. So far, we tions of Software Engineering, ACM, Helsinki, Finland,
2003, pp. 68 – 77.
have only been able to instantiate this new DSS
[4] L. Bass, P. Clements, R. Kazman, Software Architecture in
Design Model using software architecture research Practice, Addison-Wesley, Reading, MA, 1998, p. 452.
techniques that have the capability for stylistic adap- [5] K. Beck, M. Fowler, Planning Extreme Programming, Addi-
tation to domain-specific complexity. Another advan- son-Wesley Publishing Co, Reading, MA, 2000, p. 160.
tage of including software architecture methods is [6] M. Begovic, D. Novosel, M. Milisavljevic, Trends in power
that we were then able to analyze DSS prototype system protection and control, Decision Support Systems 30
(2001) 269 – 278.
behavior and interactions based on various theoreti- [7] R. Belz, P. Mertens, Combining knowledge-based systems
cal perspectives through component services. There- and simulation to solve rescheduling problems, Decision
fore, we have instantiated a multi-theoretic process Support Systems 17 (2) (1996 May) 141 – 157.
1010 R. Klashner, S. Sabet / Decision Support Systems 43 (2007) 990–1013

[8] M. Bergman, J. King, K. Lyytinen, Large scale requirements [27] S. Easterbrook, Model management and inconsistency in
analysis as heterogeneous engineering, in: C. Floyd, R. software design, in: K. Sullivan (Ed.), NSF Workshop on
Klischewski (Eds.), Social Thinking—Software Practice, the Science of Design: Software and Software-Intensive Sys-
MIT Press, Cambridge, MA, 2000. tems, NSF, Airlie Center, VA, 2003, p. 2.
[9] M. Bergman, J.L. King, K. Lyytinen, Large scale require- [28] R.T. Fielding, R.N. Taylor, Principled design of the modern
ments analysis revisited: the need for understanding the web architecture, Proceedings of the 22nd International Con-
political ecology of requirements engineering, Requirements ference on Software Engineering, Limerick, Ireland, 2000.
Engineering 7 (3) (2002) 152 – 171. [29] D. Friedman, J. Rust, The Double Auction Market: Institu-
[10] D.M. Berman, J.T. O’Connor, Who Owns the Sun: People, tions, Theories, and Evidence, Addison-Wesley Pub. Co,
Politics, and the Struggle for a Solar Economy, Chelsea Reading, MA, 1993, p. 429.
Green Pub. Co, White River Junction, VT, 1996, p. 331. [30] B.G. Glaser, A.L. Strauss, Discovery of Grounded Theory:
[11] B. Boehm, A Spiral model of software development and Strategies for Qualitative Research, Aldine de Gruyter, 1967,
enhancement, IEEE Computer 21 (5) (1988 May) 61 – 72. p. 271.
[12] B.W. Boehm, P.N. Papaccio, Understanding and controlling [31] J.A. Goguen, Formality and informality in requirements en-
software costs, IEEE Transactions on Software Engineering gineering, Proceedings of the Fourth International Confer-
14 (10) (1988) 1462 – 1477. ence on Requirements Engineering, IEEE Computer Society,
[13] T.J. Brennan, A Shock to the System: Restructuring Amer- 1996, pp. 102 – 108.
ica’s Electricity Industry, Resources for the Future, Washing- [32] J. Goguen, An introduction to algebraic semiotics, with
ton, DC, 1996, p. 138. applications to user interface design, in: C.L. Nehaniv
[14] F.P. Brooks, Is there a design of design? in: K. Sullivan (Ed.), (Ed.), Computation for Metaphors, Analogy, and Agents,
NSF Workshop on the Science of Design: Software and Soft- Springer, New York, NY, 1999, pp. 242 – 291.
ware-Intensive Systems, NSF, Airlie Center, VA, 2003, p. 2. [33] S.J. Greenspan, J. Mylopoulos, A. Borgida, Capturing more
[15] W.F. Buckley, Sociology and Modern Systems Theory, Pren- world knowledge in the requirements specification, Proceed-
tice-Hall, Englewood Cliffs, NJ, 1967, p. 227. ings of the Sixth International Conferences of Software En-
[16] W.F. Buckley, Society—A Complex Adaptive System: gineering, 1982, pp. 225 – 234.
Essays in Social Theory, Gordon and Breach, Australia, [34] S.J. Greenspan, J. Mylopoulos, A. Borgida, On formal
1998, p. 312. requirements modeling. Languages: RML Revisited, Pro-
[17] G. Burrell, G. Morgan, Sociological Paradigms and Organi- ceedings of the International Conferences of Software Engi-
sational Analysis: Elements of the Sociology of Corporate neering, 1994, pp. 225 – 234.
Life, Heinemann, London, 1979, p. 432. [35] M. Haglind, L. Johansson, M. Rantzer, Experiences inte-
[18] J.B. Bushnell, S.S. Oren, Internal auctions for the efficient grating requirements engineering and business analysis an
sourcing of intermediate products, Journal of Operations empirical study of operations and management system
Management 12 (3,4) (1995) 311 – 320. procurement, International Conference on Requirements
[19] H. Chao, S. Peck, A market mechanism for electric power Engineering (ICRE ’98), IEEE, Colorado Springs, CO,
transmission, Journal of Regulatory Economics 10 (1996) 1998, p. 108.
25 – 59. [36] J.D. Herbsleb, A. Mockus, Formulation and preliminary test
[20] J.F. Courtney, Decision making and knowledge management of an empirical theory of coordination in software engineer-
in inquiring organizations: toward a new decision-making ing, Proceedings of the 9th European Software Engineering
paradigm for DSS, Decision Support Systems 31 (1) (2001 Conference and 10th ACM SIGSOFT Foundations of Soft-
May) 17 – 38. ware Engineering, Helsinki, Finland, 2003.
[21] B. Curtis, H. Krasner, N. Iscoe, A field study of the software [37] A. Hevner, S. March, J. Park, S. Ram, Design science in
design process for large systems, Communications of the information systems research, MIS Quarterly 28 (1) (2004
ACM 31 (11) (1988 November) 1268 – 1287. March) 75 – 105.
[22] R. Cyert, J. March, A Behavioral Theory of the Firm, Pren- [38] S.R. Hiltz, M. Turoff, Structuring computer-mediated com-
tice-Hall, Englewood Cliffs, NJ, 1963, p. 252. munication systems to avoid information overload, Commu-
[23] A. Dandekar, E. Perry, Barriers to effective process architec- nications of the ACM 28 (7) (1985 July) 680 – 689.
ture—an experience report, Software Process Improvement [39] T.E. Hoff, H.J. Wenger, C. Herig, J. Robert, W. Shaw,
and Practice 2 (1) (1996 March) 13 – 20. Distributed generation and micro-grids, Proceedings of
[24] A. Dardenne, A. van Lamsweerde, S. Fickas, Goal-directed the 18th Annual North American Conference of the US
requirements acquisition, Science of Computer Programming Association for Energy Economics, San Francisco, CA,
20 (1993) 3 – 50. 1997.
[25] P.A. David, W.E. Steinmueller, Standards, trade and competi- [40] C.W. Holsapple, Knowledge management support of deci-
tion in the emerging global information infrastructure envi- sion making, Decision Support Systems 31 (1) (2001 May)
ronment, Telecommunications Policy 20 (1996) 817. 1 – 3.
[26] M.J. Denton, S.J. Rassenti, V.L. Smith, S.R. Backerman, [41] T.P. Hughes, Networks of Power: Electrification in Western
Market power in a deregulated electrical industry, Decision society, 1880–1930, Johns Hopkins University Press, Balti-
Support Systems 30 (2001) 357 – 381. more, 1983, p. 474.
R. Klashner, S. Sabet / Decision Support Systems 43 (2007) 990–1013 1011

[42] T.P. Hughes, Edison and electric light, in: D. MacKenzie, Requirements Engineering: Social and Technical Issues, Aca-
J. Wajcman (Eds.), The Social Shaping of Technology, Open demic Press, London, 1994, p. 296.
University Press, Bristol, PA, 1992, pp. 39 – 53. [61] J. McDermid, Science of software design: architectures
[43] M. Jackson, System Development, Prentice-Hall, Englewood for evolvable, dependable systems, in: K. Sullivan (Ed.),
Cliffs, NJ, 1983, p. 418. NSF Workshop on the Science of Design: Software and
[44] M. Jackson, Software Requirements and Specifications: A Software-Intensive Systems, NSF, Airlie Center, VA,
Lexicon of Practice, Principles and Prejudices, ACM Press/ 2003, p. 2.
Addison-Wesley Publishing Co., New York, NY, 1995, p. 264. [62] D.W. McDonald, Supporting nuance in groupware design,
[45] M. Jackson, A science of software design? in: K. Sullivan Moving from Naturalistic Expertise Location to Expertise
(Ed.), NSF Workshop on the Science of Design: Software Recommendation, Information and Computer Science, Uni-
and Software-Intensive Systems, NSF, Airlie Center, VA, versity of California, Irvine, CA, 2000.
2003, p. 2. [63] N. Medvidovic, R.N. Taylor, A classification and comparison
[46] I. Jacobson, G. Booch, J. Rumbaugh, The Unified Software framework for software architecture description languages,
Development Process, Addison-Wesley, New York, NY, IEEE Transactions on Software Engineering 26 (1) (2000)
USA, 1999, p. 463. 70 – 93.
[47] M. Jarke, J. Bubenko, C. Rolland, A. Sutcliffe, Y. Vassiliou, [64] N. Medvidovic, D.S. Rosenblum, R.N. Taylor, A language
Theories underlying requirements engineering: an overview and environment for architecture-based software develop-
of NATURE at Genesis, Proceedings of the IEEE Sympo- ment and evolution, Proceedings of the 21st International
sium on Requirements Engineering, RE’93, San Diego, Cali- Conference on Software Engineering (ICSE’99), Los
fornia, 1993. Angeles, CA, 1999, pp. 44 – 53.
[48] M. Jirotka, Tutorial on video supported ethnography for [65] D.M. Meira, A Model for alarm correlation in telecommuni-
requirements analysis, International Conference on Require- cations networks, Computer Science, Institute of Exact
ments Engineering (ICRE), Colorado Springs, CO, 1998. Sciences (ICEx) of the Federal University of Minas Gerais,
[49] M. Jirotka, J. Goguen, Requirements Engineering: Social and Belo Horizonte, Brazil, 1997, p. 149.
Technical Issues, Academic Press, London, 1994, p. 296. [66] H.R. Nemati, D.M. Steiger, L.S. Iyer, R.T. Herschel, Knowl-
[50] H. Kaindl, S. Brinkkemper, J.A.J. Bubenko, B. Farbey, S.J. edge warehouse: an architectural integration of knowledge
Greenspan, C.L. Heitmeyer, J.C.S.d.P. Leite, N.R. Mead, J. management, decision support, artificial intelligence and data
Mylopoulos, J. Siddiqi, Requirements engineering and tech- warehousing, Decision Support Systems 33 (2002) 143 – 161.
nology transfer: obstacles, incentives and improvement agen- [67] Open access same-time information system and standards of
da, Requirements Engineering 7 (2002) 113 – 123. conduct, In Federal Energy Regulatory Commission OASIS
[51] A. Keyhani, A. Kian, J.C. Jr., M.A. Simaan, Market moni- NOPR 889, RM95-9-000, (1996) 145.
toring and control of ancillary services, Decision Support [68] P. Oreizy, M.M. Gorlick, R.N. Taylor, D. Heimbigner, G.
Systems 30 (2001) 255 – 267. Johnson, N. Medvidovic, A. Quilici, D.S. Rosenblum, A.L.
[52] R. Klashner, Using Architecture Style to Design and Evolve Wolf, An architecture-based approach to self-adaptive soft-
Complex Integrated Information Infrastructure, Information ware, IEEE Intelligent Systems 14 (1999) 54 – 62.
and Computer Science, University of California, Irvine, CA, [69] S.S. Oren, Economic inefficiency of passive transmission
2002, p. 228. rights in congested electricity systems with competitive gen-
[53] R. Klashner, ICT and the deregulation of the electric power eration, Energy Journal 18 (1) (1997) 63 – 83.
industry: a story of an architect’s new tool, Journal of Digital [70] W.J. Orlikowski, CASE tools as organizational change: in-
Information (JoDI), To appear in a Special issue on Social vestigating incremental and radical changes in systems de-
Aspects of Digital Information in Perspective, 2004. velopment, MIS Quarterly 17 (3) (1993) 309 – 340.
[54] P.R. Kleindorfer, H.G. Kunreuther, P.J.H. Schoemaker, De- [71] L.J. Osterweil, Software processes are software too, revisited:
cision Sciences, Cambridge University Press, 1993, p. 484. an invited talk on the most influential paper of ICSE 9, 19th
[55] P.B. Kruchten, The 4+1 view model of architecture, IEEE International Conference on Software Engineering, ACM
Software 12 (1995) 42. Press, Boston, MA, 1997, pp. 540 – 548.
[56] A.M. Law, W.D. Kelton, Simulation Modeling and Analysis, [72] T.J. Overbye, J.D. Weber, K.J. Patten, Analysis and visuali-
McGraw-Hill, Boston, 2000, p. 760. zation of market power in electric power systems, Decision
[57] C.E. Lindblom, The science of muddling through, Public Support Systems 30 (2001) 229 – 241.
Administration Review 19 (1959) 79 – 88. [73] C. Pahl, M. Casey, Ontology support for web service
[58] P. Mangiameli, D. West, R. Rampal, Model selection for processes, Proceedings of the 9th European Software En-
medical diagnosis decision support systems, Decision Sup- gineering Conference and 10th ACM SIGSOFT Founda-
port Systems 36 (3) (2004 Jan) 247 – 259. tions of Software Engineering, Helsinki, Finland, 2003,
[59] M.L. Markus, A. Majchrzak, L. Gasser, A design theory for pp. 208 – 216.
systems that support emergent knowledge processes, MIS [74] D.E. Perry, Software evolution and blight Q semantics, Pro-
Quarterly 26 (3) (2002 September) 179 – 212. ceedings of the 21st International Conference on Software
[60] J.A. McDermid, Requirements analysis: orthodoxy, funda- Engineering, Los Angeles, CA, IEEE Computer Society
mentalism, and heresy, in: M. Jirotka, J. Goguen (Eds.), Press, 1999, pp. 587 – 590.
1012 R. Klashner, S. Sabet / Decision Support Systems 43 (2007) 990–1013

[75] D.E. Perry, A.L. Wolf, Foundations for the study of software [93] R.N. Taylor, N. Medvidovic, K.M. Anderson, E.J. Whitehead
architecture, ACM SIGSOFT Software Engineering Notes 17 Jr., J.E. Robbins, K.A. Nies, P. Oreizy, D.L. Dubrow, A
(4) (1992 October). component- and message-based architectural style for GUI
[76] G.P. Picco, A.L. Murphy, G.-C. Roman, LIME: Linda software, IEEE Transactions on Software Engineering 22 (6)
meets mobility, Proceedings of the 21st International Con- (1996 June) 390 – 406.
ference on Software Engineering, Los Angeles, CA USA, [94] R.J. Thomas, Restructuring the electric power business—a
1999, pp. 368 – 377. marriage of power engineering and market economics, Deci-
[77] D. Power, Building knowledge-driven DSS and mining data, sion Support Systems 30 (3) (2001) 227 – 228.
Decision Support Systems, HyperBook, 2000. [95] A. Tiwana, B. Ramesh, A design knowledge management
[78] D.J., Power, A brief history of decision support systems, system to support collaborative information product evo-
accessed in 02/18/04, 2004, DSSResources.COM, Electronic lution, Decision Support Systems 31 (2) (2001 June)
source: http://DSSResources.COM/history/dsshistory.html. 241 – 262.
[79] H.W. Rittel, J., M.M. Webber, Dilemmas in a general theory [96] W. Tracz, Confessions of a Used Program Salesman: Institu-
of planning, Policy Sciences 4 (1973) 155 – 169. tionalizing Software Reuse, Addison-Wesley Pub. Co, Read-
[80] S.A. Sabet, R. Klashner, Helping network managers meet ing, MA, 1995, p. 233.
service demands using DSS, The Proceedings of the Special [97] W. Tracz, L. Coglianese, P. Young, A domain-specific soft-
Interest Group on Decision Support, Knowledge and Data ware architecture engineering process outline, ACM Soft-
Management (SIGDSS) Workshop held in conjunction with ware Engineering Notes (1993 April) 40 – 49.
the International Conference on Information Systems [98] E. Turban, Decision Support Systems and Intelligent
(ICIS’03), Seattle, WA, 2003. Systems, Prentice-Hall Inc., Upper Saddle River, NJ,
[81] W. Scacchi, Process models in software engineering, in: J.J. 1998, p. 890.
Marciniak (Ed.), Encyclopedia of Software Engineering, [99] B.A. Turner, The use of Grounded theory for the qualitative
John Wiley and Sons, New York, NY, 2001. analysis of organizational behaviour, Journal of Management
[82] R.E. Schuler, Analytic and experimentally derived esti- Studies 20 (3) (1983) 333 – 348.
mates of market power in deregulated electricity systems: [100] M. Turoff, M. Chumer, B.V.d. Walle, X. Yao, The design of a
policy implications for the management and institutional dynamic emergency response management information sys-
evolution of the industry, Decision Support Systems 30 tem (DERMIS), Journal of Information Technology Theory
(2001) 341 – 355. and Application To appear, 2004.
[83] M. Shaw, D. Garlan, Software Architecture: Perspectives on [101] V. Vaishnavi, W. Kuechler, Design research in informa-
an Emerging Discipline, Prentice Hall, Upper Saddle River, tion systems, accessed in January 20, 2004, Electronic
NJ, 1996, p. 242. source: http://www.isworld.org/Researchdesign/drisISworld.
[84] H.A. Simon, Models of Bounded Rationality, MIT Press, htm, 2004.
Cambridge, MA, 1982. [102] Z.A. Vale, M.F. Fernandes, C. Rosado, A. Marques, et al.,
[85] H.A. Simon, The Sciences of the Artificial, MIT Press, Cam- Better KBS for real-time applications in power system con-
bridge, MA, 1994. trol centers: the experience of SPARSE project, Computers in
[86] I. Sommerville, Software Engineering, Addison-Wesley Pub. Industry 37 (2) (1998) 97 – 111.
Co, 2000, p. 693. [103] A. van Lamsweerde, Requirements engineering in the year
[87] I. Sommerville, T. Rodden, P. Sawyer, R. Bentley, M. Twi- 00: a research perspective, Proceedings of the 22nd Inter-
dale, Integrating ethnography into the requirements engineer- national Conference on Software Engineering, Limerick,
ing process, RE’93, San Diego CA, 1994. Ireland, 2000.
[88] R.H. Sprague, A framework for the development of decision [104] M. Weber, Max Weber on Law in Economy and Society,
support systems, MIS Quarterly 4 (4) (1980) 1 – 26. translated by E. Shils and M. Rheinstein, Simon and Schus-
[89] R.L. Sullivan, Power System Planning, McGraw-Hill, New ter, New York, NY, 1968.
York, NY, 1977, p. 324. [105] E.S.K. Yu, Towards modelling and reasoning support for
[90] A.G. Sutcliffe, N.A.M. Maiden, S. Minocha, D. Manuel, early-phase requirements engineering, The Proceedings of
Supporting scenario-based requirements engineering, IEEE 3rd IEEE International Symposium on Requirements Engi-
Transactions on Software Engineering 24 (12) (1998 Decem- neering, 1997, pp. 226 – 235.
ber) 1072 – 1088. [106] E.S.K. Yu, J. Mylopoulos, Towards modelling strategic actor
[91] W.J. Tauzin, August 14th Transcript of Midwest ISO con- relationships for information systems development—with
trol center from 1:00 to 5:00 pm Eastern Time. H.C.o.E.a, examples from business process reengineering, Proceedings
Commerce House of Representatives, Washington, DC, of the 4th Workshop on Information Technologies and Sys-
2003, p. 650. tems, WITS’94, Vancouver, BC, 1994.
[92] R.N. Taylor, W. Tracz, L. Coglianese, Software Development [107] J.A. Zachman, A framework for information system archi-
using Domain-Specific Software Architectures, A Curricu- tecture, IBM Systems Journal 26 (3) (1987) 454 – 470.
lum Module in the SEI style, University of California, Irvine, [108] B. Zeigler, T. Kim, H. Praehofer, Theory of Modeling and
1994. Simulation, Academic Press, 2000, p. 510.
R. Klashner, S. Sabet / Decision Support Systems 43 (2007) 990–1013 1013

[109] L. Zhou, K.M. Smedley, Unified constant–frequency integra- Sameh Sabet is a Distinguished Member of
tion control of active power filters, IEEE Applied Power Technical Staff at Tyco Telecommunica-
Electronics Conference, 2000, pp. 406 – 412. tions Laboratories. He is also an adjunct
professor at NJIT where he teaches require-
Robb Klashner is an Assistant Professor ments engineering. He is a PhD candidate
in the Information Systems Department, in Information Systems at NJIT. Previous-
College of Computing Sciences, at the ly, he was the Director of Network Sys-
New Jersey Institute of Technology. He tems Development for TyCom Labs. He
has a PhD in Information and Computer has over 10 years of experience in IT as
Science from the University of California well as software design, development and
Irvine. His current research interests in- real-time embedded systems design for
clude decision support systems, design of Network Management Systems in the telecommunications industry.
emergency response systems, theoretically His current research interests span augmenting alarm correlation
grounded analysis and design, mission with emerging decision support systems and knowledge manage-
critical infrastructure, integrated informa- ment research.
tion infrastructure, software architecture design tool environments,
software engineering, and requirements acquisition using qualita-
tive methods.

You might also like