Upload
k-n-papamichail
View
215
Download
0
Embed Size (px)
Citation preview
ORI GIN AL PA PER
Supporting distributed decision processesusing an evolution model
K. N. Papamichail Æ I. Robertson
Received: 7 September 2006 / Revised: 11 September 2007 / Accepted: 4 June 2008 /
Published online: 22 August 2008
� Springer-Verlag 2008
Abstract In recent years, emerging Information and Communication Technologies
have changed the nature and process of decision making. Decision processes are
often distributed, heterogeneous and subject to change. Business process modelling
is a key technology for analysing, representing and executing business processes. It
can be used to study distributed decision processes and improve decision making
practices. A theoretical model of the decision process has been developed in order to
better integrate the decision concept with models of business processes. The decision
model is executed using a process support system that provides a distributed Web
user interface. The enacted decision model evolves as the decision process pro-
gresses and supports decision makers if and when needed. The case study of an actual
decision process undertaken in a not-for-profit organisation is presented to highlight
the use, execution and validation of the decision model.
Keywords Business process modelling � Decision support systems �Distributed decision processes � Evolution � Process support technologies �Web-based decision support
1 Introduction
‘An executive decision is a moment in a process’, so stated Mary Parker Follett
(Follett 1941). This observation, which has unfortunately been largely ignored in the
K. N. Papamichail (&)
Manchester Business School, University of Manchester,
Booth Street East, Manchester M15 6PB, UK
e-mail: [email protected]
I. Robertson
School of Computer Science, University of Manchester,
Oxford Road, Manchester M13 9PL, UK
123
Oper Res Int J (2008) 8:279–297
DOI 10.1007/s12351-008-0019-1
intervening years, can refer to two contexts. The first is when the decision is a
commitment to action, and the commitment is only one element of the overall
process of making a decision. The second context is when the act of decision-
making may be only a small element in a very much larger process by means of
which an organisation fulfils its objectives. Contemporary activity in the area of
decision support has largely ignored the relevance of the business process to the
decision. The main focus of decision-aiding tool development has been on
supporting more elegant analyses and evaluation algorithms and integrated solutions
(i.e. one piece of software to cover all functionality).
It has been shown that the effectiveness of strategic decisions depends on the
decision steps that managers follow in order to reach them (Dean and Sharfman
1996). Particular conditions, i.e. patterns of relationships among variables such as
sufficient information and means of implementation, organisational type and
interference from the top can contribute to successful decisions (Rodrigues and
Hickson 1995). Decision processes are related to decision-specific and top-
management characteristics as well as contextual factors (Papadakis et al. 1998).
Even behavioural characteristics such as affective state or mood can affect the
decision process (Elsbach and Barr 1999). An analysis of a large number of
decision processes can help researchers and practitioners identify best and worst
practices in exploring alternatives and adopting a suitable decision approach (Nutt
2000, 2002).
Organisational decision makers often skip decision steps, execute decision steps
out of order and do not adopt a decision protocol (Mintzberg et al. 1976; Nutt 1984;
Hickson et al. 1986). Process support technologies could be used to support and
coordinate decision processes. Such technologies tend to support well-structured,
time-invariant and single-dimension workflows. However, workflows and decision
processes are influenced by actors whose behaviour is not deterministic (Volkner
and Werners 2002). More flexible technologies are therefore needed to coordinate
decision processes.
A framework called D2P has been developed to support decision processes in
organisational contexts (Oquendo et al. 2000). The D2P model represents the
decision making process at a high level. Its role is twofold: it prescribes how to take
well-researched decisions but it also describes how decision makers have actually
taken a decision (when the model is executed, it can evolve to reflect the way that a
decision process has unfolded).
In order to capture, analyse and execute complex decision making behaviour, we
have translated the framework into an enactable model that is implemented in
ProcessWeb (Warboys et al. 1999). ProcessWeb is a process support system that was
adopted for its capability to support complex processes and for its distributed Web-
based user interface. The system implements process models that are subject to
change and evolve over time.
We have explored the role of D2P in a range of settings. Previous work
investigated the use of D2P as a mechanism for understanding and influencing the
management control process within organisations (Papamichail and Robertson
2005). We have also explored the role of D2P in supporting societal decision
making (Papamichail and Robertson 2003, 2004) and e-tendering processes
280 K. N. Papamichail, I. Robertson
123
(Mohamad Noor et al. 2006). The next step is to use D2P to model a real-life
decision making process.
The concept of distributed decisions is of particular interest (Barthelemy et al.
2002). Information and Communication Technologies do not require decision
makers to be ‘statically located’ and have dramatically changed the nature and
process of decision making. This work examines the capability of D2P to adapt at
run-time in order to support distributed decision processes that involve various
participants and are subject to change. A case study is used to demonstrate the
applicability of our approach. The case study describes a decision process in a not-
for-profit organisation. This decision had to be arrived at within a specific time
frame and implementation also took place within a specific period.
The setting of the decision is a UK higher education institution. From time to
time, institutions are invited to submit bids for specific allocations of funds by their
principal funding council HEFCE (Higher Education Funding Council for England).
In responding to such requests for proposals, the institutions must identify an
individual to bring the action forward, formulate a response from identified
possibilities, and negotiate (to a greater or lesser extent) its acceptance among
concerned members of staff. The proposal, if accepted and successful in providing
funds, must also be capable of being implemented. In general, such responses are
undertaken over relatively short periods of time, making them useful for carrying
out longitudinal studies.
Of particular interest are the different models that are needed to convey
understanding of decision processes. Several modelling techniques have been
employed in this study to describe and view the HEFCE decision process (thereafter
referred to as HEFCE 00/56) from different perspectives. Some of these models can
be executed (or enacted) in order to support inherently complex human behaviour.
A core property of the models is that they are evolvable, that is, they are able to
support self-adaptation to a certain degree.
Unlike other studies that attempt to identify patterns in decision processes that
lead to successful decisions, the aim of this work is to help decision makers become
more competent at taking decisions through their understanding of the decision
process. Process support tools are employed to help actors analyse existing
processes and explore road maps, i.e. methods for creating new and more efficient
decision processes.
The structure of the paper is as follows. A review of approaches to modelling
decision processes is given in Sect. 2. Section 3 discusses our approach to
modelling decision processes and introduces a generic model whose purpose is to
support the decision-making process in organisations. The models used to analyse
the decision problem of our case study are described in Sect. 4. The relationships
between the different models of the process are discussed in Sect. 5. Section 6
describes the technology used to execute the model and Sect. 7 discusses our case
study and the implemented model. Section 8 highlights the relevance of the work to
researchers and practitioners. It is followed by the conclusions of the study in
Sect. 9.
Supporting distributed decision processes 281
123
2 Modelling decision processes
Organisational decision processes are complex phenomena and it is not always
possible to track them down and identify the points in time and space where
decisions were taken. Langley et al. (1995) refer to the case of a car manufacturing
company that hired consultants to investigate the introduction of a new model of a
car. However, it was difficult to determine who decided, what, where and how.
Various departments were involved in the process and after several very complex
decisions and actions the new model was produced.
Several contemporary descriptions of decisions are narrative text (e.g. Argyres
1999). This is probably because of the complexity of situations in the real world.
However, this approach cannot be used to define support for real world processes
mainly because such processes are very complex and natural language is
insufficiently rigorous. In other studies, examples from practice are used to support
theoretical and methodological arguments. Rich narratives and case studies are
viewed as two popular approaches for enhancing debate and learning about OR
applications (Keys and Midgley 2002).
Pidd (1999) states that ‘a model is an external and explicit representation of part
of reality as seen by the people who wish to use that model to understand, to change,
to manage, and to control that part of reality in some way or other.’ Various models
can be used to examine a decision process and enhance our understanding. These
models are often high-level descriptions of the actual process.
A range of modelling strategies such as visual mappings and grounded theory can
be used to make sense of process data (Langley 1999). Soft models act as sense-
making and interpretive devices that offer insight into how a process is being or
should be executed (Melao and Pidd 2000). Pidd (1999) gives a short review of
models and modelling and discusses several principles.
Modelling decision processes creates challenges such as deciding the level of
detail, scoping the study and determining the boundaries between patterns of
behaviour. Business process modelling (Warboys et al. 1999) is a key technology
for studying, representing and implementing business processes. It shows who are
the participants involved in a process, what activities they are undertaking and howthey interact together in order to achieve given goals. It is used in this study in order
to analyse, explain and enact the behaviour of a decision process.
Process modelling allows organisations to monitor, control and manage business
processes in real-time (Eastwood 2002). It helps people not only to redesign
processes but also invent new ones as well as share and communicate knowledge
about organisational practices (Malone et al. 1999). Process modelling tools increase
quality and predictability by standardising activities (Wreden 1998). Increased
explanatory and computational ease are additional advantages (Mackenzie 2000).
3 The approach
A decision process is a kind of transformation process where the transform activity
is to change an aspect of the real world from a state where there is a need or
282 K. N. Papamichail, I. Robertson
123
opportunity, but no means to deal with this situation (a ‘problem’), to a state where a
solution process has been identified. Subsequently, implementation of the solution
would normally lead to the original problem being ‘solved’.
Modelling of transformations can be done with Checkland’s conceptual models
(Checkland 1981; Wilson 1984), identifying activities that contribute to the
transformation and the dependencies between them. Such dependencies are usually
information flows. The diagrammatic representation of the model can also indicate
external information flows and sources, and indicate those information flows that
serve as control actions.
Generic models are the typical published models—a graphic indicating
significant activities connected by edges that depict dependency or information
flow. They generally exclude any information as to how they may be adapted to
illustrate a real-world decision process.
So far one generic process support model called D2P (Decisioning for Decision
Support in Processes) has been developed (Oquendo et al. 2000; Papamichail and
Robertson 2005). This illustrates a model that has been refined and implemented in
a technology suitable for supporting the decision process. It supports evolution in
that the methods for carrying out tasks can be substituted one for another using an
evolvable framework. Such a model does not attempt to prescribe a certain pattern
of behaviour. It is intended to support activities that can reasonably be considered in
this context.
The architectural model of D2P is shown in Fig. 1. Once a decision opportunity
or threat is identified (in ‘doing’), the D2P model is instantiated. Decision making
behaviour is then decomposed in the following phases:
Evaluating
Appraising
Formulating
Enacting Operational
Process
Information flow
Change
Levels Decisioning Doing
List of alternativesand objectives
Refined list of alternativesand objectives
Ranked list of alternatives
Refined decision model
Implementalternative
Fig. 1 Architectural model of Decisioning for decision support in processes
Supporting distributed decision processes 283
123
• Formulating. The decision problem is formulated and structured. Stakeholders
are invited to contribute their input and identify the main issues of the problem.
Decision actors devise alternative strategies and determine the main factors that
drive decision making.
• Appraising. Those alternatives and objectives that have been identified in the
Formulating phase are reviewed. Chief decision makers appraise the formulation
process and decide how to proceed. They may suggest other alternatives and
objectives before they initiate the Evaluating phase. They may be satisfied with
the results of the Formulating and/or Evaluating phases, decide a course of
action and implement their action plan. Finally, they may be unsatisfied with the
Formulating and/or Evaluating results and re-instantiate the decision process.
• Evaluating. This may include a cost-benefit analysis of those alternatives that
are under consideration, a thorough and systematic assessment using multi-
criteria decision analysis techniques or an evaluation facilitated by the use of
decision aiding technology.
Conceptual and generic models can be used to enhance understanding of
activities, their sequence and dependencies. Neither of the above models can,
however, identify the actual progression through a process and for this a timeline
diagram can place events in perspective, and identify where iterations have taken
place.
Irrespective of the models employed, we have identified a number of properties
that we believe should be captured and analysed when studying decision processes:
• Purpose of taking a decision and striving to achieve some given goals.
• Stakeholders with an inherent interest in the decision taken and possibly affected
by the outcome of the decision.
• Participants involved in the decision process. One or more participants, called
chief decision maker(s), might be responsible for taking the ultimate decision.
• Responsibilities of the decision participants and expected behavioural patterns.
• Activities undertaken to achieve some given goals.
• Temporal progression of the process, i.e. how the decision process unfolds.
• Dependencies among activities or behavioural patterns.
• Iterations of decision steps.
• Factors influencing the process.
4 Modelling a real-life decision process
HEFCE 00/56 was a consultation process initiated in 2000 by the Funding Council
followed in early 2001 by an invitation to bid for funds. The process modelled was
the series of activities set in train in order to construct and submit a credible and
implementable response.
The initial activity was to allocate the task to a senior member of staff. This was
not difficult as the scope of the request fell almost entirely within one functional
area of the organisation. There were initial presentations to members of senior
284 K. N. Papamichail, I. Robertson
123
management and subsequently the activity consisted of identifying areas within the
scope of the request that would benefit from an infusion of funds with results that
would again fall within the scope of the request. A concern was that a commitment
to undertake an improvement in an area could be implemented and the benefits
measured.
Of course there are invariably decisions about decisions and these aspects were
not modelled although they had been considered in model development. Three
models were necessary to obtain a clear understanding of the process as it took place
and how it might be supported. They were a system conceptual model, a timeline
model, and an active model.
The conceptual model is shown in Fig. 2. It is an easily understandable model
that indicates state transitions, dependencies between transitions, as well as
stakeholders and other interactions with behaviours in the system environment. It is
based on Checkland’s models and his Soft Systems Analyses (Checkland 1981;
Wilson 1984). By virtue of the fact that it is easily understandable it is a useful
vehicle for developing other models. The model exhibits the following decision-
making properties:
• Stakeholders
• Participants
• Activities
• Dependencies
• Iterations
• Factors influencing the process
The Timeline model is illustrated in Fig. 3. This is a model of an actual example
of the process, and indicates the sequence of certain activities or events that took
Fig. 2 Conceptual model of HEFCE 00/56 decision process instance
Supporting distributed decision processes 285
123
place, and the extent to which the process had to be iterated over before an outcome
was arrived at. The model highlights the following decision-making elements:
• Activities
• Temporal progression
• Dependencies
• Iterations
The active model (Fig. 1) is a generic model that supports decision processes. It
is active in the sense that it can be implemented in a technology (see Sect. 6 and 7).
The concern here is to have a succinct and reusable model that is less conducive to
easy understanding. It represents:
• Participants
• Responsibilities
• Activities
• Temporal progression
• Dependencies
5 Relating the models one to another
The relationship between the timeline and conceptual models is clear. The
sequences that are identified in the timeline model must be logically possible in the
conceptual model. Where an activity or an event is mentioned in one but not the
other is usually explained in terms of significance in the context. For example, the
activity ‘solicit views’ is implicit in the timeline model but it is explicitly
represented in the conceptual model because it highlights an important behavioural
pattern.
The relationship between the conceptual model and the active model is more
complex, in that we know from the outset that, because the active model is generic it
can only support the instance of the decision process illustrated after having been
adapted. D2P represents decision-making activities that are likely to occur at a high-
level of abstraction. Decision-makers, however, might skip decision steps.
Meetingwith UCEA
ProduceHEFCEreport
Produceproposal
Meetingwith Deans
MeetingwithDeans
Write report
MeetingwithSMT
Meetingwith PC
MeetingwithCouncil
MeetingwithTU
Revise report
UCEA - University Continuing Education Association
TU - Trade Unions
PC - Personnel Committee
SMT - Senior Management Team
Sendreport to HEFCE
t
Fig. 3 Timeline model of HEFCE 00/56 decision process instance
286 K. N. Papamichail, I. Robertson
123
Therefore, the actual decision process as represented by the conceptual model is a
specialisation or adaptation of the D2P. Any time a decision process is instantiated,
the active model, which is essentially an executable D2P instance, needs to adapt in
order to support the process.
Adaptations must take place in two dimensions. The first is in the detailed model
that defines behavioural support for D2P operations, and the second is in the
insertion of D2P models where decisions are needed within the decision process, i.e.
a kind of meta-decisions: decisions about aspects of the decision process itself. This
is the normal situation that we cope quite well with in our day-to-day activities.
In our case study, decision-making behaviour can be decomposed into the
following phases (see Figs. 1, 2):
• Formulating. This is the generation of alternatives. The core of the response was
determined by identifying where the request coincided with initiatives that were
already a part of the university’s strategic plans. Once this was identified,
consideration was given to what activities could be undertaken in order to
address the remaining aspect of the 00/56 request. This consisted of soliciting
views from the community at large, and, more specifically, interacting with pro-
active stakeholders. The stakeholders included members of the university’s
senior management team, the Personnel Committee and particularly the Deans
of Faculties representing the academic membership of the university.
• Appraising. Appraising is initially a refining activity—testing the adequacy and
feasibility of alternatives generated in Formulating and possibly extending them.
The conventional Appraising is comprised of two parts: one dealing with the
transfer of the decision model to Evaluating, and the other dealing with taking
the ranked list of alternatives from Evaluating and either implementing one of
them or returning the decision model to Formulating for cycling again. In this
study, one alternative was iteratively negotiated with the stakeholders so that at
the end there was no need to defend it. Therefore, the focus was placed upon
refining the previous draft of the solution rather than refining a list of
alternatives. In fact there was a cycling between ‘Formulating’ and ‘Appraising’
of the D2P model. This can be supported by the active model by skipping the
‘Evaluate’ activity until it is really needed and utilising the ‘refined decision
model’ feedback link to Formulating.
• Evaluating. This is conventionally the behaviour of ranking the alternative
solutions in terms of participants’ values and weighting. In the 00/56 context, the
activity here was one of testing the adequacy of the draft proposal with regard to
the requirements of 00/56 and the perceived needs of the university. If the test
failed, the decision model had to be refined, and it was returned to Formulating
and Appraising.
6 The technology
D2P is a generic role model of the decision process that can be implemented in a co-
ordination technology, a technology that facilitates co-ordination among users in an
Supporting distributed decision processes 287
123
organisational context. It is an active model, which means that, when executed, the
system prompts users for their contribution to the process at the appropriate point. It
thus effectively passes control of the process between users to progress the process
to a conclusion. The conclusion in this case is the identification (and rationale) of an
alternative (i.e. a potential solution) which is deemed to be ‘best’ for dealing with
the stated problem.
The technology is ProcessWeb (Warboys et al. 1999), which is characterised by
its persistence, and the facilities that it provides for run-time evolution of
implemented models. It is a process support system designed to work across the
World Wide Web. Users can access the system through their Web browsers and
enter data or submit queries to check on the status of the enacted process. Any time
the state of the process changes, the system propagates the changes and the user
displays are updated to reflect the change.
ProcessWeb provides a concise expression of the pattern of activities nominally
required to arrive at a decision by choosing amongst alternatives. For example,
during a decision analysis process, alternatives need to be evaluated. Depending on
the evaluation method used, the alternatives can be scored against criteria such as
quality, implementation time, priority and cost.
A feature of ProcessWeb, which makes it very suitable for experiments on
decision processes, is its ability to evolve at run-time. This means that the pattern of
activities that are supported can be changed whilst decision-makers are interacting
in certain activities. The final steps of a process do not have to be established until
just before they are needed.
7 The implemented model
The implemented computer model was based on the earlier version (Oquendo et al.
2000) developed in the PIE project (PIE 2001) and was extended to make it capable
of supporting the case study process. There were two main areas of difference:
1. The HEFCE case study did not have, as its main task, the choosing among
alternative responses to the HEFCE request. The broad structure of the response
was accepted by all parties, and the task in hand was to best shape this response
so as to maximise benefit to the institution and to maximise likelihood of it
being accepted by HEFCE. In the former aspect, this was to align it as far as
possible with existing strategies, and in the latter aspect to ensure that all the
constraints imposed on their funding were met. This was achieved by a process
of negotiation with the principal stakeholders, the Deans, who were represented
on the Senior Management team.
2. There was no ‘ranking of alternatives’. Once that approach had been negotiated,
the choice for the management team was to proceed, or not to proceed, with the
bid as prepared. Final approval of the bid was the responsibility of Council, a
senior committee of the university.
The additional support was provided by adapting the original model such that it
could support decisions that involved the choosing among alternatives, or decisions
288 K. N. Papamichail, I. Robertson
123
that involved the development of a unique solution. In our case study (see Sect. 4),
the chief decision maker was the human resource manager of a large higher
education organisation who had to put together a bid for funding. His aim was to
devise a detailed human resource management strategy and submit a bid. If
successful, he would obtain additional funds that would allow him to implement the
devised strategy. The main task in this effort was to decide on funding percentages,
i.e. to explain how the funds would be distributed to support initiatives such as
recruiting highly skilled staff, improving retention levels and increasing salaries of
staff on low income. In order to justify the percentages of funding, he invited
stakeholders (e.g. the Deans and the Senior Management Team) to comment on and
contribute to his proposals. He was constrained however, by the amount of time he
could spend on the preparation of the bid as he had to meet the submission deadline.
The decision process in our case study starts with the chief decision maker
becoming aware of the opportunity to obtain further funding and devising a draft bid
that includes funding percentages. He logs into ProcessWeb using login id and
password facilities (Fig. 4) and initiates an instance of the D2P process model. He
assigns himself to the Appraising role and sends the draft bid to those actors that are
responsible for formulating the problem, i.e. those assigned to the Formulating role.
The ‘Formulating’ role actors receive notification that a draft proposal has been
submitted and login to the system. Only those parties that have been assigned to the
Formulating role (i.e. Personnel Committee members, Senior Management Team
members, Trade Union representatives and the Deans) can submit their views
through the system. They essentially have to review the draft proposal and suggest
criteria, e.g. cost of implementing the proposed strategy and quality of proposal. The
user interface (see Fig. 5) gives them the flexibility of proposing new alternatives
for comparison (even though this activity did not take place in our case study). It
should be noted that other users (e.g. the human resource manager) can check on the
status of the process and indentify which parties may be causing delays but cannot
interact with the Formulating user interface.
Fig. 4 The login interface of ProcessWeb
Supporting distributed decision processes 289
123
Once all the interactions with the stakeholders have been completed, the human
resource manager is notified. He connects to ProcessWeb and views the results of
the consultation process on the Appraising user interface (Fig. 6). As the chief
decision maker, he can still make changes and delete or add new criteria such as
‘priority’, i.e. how well the proposal meets the priorities set by the funding body. He
is then able to make one of four choices (see Fig. 1): Firstly, if he is happy with the
appraisal, he can submit the final draft, complete the ‘Decisioning’ process and
return to ‘Doing’. Secondly, he can submit the proposal for evaluation. Thirdly, he
can initiate a further consultation cycle by returning control to Formulating and
inviting stakeholders to resubmit their views. Fourthly, he can have the proposal
appraised by someone else.
In our case study, the human resource manager is focused on incremental
revisions of his bid rather than devising alternative proposals and conducting a
thorough evaluation of the alternative bids. Therefore, the evaluation decision step
was implicit. The implemented D2P model, however, is prescriptive. It offers the
possibility of thoroughly assessing alternatives and it is up to the decision actors
whether they would like to do so. This illustrates the point made in Sect. 5 that an
actual decision process is an adaptation or specialisation of the D2P.
There are two options in the ‘Evaluating’ role: a voting facility (yes/no, with or
without comments, majority/unanimous voting) or a ranking system connected to a
library of multi-criteria decision analysis models (currently populated with a multi-
attribute value theory model). The two options appear in Fig. 7 and 8, respectively.
In Fig. 7, the human resource manager opts for the voting facility and assigns
stakeholders/actors to the Evaluating role. These decision actors are given the
choice of approving the main parts of the proposal. If the bid is approved by the
majority of them, a message is sent to the human resource manager (assigned to
Appraising) who can then proceed with the submission of the bid. If the proposal is
not approved or the chief decision maker is not happy with some aspects of the
decision process, the D2P cycle is repeated.
Fig. 5 The Formulating user interface
290 K. N. Papamichail, I. Robertson
123
The second option in the Evaluating role is to rank alternatives. For example, if
the decision actors suggested alternative strategies, the Evaluating user interface
would look like in Fig. 8. Whoever is assigned to Evaluating, for example the chief
decision maker in consultation with a decision analyst, would then have to score the
alternatives (on a scale from 1 to 100) and decide the weights of the criteria (i.e.
cost, time, quality). A multi-attribute value function would be used to rank the
alternatives.
After the completion of Evaluating, the results of the voting/ranking process are
then sent to the actor(s) assigned to the ‘Appraising’ role and displayed on their
Web pages when connected to ProcessWeb. A future plan is to explore the
capability of D2P to invoke external decision aiding tools and display their outputs
Fig. 6 The Appraising user interface
Fig. 7 The Evaluating user interface (voting system)
Supporting distributed decision processes 291
123
on ProcessWeb. In our case study, the final step is to have the decision approved by
the Council and complete the D2P process.
8 Relevance to researchers and practitioners
The aim of this work is not to aid decision makers in taking better decisions, but
rather to help them become more competent at taking decisions through their
understanding of the underlying process. Process-support technology has been used
for representing decision processes with the aim of improving decision-making
practices. Decisions of particular interest are those that are taken over a period of
time. These decision processes are often distributed, heterogeneous and subject to
evolution because of the ever changing organisational environment.
This paper describes D2P, a framework designed to coordinate distributed
decision processes and implemented using Web-based process support technology.
The setting is the formulation of a human resource management strategy for
retaining and training staff in a not-for profit organisation. We believe the work is
relevant to researchers and practitioners who are interested in modelling decision
processes and developing decision support technologies that coordinate complex
and distributed decision processes which evolve over time. It is of particular interest
to managers that are trying to increase acceptability and trust in organisational
Fig. 8 The Evaluating user interface (ranking system)
292 K. N. Papamichail, I. Robertson
123
decisions by making the underlying decision processes transparent and consistent as
well as encouraging stakeholders to participate in the process.
The HEFCE case study allowed us to demonstrate how to map and execute a
longitudinal decision process in a real-life setting as well as explore the validity and
utility of D2P. Along the lines of other studies (Mintzberg et al. 1976; Nutt 1984;
Hickson et al. 1986), we found that decision makers skip or iterate decision steps
and do not follow a well-structured process. The HEFCE decision process was
found to be a specialisation of the D2P model. Therefore, the D2P model can
represent decision processes at a high level but when enacted, it can adapt to reflect
the actual decision process.
Several models of decision making have been suggested in the literature (see for
example Simon 1960; Pounds 1969; Keeney 1982; Holtzman 1989; Pokras 1989;
Couger 1995; Altier 1999; Shim et al. 2002). What differentiates D2P from such
models is that it is executed using process support technology and adapted at run-
time to reflect changes in the process. It should be noted that this work is intended to
make a contribution to the area of decision modelling and decision support rather
than the area of computer-supported cooperative work. A wide range of Web-based
workflow and process-centred systems (Fuggetta and Ghezzi 1994) have been
developed over the years to support teamwork (Ambriola et al. 1997). D2P is a
specialised process-centred system that coordinates distributed decision processes
that involve multiple decision makers and stakeholders.
There has recently been interest in the development of knowledge repositories
that codify contextual information. KnowledgeScope (Kwan and Balasubramanian
2003) and KOPeR (Ramesh et al. 2005) are two knowledge-based systems that use
process support technology to facilitate knowledge sharing and access to a
repository of solution options and process templates. D2P can capture contextual
knowledge as the decision making process unfolds. It can be used to codify and
retrieve contextual decision making knowledge such as who was responsible for a
decision making task, when and how they contributed to the process and what was
the reasoning behind their actions. In the long run, this could help decision makers
to access and codify good practice in decision making.
WOODSS (Pinto et al. 2003; Medeiros et al. 2005) is the only other Web-based
decision support system we have come across that uses workflow technology to
capture and codify user interactions. These are stored in a repository of evolvable
and executable process elements to help decision makers synthesise alternatives.
The process engine executes and dynamically updates decision processes as well as
coordinates actors. The main strength of the executable D2P over other decision
support systems is that it offers prescriptive advice by encouraging decision makers
to follow certain decision steps and thoroughly explore the decision problem. It
assists distributed decision actors throughout all the phases of decision analysis and
implementation including the formulation, evaluation and appraisal of a decision
model. Its main weakness is its user interface that is not as sophisticated as that of
other workflow based tools.
In order to determine the attitude of potential users towards D2P, we have
demonstrated the model and received qualitative feedback through discussions. The
subjects of our study felt that the D2P enhances learning by facilitating
Supporting distributed decision processes 293
123
understanding of the decision content, context and process. It extends organisational
memory by codifying knowledge in the form of problem solving strategies and
information about the decision problem.
Potential users of D2P range from expert decision makers, facilitators and
decision analysts to novices that need guidance through the decision process. We
believe that D2P appeals to those managers that would like to exercise control over
the decision process and the type of knowledge codified. Some subjects stated that
D2P is best suited to organisations with hierarchical structures in which decisions
are taken at different levels and multiple decision makers are involved in different
stages.
The active D2P model can be used to invite actors to participate in the decision
process at given decision points or elicit the views of stakeholders. This is of
particular importance given that in the public sector, there is currently the need to
involve stakeholders when devising public policy while in the private sector,
shareholders increasingly demand to be involved in taking key decisions. This
feature coupled with the ability to map decision processes as they unfold makes D2P
best suited to organisations with a learning culture that encourage and support
knowledge sharing and distribution.
9 Conclusions
A theoretical framework called D2P has been used to support and coordinate
decision processes. The framework has been translated into a formal model that can
be executed by a process support system. This section discusses the potential and
limitations of the D2P model and process support technology.
The D2P was developed as an attempt to model a generic decision process. How
successful it was can only be determined in using the model in a variety of different
contexts. This article describes a decision process carried out in a not-for-profit
organisation and is one such context.
The models used in the study perform very different and complementary
functions. A timeline model gives an indication of the significant events (i.e.
changes of state of the decision process) but does not indicate behavioural
dependencies. This is provided by the conceptual model which, however, suffers
from offering no temporal indications or of the coordination that actually makes it
happen. This brings us to the third model—the active model. This defines
coordinations and a general route through the process but offers the participants the
possibility of tailoring the process to suit their own needs.
Because it supports self adaptation, the potential of the model is constrained
solely by the interpretations put on this by the model developers, and by the
availability of suitable methods to address the needs of the participants. One aspect
of the model which has not been exploited is the ease with which it could interact
with external tools. This was demonstrated in an earlier project, however, now the
potential exists for suitable methods, or even algorithms, to be furnished by a
method provider, or even run by the methods provider and the results made
available to the decision makers. Such methods could be provided via web services.
294 K. N. Papamichail, I. Robertson
123
The D2P is a research vehicle and is not itself intended for or suitable for use as a
tool of general application. It does, however, provide a template for commercial
exploitation.
Cascading decisions is an aspect of decision processes that has never before been
supportable using contemporary technologies. However, some decision points in the
process often involve taking other decisions. D2P offers the possibility of support at
any level in the process: wherever it is needed.
The primary drawback to this technology is its experimental nature. It is not yet
sufficiently intuitive to use, requires a high level of user application to master it
(which individuals might be reluctant to commit in what might be a time-limited
decision), and the library of methods is still quite primitive.
Societal problems are almost invariably complex and encompass power,
legitimacy and urgency structures (Liebl 2002). There is no mechanism in this
model to address social and political issues and support one form of power structure
over another, or of attempting to elaborate on the ‘why’s’ of an actual decision
process. The authors believe that, in the first instance, behavioural support is itself a
major challenge without yet addressing political support.
Human behaviour is, to a large extent indeterminate. The goals and motivations
that are observed may or may not accord with reality. There is clearly the need for
IT to be embedded in human activity. The aim of technology, however, is to support
that activity that can benefit from IT support, and not to hinder that insightful,
imaginative, and purposeful aspect of human behaviour which contemporary
technology is ill-equipped to deal with.
Acknowledgments We would like to thank all those individuals who willingly set aside their time to
enlighten us about the HEFCE 00/56 decision process and also to participate in interesting discussions
about the decision-making practices within their organisation. We would also like to acknowledge the
support of our colleagues. In addition, special thanks are due to Melvin Loh who contributed to the
implementation of the active model. This work was funded by the Research Support Fund (University of
Manchester). Ian Robertson sadly passed away before this article was published. Nadia Papamichail is
grateful for his insightful discussions and wise comments. He was a great man, a trustworthy friend and
an extraordinary colleague. Trained as civil engineer, he worked for several years in the Middle East
before being appointed as Research Fellow in the School of Computer Science, University of Manchester.
He was an active member and representative of the Association of University Teachers (Manchester
University branch) and an elected local Councillor for the Rochdale Metropolitan Borough Council in the
area of Greater Manchester. Ian’s motto in life was ‘what is popular may not be right and what is right
may not be popular’. He will be sadly missed and fondly remembered by all his colleagues.
References
Altier W (1999) The thinking manager’s toolbox. Oxford University Press, Oxford
Ambriola V, Conradi R, Fuggetta A (1997) Assessing process-centered software engineering
environments. ACM Transactions on Software Engineering and Methodology 6(3):283–328
Argyres NS (1999) The impact of information technology on coordination: evidence from the B-2
‘‘Stealth’’ bomber. Organization Science 10(2):162–180
Barthelemy JP, Bisdorff R, Coppin G (2002) Human centered processes and decision support systems.
Eur J Oper Res 136(2):233–252
Checkland PB (1981) Systems thinking, systems practice. John Wiley, Chichester
Couger JD (1995) Creative problem solving and opportunity finding. Boyd and Fraser, Danvers
Supporting distributed decision processes 295
123
Dean JW, Sharfman MP (1996) Does decision process matter? A study of strategic decisionmaking
effectiveness. Academy of Management Journal 39(2):368–396
Eastwood G (2002) Part of the process. Computer Business Review 10(2):27–30
Elsbach KD, Barr PS (1999) The effects of mood on individuals’ use of structured decision protocols.
Organization Science 10(2):181–198
Follett MP (1941) An executive decision is a moment in a process. In: Metcalf H, Urwick L (eds)
Dynamic Administration: The Collected Papers of Mary Parker Follett. Pitman, London
Fuggetta A, Ghezzi C (1994) State of the art and open issues in process-centered software engineering
environments. J Syst Softw 26(1):53–60
Hickson DJ, Butler RJ, Gray D, Malory GR, Wilson DC (1986) Top decisions: strategic decision making
in organisations. Blackwell, Oxford
Holtzman S (1989) Intelligent decision systems. Addison-Wesley, Reading, MA
Keeney RL (1982) Decision analysis: an overview. Operations Research 30(5):803–838
Keys P, Midgley G (2002) Part special issue editorial: the process of OR. Journal of the Operational
Research Society 53(2):123–125
Kwan MM, Balasubramanian P (2003) KnowledgeScope: managing knowledge in context. Decision
Support Systems 35(4):467–486
Langley A (1999) Strategies for theorizing from process data. Acad Manag Rev 24(4):691–710
Langley A, Mintzberg H, Pitcher P, Posada E, Saintmacary J (1995) Opening up decision-making—the
view from the black stool. Organization Science 6(3):260–279
Liebl F (2002) The anatomy of complex societal problems and its implications for OR. Journal of the
Operational Research Society 53(2):161–184
Mackenzie KD (2000) Processes and their frameworks. Manag Sci 46(1):110–125
Malone TW, Crowston K, Lee J, Pentland B, Dellarocas C, Wyner G, Quimby J, Osborn CS, Bernstein A,
Herman G, Klein M, O’Donnell E (1999) Tools for inventing organizations: Toward a handbook of
organizational processes. Manag Sci 45(3):425–443
Medeiros CB, Perez-Alcazar J, Digiampietri L, Pastorello GZ, Santanche A, Torres RS, Madeira E,
Bacarin E (2005) WOODSS and the Web: annotating and reusing scientific workflows. Sigmod
Record 34(3):18–23
Melao N, Pidd M (2000) A conceptual framework for understanding business processes and business
process modelling. Information Systems Journal 10(2):105–129
Mintzberg H, Raisinghani D, Theoret A (1976) The structure of ‘unstructured’ decision processes.
Administrative Science Quarterly 21(2):246–275
Mohamad Noor MM, Papamichail KN, Warboys B (2006) Implementation of the adaptive decisioning of
decision support in process (D2P). Paper presented at the IADIS International e-Society Conference,
Dublin, Ireland
Nutt PC (1984) Types of organizational decision-processes. Administrative Science Quarterly 29(3):414–
450
Nutt PC (2000) Decision-making success in public, private and third sector organizations: finding sector
dependent best practice. Journal of Management Studies 37(1):77–108
Nutt PC (2002) Making strategic choices. Journal of Management Studies 39(1):67–96
Oquendo F, Papamichail KN, Robertson I (2000) Overcoming inadequacies in process modelling: the
need for decisioning be a first-class citizen. Lecture notes in computer science (Software Process
Technology) 1780:84–89
Papadakis VM, Lioukas S, Chambers D (1998) Strategic decision-making processes: the role of
management and context. Strategic Management Journal 19(2):115–147
Papamichail KN, Robertson I (2003) Supporting societal decision making: a process perspective. Journal
of Multi-Criteria Decision Analysis 12(2–3):203–212
Papamichail KN, Robertson I (2004) A process support system to coordinate societal decision processes.
Lect Notes Comput Sci 3183:566–567
Papamichail KN, Robertson I (2005) Integrating decision making and regulation in the management
control process. Omega-International Journal of Management Science 33(4):319–332
Pidd M (1999) Just modeling through: a rough guide to modeling. Interfaces 29(2):118–132
PIE (2001) Process instance evolution. European Commission Framework IV. Project number 34840, vol
2003
Pinto GDR, Medeiros SPJ, De Souza JM, Strauch JCM, Marques CRF (2003) Spatial data integration in a
collaborative design framework. Commun ACM 46(3):86–90
296 K. N. Papamichail, I. Robertson
123
Pokras S (1989) Systematic problem-solving and decision-making: rational methods for problem-solving
and decision-making. Crisp Publications, Los Altos
Pounds W (1969) The process of problem finding. Industrial Management Review 11:1–19
Ramesh B, Jain R, Nissen M, Xu P (2005) Managing context in business process management systems.
Requirements Eng 10(3):223–237
Rodrigues SB, Hickson DJ (1995) Success in decision-making—different organizations, differing reasons
for success. Journal of Management Studies 32(5):655–678
Shim JP, Warkentin M, Courtney JF, Power DJ, Sharda R, Carlsson C (2002) Past, present, and future of
decision support technology. Decision Support Systems 33(2):111–126
Simon HA (1960) The new science of management decisions. Harper and Row, New York
Volkner P, Werners B (2002) A simulation-based decision support system for business process planning.
Fuzzy Sets and Systems 125(3):275–287
Warboys BC, Kawalek P, Robertson I, Greenwood RM (1999) Business information systems: a process
approach. London
Wilson B (1984) Systems: concepts, methodologies and applications. Wiley, Chichester
Wreden N (1998) Model business processes—companies turn to modelling tools and reengineering
techniques to analyze business goals and improve the efficiency of business processes. Information
Week 98(702):1A–8A
Supporting distributed decision processes 297
123