Content uploaded by Michael C Jackson
Author content
All content in this area was uploaded by Michael C Jackson on Feb 18, 2019
Content may be subject to copyright.
Towards a System of Systems Methodologies
Author(s): M. C. Jackson and P. Keys
Source:
The Journal of the Operational Research Society,
Vol. 35, No. 6 (Jun., 1984), pp.
473-486
Published by: Palgrave Macmillan Journals on behalf of the Operational Research Society
Stable URL: https://www.jstor.org/stable/2581795
Accessed: 05-02-2019 13:35 UTC
REFERENCES
Linked references are available on JSTOR for this article:
https://www.jstor.org/stable/2581795?seq=1&cid=pdf-reference#references_tab_contents
You may need to log in to JSTOR to access the linked references.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms
Operational Research Society, Palgrave Macmillan Journals
are collaborating with JSTOR
to digitize, preserve and extend access to
The Journal of the Operational Research Society
This content downloaded from 5.198.35.247 on Tue, 05 Feb 2019 13:35:10 UTC
All use subject to https://about.jstor.org/terms
J. Opl Res. Soc. Vol. 35, No. 6, pp. 473-486, 1984 0160-5682/84 $3.00+ 0.00
Printed in Great Britain. All rights reserved Copyright ? 1984 Operational Research Society Ltd
Towards a System of Systems Methodologies
M. C. JACKSON and P. KEYS
Department of Operational Research, University of Hull
This paper sets out to consider O.R. as a problem-solving methodology in relation to other
systems-based problem-solving methodologies. A 'system of systems methodologies' is developed as
the interrelationship between different methodologies is examined along with their relative efficacy in
solving problems in various real-world problem contexts. In a final section the conclusions and benefits
which stem from the analysis are presented. The analysis points to the need for a co-ordinated research
programme designed to deepen understanding of different problem contexts and the type of
problem-solving methodology appropriate to each.
INTRODUCTION
Ackoff's papers on the present state and future prospects of O.R.1,2 have provided the
stimulus needed for a re-examination of the nature of O.R. as an activity. These papers,
together with the recent contributions of Dando and Bennett3 and Rosenhead and
Thunhurst,4 have begun to establish a body of literature which examines O.R. in a more
rigorous way and in terms of a wider range of issues than has previously been the case.
This paper seeks to add to this literature by investigating the relationship between O.R.
and some other problem-solving methodologies. O.R. is not the only systems-based
problem-solving methodology, and the aim of the following is to establish how O.R.
compares to some other methodologies of this general type. A problem solver facing a
problem context must address himself to the question of which is the appropriate
methodology to use. By considering this issue in detail, it is hoped to provide some valuable
insights into the nature, strengths and weaknesses of the methodologies considered.
The first section develops a systematic analysis of problem contexts. A problem context
is defined to include the individual or group of individuals who are the would be problem
solvers, the system(s) within which the problem lies and the set of relevant decision makers.
This set contains all of the elements which can make decisions which may affect the
behaviour ofthe system(s). In particular, the problem solvers may also be decision makers.
A classification of problem contexts which is relevant to the issue of which methodology
is most appropriate for the problem solvers to use is developed.
In the second section, the methodologies most suitable for the different classes of problem
context are identified.
These two sections therefore provide a 'system of systems methodologies' since the
inter-relationship between different methodologies is probed, as is the relationship of these
methodologies to real-world problem contexts.
In the third section, some ideas which stem from the analysis are presented, and some
conclusions are drawn as to the role of O.R. and its relationship to other systems-based
methodologies. Some practical implications of the analysis are also considered.
A CLASSIFICATION OF PROBLEM CONTEXTS
The choice of suitable criteria to differentiate types of problem context will play a crucial
role in determining the success or otherwise of a study relating those problem contexts to
different problem-solving methodologies. Good criteria will result in similarities and
differences being revealed which are very pertinent to the questions being asked in the
study. Poor criteria will not allow much, if any, progress to be made. In this paper, the
study is concerned with problem-solving methodologies, and the criteria for classifying
problem contexts must therefore identify relevant similarities and differences in problem
473
This content downloaded from 5.198.35.247 on Tue, 05 Feb 2019 13:35:10 UTC
All use subject to https://about.jstor.org/terms
Journal of the Operational Research Society Vol. 35, No. 6
contexts which are important with respect to problem-solving methodologies. It is the
purpose of this section to identify such criteria and to draw up a classification of problem
contexts based on these criteria.
The management and improvement of systems requires that attention be paid to two
sets of processes?the planning process (broadly defined) and the control process (broadly
defined). Problems can occur in relation to either of these processes. A problem in relation
to the planning process might concern whether the relevant system(s) are pursuing the
correct goals from the decision makers' point of view. A problem in relation to the control
process might concern whether the system(s) are pursuing their goal in the most efficient
manner. Ackoff5 calls problems related to the planning process 'developmental' and
problems related to the control process 'evaluative'. According to Ackoff6 the minimal
necessary and sufficient conditions for the existence of a problem of either type are:
(1) an individual who has the problem: the decision maker;
(2) an outcome that is desired by the decision maker (i.e. an objective);
(3) at least two unequally efficient courses of action which have some chance of yielding
the desired objective;
(4) a state of doubt in the decision maker as to which choice is 'best';
(5) an environment or context of the problem.
Complications do, of course, arise which make problems considerably more complex than
is suggested in the above. Ackoff goes on to list the conditions from which these
complications arise. When this list is examined, it is found that all these complications arise
either from changes in the nature of the decision maker(s) or changes in the nature of the
system(s) in which the problem is located. With regard to the decision makers, compli?
cations arise when a group of decision makers, rather than one, makes the decision; when
some decision maker(s) make the decision but others carry it out; when some decision
makers not a party to a particular decision react against it; or when the decision makers
objectives are not consistent or change with time. With regard to the system(s) in which
the problem is located, complications arise if the system becomes more complicated and
more difficult to understand, because the number of possible courses of action available
to the decision maker(s) then becomes very large, even infinite.
It can be seen, therefore, that two aspects of problem contexts (decision makers and
systems) seem to have a particularly important effect on the character of the problems
found within them. This is so whether the problems are related to the planning process
or the control process. A good way to classify problem contexts generally, therefore, would
seem to be in terms of the nature of the decision makers and in terms of the nature of
the system(s) in which the problem is located. This will yield a very broad classification
of problem contexts of particular use in this paper. The concern of the paper is to relate
in the most general terms systems-based problem-solving methodologies to the contexts
in which problems are found. The emphasis is on the key variables in problem contexts
which can, in changing their character, lead to qualitative changes in such contexts,
affecting the problems therein and thereby demanding a significant re-orientation in
problem-solving approach. Techniques developed to tackle problems relating to specific
areas of either the planning process or the control process (e.g. cognitive mapping7 or
hypergame analysis8 to aid strategic planning) will not be considered.
The classification of problem contexts can now be developed. It will clearly be useful
to proceed by classifying systems and decision makers separately. The overall classification
of problem contexts will then be a synthesis of these two classification schemes.
The set of all systems can be classified in a variety of ways. Boulding9 classifies systems
into nine categories using the criterion of complexity; while Beer10 generates a classification
using susceptibility to control as the criterion for differentiation. Jordan11 draws up a
systems taxonomy consisting of eight cells into which systems can be fitted, according to
where they lie along three system dimensions; while Checkland12 identifies five classes of
systems which make up a "systems map ofthe universe". Other classification schemes refer
to more specific types of system. Such schemes usually exist within particular areas of
474
This content downloaded from 5.198.35.247 on Tue, 05 Feb 2019 13:35:10 UTC
All use subject to https://about.jstor.org/terms
M. C. Jackson and P. Keys?A System of Systems Methodologies
study?botany and zoology provide many examples. In the present case, a classification
of systems will be provided which hopefully identifies relevant similarities and differences
in systems with respect to problem-solving methodologies.
Problems which are set in systems which are perceived to be simple are often taken to
be easier to solve than problems which exist in complex systems. In terms of problem
solving, a classification based upon the simple-complex dichotomy will therefore be useful.
There is, of course, a difficulty. The classification of a system as complex or simple will
depend upon the observer of the system and upon the purpose he has for considering the
system. If his purpose is to solve a problem which is perceived to exist in the system, then
the type of problem he is tackling will be a crucial factor in determining the perspective
taken. Often the same system may be seen as being simple or complex, depending upon
the problem. The labour market, for example, can be treated as a simple supply-demand
system and is often so treated in the models of the national economy used by the Treasury,
the Cambridge group and the N.I.E.S.R. In this case the labour market is being treated
in a highly aggregate manner because the problems being considered refer to macro?
economic issues such as inflation, interest rate fluctuations, levels of export trade and the
like. Alternatively, the labour market may be taken to comprise many individuals, each
acting as parts of a highly complex system. This approach is useful if the problem being
addressed concerns the individual?issues such as benefit payments, early retirement or
redundancy (see, for example, Clarke and Keys13).
Granted the observer-dependent nature of the simple-complex criterion, some points
contributing to a distinction between simple and complex systems can still be established.
A simple system will be perceived to consist of a small number of elements, and the
interactions between these elements will be few, or at least regular. A complex system will,
on the other hand, be seen as being composed of a large number of elements, and these
will be highly interrelated.
Despite the difficulty, therefore, the simple-complex criterion for classifying systems
remains of some use in helping the identification of 'easy' problems and 'difficult'
problems. More, however, is needed in order to understand fully why some systems pose
'easy' problems while others pose 'difficult' problems. It is possible to identify other points,
which, although connected to the simple-complex distinction, are not fully embraced by
it. These additional points give us an enhanced version of the simple-complex distinction.
In the analysis that follows, it is necessary to refer to simple-etc. systems and complex-etc.
systems to make this clear.
The starting point of the analysis is that, in general, problems can be regarded as 'easy'
if it is relatively straightforward to understand the system(s) in which they are found.
Vemuri14 identifies four reasons why some systems are more difficult to understand than
others. Each of these points contributes to the make-up of a complex-etc. system and has
implications for the problem solver.
Firstly, in complex-etc. systems, not all of the attributes of the parts of the system will
be directly observable. As a result it is difficult to understand the nature of the system
completely. The causes of any problem may be hidden, and this will impede the ability
of the problem solver to identify useful solutions. It will also be difficult to establish the
effects of any solution to the problem without actually implementing that solution.
Secondly, in complex-etc. systems, even if laws can be established relating the actions
of different parts of the system, they will invariably be only probabilistic in nature. Any
attempt to use a quantitative approach to aid the solution of a problem can, therefore,
only give information about the likely effects rather than the exact effects of a proposed
solution.
Thirdly, complex-etc. systems evolve over time. This evolution stems in large part from
the fact that such systems are in constant interaction with the environment?they are
'open' rather than 'closed'. Social systems exist in increasingly turbulent environments and
this makes it difficult for the problem solver to predict system-environment interactions.
Moreover, in social systems, in order for evolution successfully to occur, the parts of the
system must have a certain amount of freedom of action. The parts of the system are
475
This content downloaded from 5.198.35.247 on Tue, 05 Feb 2019 13:35:10 UTC
All use subject to https://about.jstor.org/terms
Journal of the Operational Research Society Vol. 35, No. 6
purposeful, and it is this characteristic which allows the system as a whole to be adaptive
to the environment. This autonomy of the parts of the system, of course, poses difficulties
for the problem solver. 'Solutions' to problems may evoke unpredictable responses.
Fourthly, complex-etc. systems inevitably involve more 'behavioural' problems.
Decisions made in the system will be affected by political, cultural, ethical and similar
factors. This makes it difficult for the problem solver to fully understand the 'rationale'
behind decisions made by actors in the system. Changing values are an important internal
source of change in such systems.
Complex-etc. systems therefore pose difficult problems because they are often only
partially observable, probabilistic, open, have purposeful parts and are subject to
behavioural influences. Simple-etc. systems, on the other hand, are likely to pose easy
problems because they are fully observable, are governed by well-deflned laws of
behaviour, are relatively closed to the environment, have sub-systems which are passive
and do not pursue their own goals, and such systems are not affected by behavioural
influences. These characteristics of simple-etc. systems make problem solving easier since
such systems are more straightforward to analyze, and it is feasible to establish the likely
effect of solutions to problems without actually putting them into practice. The usefulness
of the simple-complex criterion for system differentiation is therefore increased when it
is extended to include those further factors identified by Vemuri.
Ackoff15 has used the terms "machine-age" and "systems-age" to refer to eras which
demonstrated a concern for two different system types. The machine-age was concerned
with systems which were closed, had passive parts, were fully observable and could be
understood using the reductionism of the traditional scientific method. The systems-age
must concern itself with systems which are open, have purposeful parts, are only partially
observable and cannot be understood using the methods of reductionism. Applying
Ackoff's terminology, reference will be made in what follows to mechanical problem
contexts, which contain simple-etc. systems manifesting relatively easy problems and
systemic problem contexts, which contain complex-etc. systems manifesting difficult
problems.
The system in which the problem exists is not the only factor which determines the
character of the problem context. The nature of the decision makers will also greatly affect
the type of solution needed to problems and the problem-solving methodology needed to
reach that solution. The major factor of interest here concerns the objectives ofthe decision
makers. The criterion to be used in classifying decision makers in particular problem
contexts is whether they are a unitary or a pluralist set in respect of their objectives. A
set of decision makers is unitary if they all agree on a common set of goals for the whole
system and make their decisions in accordance with these goals. A set of decision makers
is pluralist if they cannot agree on a common set of goals and make decisions which are
in accordance with differing objectives. A problem context will therefore be called unitary
if the set of decision makers is unitary; a problem context is pluralist if the set of decision
makers is pluralist.
A problem solver acting in a unitary problem context will be acting in a more 'stable'
environment than if he were acting in a pluralist problem context. The stability is derived
from the lack of conflict between the decision makers and the resulting cohesiveness of
that group of decision makers. A second feature which characterizes the unitary
problem context is the ease with which the behaviour of the system can be understood
relative to the situation in the pluralist problem context. The fact that all of the decision
makers are acting towards the same overall objectives means that the behaviour of the
system will be more unified, and this will ease the understanding of overall systems
behaviour. A third benefit of operating in a unitary environment, as far as the problem
solver is concerned, is that implementation of the solution to the problem will be
acceptable to all parts of the system. Since the objectives are common to all decision
makers, a solution which attempts to achieve these objectives will be acceptable to all
decision makers. In a pluralist problem context, a solution which is acceptable to some
decision makers will not necessarily be acceptable to others. In order to arrive at a
476
This content downloaded from 5.198.35.247 on Tue, 05 Feb 2019 13:35:10 UTC
All use subject to https://about.jstor.org/terms
M. C. Jackson and P. Keys?A System of Systems Methodologies
'solution' in this case, two approaches can be used, both of which involve difficulties. Either
a solution can be arrived at after the various decision makers reach some compromise
about overall objectives. Difficulties arise here in trying to achieve a compromise. Or a
solution can be imposed if a subset of the decision makers has sufficient power. But the
context would not then be pluralist in any meaningful sense. Important ethical problems
arise for problem solvers in these circumstances. This point will be taken up in the final
section (implication e).
Problem contexts, therefore, can be seen as being in one of four categories:
mechanical-unitary, systemic-unitary, mechanical-pluralist and systemic-pluralist. Each
problem context differs in a meaningful way from the others. The test of how useful this
classification is lies in the insights it provides about what methodology to use in solving
problems in each category. In the next section, the type of methodology useful for each
type of problem context will be identified.
TYPES OF PROBLEM-SOLVING METHODOLOGY
(The word 'methodology' is here used in its broadest sense?to refer to any kind of advice
given to analysts about how they should proceed to intervene in the real world.) It is now
possible to ask whether there are methodologies available which are capable of helping
with problem solving in the four different kinds of problem context identified above. It
will be clear that no one problem-solving methodology is likely to be of use in all
circumstances?the types of problem and problem context found are too diverse and offer
different difficulties to the problem solver. At best there will be a number of different
approaches, often developed independently of each other, each being suited to solving
problems in only one of the kinds of problem context outlined. Difficulties are almost
certain to occur when methodologies suited to particular problem contexts are transferred
and adopted for use in problem situations for which they were not designed. All it is
possible to do in this section is to see if, for each of the four problem contexts identified,
there exists a methodology suitable for problem solving therein.
Problems within a mechanical-unitary problem context may be adequately dealt with
using the techniques of classical O.R. Because ofthe unitary nature ofthe context, there
will be general agreement about the goals to be achieved. The problem solver will therefore
be able to establish without too much difficulty the objectives of the system in which the
problem resides. Depending on the relative complexity of the mechanical system involved,
the problem solver will employ deterministic or stochastic O.R. techniques to model the
system. If the system is genuinely closed, there are few elements, and these interact little
(or in a regular way), then deterministic techniques can be used to arrive at a mathematical
representation of the system for use by the problem solver. If the system is more prone
to environmental disturbance (but still 'closed' in von Bertalanffy's sense16), the elements
are greater in number and exhibit more interdependence, then the use of simulation and
the techniques of stochastic O.R. may be more appropriate.
The characteristics ofthe O.R. methodology reflect its appropriateness for use in dealing
with problems in the mechanical-unitary context.1718 The first step is to formulate the
problem in the light of the objectives of the system under study. The system is then
represented in a quantitative model (deterministic or stochastic) which will simulate its
performance under different operational conditions. The particular design that optimizes
the performance of the system in pursuit of its objectives will then be chosen for
implementation.
Two 'sister disciplines' to classical O.R. are systems engineering (S.E.) and systems
analysis (S.A.). These are, like classical O.R., probably most appropriate for solving
problems in mechanical-unitary contexts.
S.E. developed in the 1950s from the work of the Bell Telephone Laboratories on the
design of engineering systems (see the classic exposition of S.E. by Hall19). S.E. has been
defined by Jenkins as "... the science of designing complex systems in their totality to
ensure that the component sub-systems making up the system are designed, fitted together,
O.R.S. 35/6?B 477
This content downloaded from 5.198.35.247 on Tue, 05 Feb 2019 13:35:10 UTC
All use subject to https://about.jstor.org/terms
Journal of the Operational Research Society Vol. 35, No. 6
checked and operated in the most efficient way."20 Jenkins recognizes close similarities
between S.E. and O.R. but suggests that, while S.E. places emphasis on the design ofthe
total system, O.R. tends to be content to tinker at the level of the more mechanical
sub-systems. For Jenkins, therefore, S.E. is broader than O.R.?apparently capable of
solving problems in systemic contexts. In Jenkins, too, there is some attempt to take
account of the pluralistic nature of some problem contexts. Thus, according to Jenkins,
"... systems have conflicting objectives". However, it is taken for granted that the systems
engineer will be able to conjure up some compromise between those objectives. There are
some differences, therefore, between S.E. and O.R., but in the most important aspects
Jenkins' S.E. methodology is very like the O.R. methodology.
S.A. (which spawned such variants as planning-programming-budgeting-systems)
originated, like O.R., in wartime military operations planning (see Checkland21). It is
designed to appraise the costs and other implications of alternative means of reaching a
goal. In the hands ofthe RAND Corporation, it began to be used in non-military settings.
It is said by its proponents to be a broader and more refined methodology than O.R. As
with S.E., there have been attempts from within the S.A. tradition to enlarge the approach
in such a way as to make it appropriate for tackling problems in systemic contexts.
Optner22 makes use of certain cybernetic notions in his systems analysis; while Krone23
presents a systems-analysis-policy-sciences mixture of concepts capable of analyzing
"... Many areas where systems analysis faltered on the limitations of its own overly
rational methodologies in the past. ..."
Checkland, recognizing the similarities between classical O.R., S.E. and S.A., has
labelled the approach they share in common 'hard systems thinking'. This approach is
"... based upon the assumption that the problem task they tackle is to select an efficient
means of achieving a known and defined end."21
It is necessary now to consider whether there is a methodology available which will
enable a problem solver to tackle problems arising in systemic-unitary contexts. The
systems of concern are what Beer10 calls exceedingly complex, probabilistic systems. They
have many elements in close inter-relationship and exhibit behaviour which is difficult to
predict. Furthermore (and this is clear in the above classification but not in Beer's), these
systems are open to their environments and include purposeful parts. There is, however,
full agreement about the goals of the system(s) (unitary).
Following Beer, it is argued that the tools provided by cybernetics give the problem
solver the best chance of dealing with difficulties encountered in this type of problem
context. According to Morgan,24 cybernetics can be treated as a technique for
"... improving the design, control and performance of systems geared to the achievement
of pre-determined ends" (as Morgan makes clear, there is another side to cybernetics. It
can also treat organizations as morphogenic systems, as learning systems and as ecologies).
The tools employed by cyberneticians for dealing with systemic-unitary contexts are the
black-box technique, variety engineering (based on information theory) and negative
feedback. All of these tools receive full articulation in the model of any viable system
constructed by Beer in Brain of the Firm15 and The Heart of Enterprise.2b
Briefly, a system is viable for Beer if it is capable of responding to environmental
changes, even if they were not foreseen at the time the system was designed. In order to
achieve this capacity for self-regulation, a system needs to achieve 'requisite variety' (at
a level concordant with its effective performance) with the complex environment with
which it is faced. It should use the black-box technique to acquire an adequate model of
that environment and should reflect possible environmental states in its own organization.
Also, because of the great complexity of the environment, if the system is to achieve
self-regulation, its goals need to be broken down into sub-goals and these allocated to
different sub-systems, which are themselves viable systems. These sub-systems will have
discretion in relation to the achievement of these sub-goals. It is the use of negative
feedback which allows both the system as a whole and the sub-systems to monitor their
performance in relation to their goals. According to Beer, all viable systems need to possess
five functions, which may be labelled policy, intelligence, operational control, coordination
478
This content downloaded from 5.198.35.247 on Tue, 05 Feb 2019 13:35:10 UTC
All use subject to https://about.jstor.org/terms
M. C. Jackson and P. Keys?A System of Systems Methodologies
and implementation.27 It is essential that all these functions be adequately catered for in
systems and their viable sub-systems. Great importance is also given to the design of the
information channels which link the different functions and the organization and its
environment (the technique of variety engineering is crucial here). The model of any viable
system can be used as a diagnostic tool to check for the existence and proper performance
of the five functions and the communications channels in any actual system. This should
point to the reason for any problems and suggest cures.
The limitations on the applicability of cybemetics as a problem-solving methodology are
brought out in a critique of Beer's work in Chile (Project CYBERSYN?the name afforded
to the Chilean experiment in the cybernetic regulation of the social economy) developed
by Ulrich.28 Using Ulrich's analysis, it is clear that cybemetics could be employed to solve
problems, in systomic-pluralist contexts only if it were capable of generating both 'intrinsic
control' and 'intrinsic motivation'. Project CYBERSYN went some way to generating
intrinsic control?spreading the sources of control throughout the architecture of the
system. But it could not generate intrinsic motivation?distribute the source determining
the system's goal state and purpose throughout the system. It was only the top
controller?the government?which was in a position to change CYBERSYN's design
according to its political purposes. There are certainly authoritarian implications in the use
of cybemetics for keeping "everything under control".29 However, if there is a genuine
agreement about goals, cybemetics can be a very effective method for problem solving in
systemic-unitary contexts. In Chile, of course, Project CYBERSYN ultimately failed
because there was not this agreement about goals. Whether such agreement might have
been achieved had not destructive pluralism been 'pumped-in' from the outside is a matter
for speculation.
It is worth mentioning here that the whole area of work known as socio-technical
systems thinking may be regarded as another attempt to come to terms with problems in
systemic-unitary contexts. Socio-technical systems theorists believe30 that the 'primary
task' of a system (the assumed unitary goal) is best achieved if there is joint-optimization
of the technical sub-system and the social-psychological sub-system. Once the system
becomes complex, the sub-tasks which contribute to the primary task are best
differentiated and put under the control of semi-autonomous work groups.31 Managers
service these work groups by overseeing exchanges across sub-task boundaries. The
semi-autonomous work groups are assumed to be better equipped to deal with variances
that arise from their tasks and, with the help of the managers, with environmental
change?in other words, they are an effective means of coping in systemic contexts.
Mechanical-pluralist problem contexts will now be considered. Because of the nature
of the parts of mechanical systems (passive not purposeful) the pluralism must concern
differences amongst decision makers outside the system (of whom the problem solver may
be one) about the goals to be served by that system. If this disagreement amongst the
decision makers can be resolved, then the problems remaining can be solved using the
'hard' systems methodologies previously examined.
Perhaps the most sophisticated attempt to develop a methodology designed to bring
about a 'synthesis' among decision makers so that action can be taken is that of
Churchman.32 The philosophy underlying the approach is based primarily on the work of
Hegel and Singer, as interpreted by Churchman.33 The notions of thesis, antithesis and
synthesis contained in the Hegelian dialectic are employed in a methodology designed to
develop the world-views or Weltanschauungen of decision makers until a temporary
consensus is achieved. Since this process is never-ending and must be endured by the
decision makers in an 'heroic mood', it can be seen as contributing to a 'Singerian inquiring
system'. Churchman and Schainblatt34 have also investigated the important sub-problem
of how to bring about an effective relationship between problem solver ('researcher') and
the key decision makers ('managers').
Mitroff, Mason et al.35~31 have helped to develop Churchman's ideas into a rigorous
methodology?sometimes referred to as strategic assumption surfacing and testing
(SAST). The advocates of this methodology claim that it is appropriate to use it to tackle
479
This content downloaded from 5.198.35.247 on Tue, 05 Feb 2019 13:35:10 UTC
All use subject to https://about.jstor.org/terms
Journal of the Operational Research Society Vol. 35, No. 6
complex, 'wicked' problems in systemic -pluralist contexts?and it has indeed been applied
to problem areas such as 'strategic planning'.38 However, the whole emphasis of the
approach is on tackling the pluralist aspects of such contexts and not on the systemic
aspects. There seems to be an unwarranted assumption that, once the pluralism has been
dissolved, the problems stemming from the systemic nature of the context will disappear
as well. Thus, Mason and Mitroff37 support Rittle's conclusion that "Every formulation
ofa wicked problem corresponds to a statement of solution and vice versa. Understanding
the problem is synonymous with solving it." The analysis in this paper points to the need
to see the systemic characteristics of some problem contexts as presenting peculiar
difficulties of their own. Since SAST tends to ignore these difficulties, pretending that all
the difficulties stem from pluralism, it is legitimate to see SAST primarily as an aid to
solving problems in mechanical-pluralist contexts and to discuss it here.
SAST is a dialectical approach designed to help decision makers understand the different
points of view which may exist concerning a problem. Four major stages may be identified.
The first stage is 'group formation'. At least two different groups are formed to be
advocates of very different perceptions ofthe problem and system. These groups may be
made up of managers from different functional areas or different organizational levels. The
second stage is 'assumption surfacing'. This consists of various techniques designed to help
each group to uncover and analyze the key assumptions on which its view of the system
and its proposals are based. Each group comes to understand its own key assumptions?its
Weltanschauung (W). The third stage is 'dialectical debate'. Each group develops the best
possible case for its position. The key assumptions underlying its position are presented
and argued for. Each group may be asked to interpret some agreed 'data' according to
its perspective. Once this stage is complete, and each group begins to understand the
assumptions of the others, the fourth, 'synthesis', stage can begin. The aim is that the
different groups should arrive at a synthesis incorporating all their different world-views
and going beyond them as well. This synthesis can then become the basis for detailed
planning and problem solving.
Finally in this section, the solution of problems in systemic-pluralist contexts must be
addressed. It is here that the 'soft' systems approaches of Ackoff and Checkland seem most
appropriate.
According to Ackoff, problem solvers are increasingly faced?in the systems-age15?not
with separable problems, but with 'messes', systems of interdependent problems. The
problem-solving orientation of O.R. should therefore be replaced by one that focusses on
planning for and design of systems.1,2 Moreover, organizations are purposeful systems
which contain purposeful parts and which are themselves parts of larger purposeful
systems. Hence organizations have responsibilities to their own purposes, to the purposes
of their parts and to the purposes of the larger systems of which they are parts.39 These
responsibilities may often seem to conflict (pluralism). The principle task of the manager
and management scientist is to learn how to remove any conflict between these levels of
purpose. AckofFs 'interactive planning' is therefore an elaborate attempt to come to terms
with problem solving in systemic-pluralist contexts.
Interactive planning has three operating principles.40 The first is the participative
principle. If possible, all the stakeholders should participate in the various phases of the
planning process. Stakeholders here include representatives of the purposeful parts of the
system and of the larger system. The second is the principle of continuity. Because values
change and because events occur which could not have been foreseen, plans need to be
continually revised. The final principle is the holistic principle. A plan should involve
simultaneously and interdependently as many parts and levels of the system as possible.
These operating principles are reflected in the five phases of the planning process
itself?formulating the mess, ends planning, means planning, resource planning and
'implementation and control'. The unique feature of the approach is undoubtedly phase
two, in which the interactive planner and the stakeholders participate in designing an
'idealized future' for the system with which they are concerned. This designing of a
desirable future is seen as a way of eliminating petty differences between stake holders and
480
This content downloaded from 5.198.35.247 on Tue, 05 Feb 2019 13:35:10 UTC
All use subject to https://about.jstor.org/terms
M. C. Jackson and P. Keys?A System of Systems Methodologies
of concentrating everyone's mind on the broader, long-term interests they all share in
common. The idealized future is a design of the future that begins from scratch. It is a
statement of the system the planners and stakeholders would create if they were free to
create any system they wanted. If pursued positively, interactive planning can lead to the
interests of the different parts of the organization, of the organization itself, and of the
wider society being reconciled. The problems which contributed to the original mess will
be 'dissolved' in the new design. They will, of course, be replaced by new problems. Hence
the need for continuous planning.
Checkland's methodology41 is similarly adept at dealing with systemic-pluralist problem
contexts. It grew out of the frustration experienced by consultants trying to use hard
systems methodologies in such contexts. A 'problem situation' is first analyzed and a 'rich
picture' ofthe situation is built up. The aim is not to delimit particular problems 'out there'
in the real world, but to gain an understanding of " .. . a situation in which various actors
may perceive various aspects to be problematical".
The various 'Ws' operating in the system (plus any the analyst may wish to introduce)
are expressed in 'root definitions' of systems relevant to the problem situation. 'Conceptual
models' are then constructed ofthe various systems enshrined in the root definitions. These
are used for comparison with what is perceived to exist in the real world. This comparison
helps to structure a debate about possible change among the actors concerned with the
problem situation. The analyst and the various actors should now be able to agree on
changes which are both 'desirable and feasible'. The original problem situation is
successfully tackled, although a different problem situation simultaneously emerges.
In a sense, the systemic-pluralist problem context embraces the other three types of
problem context as special cases. It follows that, in theory, the Ackoff and Checkland
methodologies should be able to address problems in all four problem contexts. To use
these methodologies in contexts other than the systemic-pluralist would, however, be very
inefficient. If they were used in either the mechanical-unitary or systemic-unitary contexts,
resources would be wasted in reaffirming an already existing consensus on objectives. If
they were used in either the mechanical-unitary or mechanical-pluralist contexts, efforts
would be wasted attempting to deal with a complexity that did not exist. Although,
therefore, these methodologies are potentially able to address problems in all problem
contexts, in practice they are only appropriate for one?the systemic-pluralist context.
IMPLICATIONS OF THE ANALYSIS
The purpose of this section is to consider what benefits of an intellectual or practical nature
can be derived from the analysis pursued up to this point. Because the analysis seems so
fruitful in terms of opening up new insights and new avenues of enquiry, it is impossible
within the scope of this paper to do more than point to the existence of some of its more
important implications. These implications can roughly be considered as being of two
kinds: first, theoretical implications which have a bearing on the current debate about the
nature of O.R. as an activity; second, practical implications concerning the conduct ofthe
problem-solving venture.
Theoretical implications
(a) The analysis presented places a different perspective on the 'O.R. in crisis' debate.
Instead of seeing different problem-solving methodologies as competing for the same
problem contexts (as Dando and Bennett3 do), different problem-solving methodologies
are presented as being appropriate for dealing each with one type of problem context. A
'crisis' will not therefore occur if a methodology is only used for its appropriate problem
context. A diversity of problem-solving methodologies, each with a defined area of
competence, should be encouraged. Classical O.R. will be one of these with its own unique
competence. The way ahead is to work for the establishment of an 'avant-garde O.R.'
embracing all the problem-solving approaches discussed above.
481
This content downloaded from 5.198.35.247 on Tue, 05 Feb 2019 13:35:10 UTC
All use subject to https://about.jstor.org/terms
Journal of the Operational Research Society Vol. 35, No. 6
(b) Different problem-solving methodologies, instead of being compared one with
another in relation to their ability to solve the generality of problem types, should be
evaluated in relation to their success in solving those problems for which they are best
suited. This should encourage mutual respect among the proponents of the different
approaches. This opens up a much better prospect for future debate and progress than
the rather carping criticism of particular problem-solving approaches from other perspec?
tives, which is currently common.
(c) Once different problem-solving methodologies have been identified, work can begin
on establishing the usefulness of each approach and on developing each approach to its
natural limits. Methodologies should be evaluated in terms of their success in tackling the
problems for which they are suited and in terms of the 'size' of the problem domain in
which they can be used. Two types of work will be necessary: first, theoretical work which
attempts to establish the underlying assumptions about problem contexts which underpin
the different approaches?this should help in the identification of those problem contexts
for which the different methodologies are appropriate (a start has already been made on
this work?see Checkland21 and Jackson42); second, practical work?trying out the
different methodologies in cases where you would expect them to be successful and cases
where you would expect them to fail. Both types of work should see improvements made
to the different methodologies which will facilitate their success in their own domains and
maybe broaden those domains.
(d) Raitt43 suggests that O.R., and therefore all of the complementary methodologies,
are technological activities. Raitt's argument is based upon a distinction between models
and theories, an overriding concern with the former indicating a technological activity, a
dominant concern with the latter indicating a scientific activity. To make such an assertion
as this is to simplify the distinction between science and technology. The interactions
between science, applied science and technology are complex, and to argue that one can
exist in some sort of isolation from the others is to ignore these interactions.44 As the
analysis makes clear, any problem-solving methodology must take into account the
behaviour ofthe system in which the problem exists. This involves the creation of a model
of that system. This model may take one of many forms but it is essentially a representation
of the system. In order to create such a model, it is necessary to formulate some ideas
concerning the relationships and processes which are embodied in the system. This activity
is a theory-building activity, so the model is theory-dependent; it is a representation of the
theory which has been built up of the real world.
The question arises, therefore, of where systems methodologies can turn for the
theoretical support that seems to be required. One answer might be to turn to the science
of systems, general systems theory (G.S.T.). G.S.T. may provide a framework for
discussing features common to many problem contexts. As Naughton45 and Checkland46
have convincingly argued, however, G.S.T. is unlikely to be able to provide the theoretical
support necessary. As an alternative to G.S.T. it might be better to turn to the particular
sciences which are concerned with explaining the nature of the system(s) that exist in the
different problem contexts. This, of course, would involve methodologies concerned with
systemic-pluralist contexts turning to the social sciences. There are problems. Vickers47 has
lamented the lack of support the professions which manage human systems derive from
the social sciences. But Vickers may not be being altogether fair here. There is a mass of
relevant theory available in the social sciences, although this is certainly in need of being
sorted out.
(e) In the analysis, problem contexts of a certain type have been deliberately left out of
account. These are problem contexts characterized by contradictions between different
political and economic interests. There will be structural conflict between the different
groups involved. The way such contexts are structured constrains human development.
The systems involved are ripe for radical change. There is a clear difference between this
type of problem context and the systemic-pluralist context. In a systemic-pluralist context,
it is possible for some compromise solution or some appeal to a desirable future to bring
about a genuine consensus among the parties involved. In what might be called
482
This content downloaded from 5.198.35.247 on Tue, 05 Feb 2019 13:35:10 UTC
All use subject to https://about.jstor.org/terms
M. C. Jackson and P. Keys?A System of Systems Methodologies
systemic-coercive contexts, any cohesion that does exist will be achieved by the exercise
of power and by domination (overt or more or less concealed) of one or more groups over
others. Walsh et al.4S offer a good account of power in organizations and the effect it can
have on any 'consensus' achieved.
Systemic-coercive contexts have been left out of account, not because it is believed they
are rare (it is arguable that the majority of organizations in capitalist societies provide such
contexts?see Rosenhead and Thunhurst4), but because it is believed that the drastic
problems which exist in such contexts are unlikely to succumb to the remedies of
problem-solving methodologies. And it will be remembered that the classification of
problem contexts in the earlier section was constructed only with such problem-solving
methodologies in mind.
The analysis can, however, help explain why problem-solving methodologies are
inappropriate in systemic-coercive contexts. Examples might be offered. Imagine a
problem solver with a problem context characterized by structural conflict, contradiction
and domination trying to apply a hard-systems methodology. Even if the methodology
could be made to work, all the analyst could achieve would be to help the dominant group
or groups to strengthen an enforced cohesion. In so doing, he would be contributing to
the survival of a situation in which human development was being constrained. Even if
he tries to employ a Checkland-type approach, he is likely to do more to buttress the status
quo than to challenge it.42,49 If the methodology could be made to work, agreement will
eventually be reached on feasible and desirable changes. However, as these changes have
been agreed by the dominant group or groups in the system, they are likely to be changes
which support rather than threaten their authority. Other groups may agree to such
changes because of fear of the consequences of opposing them or because they fail to
recognize their own true interests (they are subject to the 'third dimension of power'?see
Lukes50). Again, the problem solver will only tend to contribute to the survival of a
situation in which human development is being constrained.
The use of O.R./systems methodologies in systemic-coercive contexts can therefore only
prolong the existence of systems which benefit some groups at the expense of others. If
Rosenhead and Thunhurst's analysis of capitalist society as replete with systemic-coercive
contexts is correct, then the use of the methodologies in these contexts only services and
legitimates capitalism. To prepare the ground for the abolition of such contexts and the
legitimate use of O.R./systems methodologies, O.R. workers and systems researchers must,
as Rosenhead and Thunhurst suggest, "participate in the struggle of labour against
capital".4
Practical implications
(f) If the analysis is taken seriously, it will lead the problem solver to ask, on each
occasion he is confronted with a problem, what methodology is appropriate to this
problem context. It should make it more difficult for him to get away with employing some
favourite methodology in all circumstances. There should, for example, be less cause for
clients to complain that O.R. workers always see the world in terms of their favourite
problem-solving technique.
(g) The question of identifying problem contexts correctly obviously becomes crucial.
Some work needs to be done to help problem solvers with this difficult task. Problem
contexts in the real world rarely announce their character unambiguously. The way any
problem context is perceived is going to depend very much on the individual who is
observing it (and in the case of a consultant, on the information he can gain from the
client?see Eilon51). For example, the wider he draws the system boundaries, the more
likely is the problem context to become systemic-pluralist. A problem solver's 'W will very
largely determine the way he sees and approaches problem contexts. Problem solvers, of
course, may not be aware ofthe 'W which conditions the way they view problem contexts.
Often, the only expression of this 'W will be an adherence by the individual to some
favourite methodology which acts as the bearer of an unconsciously held 'W\ It seems
necessary, therefore, for the problem solver always to be aware that there are different 'W's
483
This content downloaded from 5.198.35.247 on Tue, 05 Feb 2019 13:35:10 UTC
All use subject to https://about.jstor.org/terms
Journal of the Operational Research Society Vol. 35, No. 6
around?different ways of viewing and approaching systems and problem contexts. The
problem solver needs to stand back and examine problem contexts in the light of different
'W's. Perhaps he can then decide which 'W' seems to capture the essence ofthe particular
problem context he is faced with. This whole process needs formalizing if it is to be carried
out successfully.
The problem solver needs to be aware of different paradigms in the social sciences, and
he must be prepared to view the problem context through each of these paradigms. Burrell
and Morgan52 provide a framework of sociological paradigms which might be a useful
starting point for the work that needs to be done in helping problem solvers identify
problem contexts correctly.
(h) Some problem contexts will, of course, not fit exactly into any one of the above four
categories. Faced with such an intransigent problem context, the problem solver may still
gain benefits from the analysis. It will be possible, using the analysis, to see how a
particular methodology might be extended by making use of aspects of other approaches.
For example, a problem solver who is armed with a soft-systems methodology appropriate
for a systemic-pluralist context may find it possible to 'harden up' his methodology for
a problem context which has some mechanical-pluralist aspects. The resolution of conflict
over objectives may be helped by the use of a quantitative approach to aid the decision
makers in investigating the effects of their own preferred solutions relative to the solutions
of others. A methodology suited to mechanical-unitary problem contexts may be 'softened
up' in a way which makes it more effective in dealing with problem contexts which exhibit
some systemic-unitary features. For example, the use of quantitative methods based on
psychological or sociological theory may be used to deal with certain behavioural aspects
of a systems behaviour. Some examples of such a 'behavioural O.R.' approach are given
in Keys.53
(i) Where the problem solver does happen to begin with a methodology which he later
begins to perceive as inappropriate to the problem situation encountered, the analysis can
indicate why it is inappropriate and can point to the direction in which he must move to
get a correspondence between problem context and methodology. For example, if he is
using the O.R. methodology and encounters conflicting objectives, he might try using the
SAST approach to remove this conflict.
(j) Finally, the analysis aids the understanding of exactly what goes wrong when an
inappropriate problem-solving approach is employed in a particular problem context. For
example, the attempt by Governor Brown of California to apply systems analysis to social
problems in the early 1960s (documented by Hoos54,55) can be understood as an attempt
to apply a methodology appropriate for problems in a mechanical-unitary context to solve
problems in a systemic-pluralist context. More recently, Rosenhead has criticized the
attempt to apply classical O.R. to the systemic-pluralist contexts of the health service56
and urban planning.57 Churchman recounts58 how he attempted to use his methodology,
which can be seen as most appropriate to mechanical-pluralist contexts, to examine the
NASA programme to put a man on the moon. This 'problem' was conceived by NASA
to be set within a mechanical-unitary context. Churchman's analysis received from one
NASA group monitoring the work an 'F' in relation to relevance to NASA's mission and
an 'A' for interdisciplinarity! Beer's work in Chile can be seen (as previously mentioned)
as an attempt to apply a methodology suited to a systemic-unitary context in a situation
which was becoming (perhaps artificially) systemic-pluralist. The debate and consensus
building which is the hallmark of the soft-systems approach would be a waste of time in
mechanical-unitary contexts.
CONCLUSION
O.R. is regarded by many as being in crisis. If O.R. is taken to be 'classical O.R.', this
is indisputable. 'Classical O.R.' provides the practitioner with an approach suitable for
solving problems only in mechanical-unitary contexts. If, however, the definition of O.R.
is widened to embrace other systems-based methodologies for problem solving, then a
484
This content downloaded from 5.198.35.247 on Tue, 05 Feb 2019 13:35:10 UTC
All use subject to https://about.jstor.org/terms
M. C. Jackson and P. Keys?A System of Systems Methodologies
diversity of approaches may herald, not crisis, but increased competence and effectiveness
in a variety of different problem contexts. This paper has suggested that different kinds
of problem context exist in the real world. It is essential to develop different methodologies
to cope with these. In the diversity of methodologies around can be found attempts to
come to terms with each ofthe different types of problem context discussed. The analysis
provided can be used as a starting point for a coordinated research programme designed
to deepen our understanding of these different problem contexts and the type of problem
solving methodology appropriate to each.
REFERENCES
'R. L. Ackoff (1979) The future of operational research is past. J. Opl Res. Soc. 30, 93-104.
2R. L. Ackoff (1979) Resurrecting the future of operational research. J. Opl Res. Soc. 30, 189-200.
3M. R. Dando and P. G. Bennett (1981) A Kuhnian crisis in management science? J. Opl Res. Soc. 32, 91-104.
4J. Rosenhead and C. Thunhurst (1982) A materialist analysis of operational research. J. Opl Res. Soc. 33,
111-122.
5R. L. Ackoff (1962) Scientific Method, p. 24. Wiley, New York.
6R. L. Ackoff (1962) Scientific Method, p. 30. Wiley, New York.
7C. Eden, Thinking in Organisations. Macmillan, London.
8P. G. Bennett (1980) Hypergames: developing a model of conflict. Futures 12, 489-507.
9K. E. Boulding (1956) General Systems Theory?-the skeleton of science. Mgmt Sci. 2, 197-208.
10 S. Beer (1967) Cybemetics and Management. English Universities Press, London.
11N. Jordan (1981) Some thinking about 'system'. In Systems Thinking (F. E. Emery, Ed.), Vol. 2. Penguin,
Harmondsworth.
12 P. B. Checkland (1971) A systems map of the universe. J. Systems Engng 1, No. 2.
13 M. C. Clarke and P. Keys (1982) What happens when public sector employees are made redundant: some
exploratory results using a micro-simulation model. Working Paper 298, School of Geography, University of
Leeds.
14V. Vemuri (1978) Modeling of Complex Systems. Academic Press, New York.
15 R. L. Ackoff (1974) The systems revolution. Long Range Planning 7, 2-20.
16 L. von Bertalanffy (1950) The theory of open systems in physics and biology. In Systems Thinking (F. E.
Emery, Ed., 1969). Penguin, Harmondsworth.
17 C. W. Churchman et al. (1957) Introduction to O.R. Wiley, New York.
18 R. L. Ackoff and M. W. Sasieni (1968) Fundamentals of O.R. Wiley, New York.
19 A. D. Hall (1962) A Methodology for Systems Engineering. Van Nostrand, New York.
20 G. M. Jenkins (1969) The systems approach. In Systems Behaviour (J. Beishon and G. Peters, Eds), 2nd edn,
p. 82. Harper & Row, London.
21P. B. Checkland (1978) The origins and nature of'hard' systems thinking. J. appl. Systems Analysis 5, 99-110.
22 S. L. Optner (1965) Systems Analysis for Business and Industrial Problem Solving. Prentice-Hall, Englewood
Cliffs, New Jersey.
23 R. M. Krone (1980) Systems Analysis and Policy Sciences. Wiley, Chichester (preface).
24 G. Morgan (1982) Cybemetics and organisation theory: epistemology or technique. Human Relations 35,
521-537.
25 S. Beer (1972) Brain ofthe Firm. Wiley, Chichester.
26 S. Beer (1979) The Heart of Enterprise. Wiley, Chichester.
27 R. Espejo (1977) Cybemetics in management and organisation. University of Aston Management Centre,
Working Paper Series, No. 76.
28 W. Ulrich (1981) A critique of pure cybernetic reason: the Chilean experience with cybemetics. J. appl. Systems
Analysis 8, 33-59.
29 J. G. U. Adams (1973) Chile: everything under control. Science for People 21.
30 E. L. Trist et al. (1963) Organisational Choice: Capabilities of Groups at the Coalface under Changing
Technologies. Tavistock, London.
31 A. K. Rice (1958) Productivity and Social Organisation: The Ahmedabad Experiment. Tavistock, London.
32 C. W. Churchman (1979) The Systems Approach. Dell, New York.
33 C. W. Churchman (1971) The Design of Inquiring Systems. Basic Books, New York.
34 C. W. Churchman and A. H. Schainblatt (1965) The researcher and the manager: a dialectic of
implementation. Mgmt Sci. 11, B69-B87.
351. I. Mitroff and J. R. Emshoff (1979) On strategic assumption-making: a dialectical approach to policy and
planning. Acad. Mgmt Rev. 4, 1-12.
361. I. Mitroff et al. (1970) Assumptional analysis. A methodology for strategic problem solving. Mgmt Sci.
25, 583-593.
37 R. O. Mason and I. I. Mitroff (1981) Challenging Strategic Planning Assumptions. Wiley, New York.
38 R. O. Mason (1969) A dialectical approach to strategic planning. Mgmt Sci. 15, 403-414.
39 R. L. Ackoff (1974) The social responsibility of O.R. Opl Res. Q. 25, 361-371.
40 R. L. Ackoff (1981) Creating the Corporate Future. Wiley, New York.
41P. B. Checkland (1981) Systems Thinking, Systems Practice. Wiley, Chichester.
42 M. C. Jackson (1982) The nature of'soft' systems thinking: the work of Churchman, Ackoff and Checkland.
J. appl. Systems Analysis 9, 17-29.
43 R. A. Raitt (1979) Viewpoints. J. Opl Res. Soc. 30, 835-836.
44D. Ihde (1979) Technics and Praxis. D. Reidel, Dordrecht.
485
This content downloaded from 5.198.35.247 on Tue, 05 Feb 2019 13:35:10 UTC
All use subject to https://about.jstor.org/terms
Journal of the Operational Research Society Vol. 35, No. 6
45 J. Naughton (1979) Anti-G.S.T. An evolutionary manifesto. Paper for the Silver Anniversary Meeting of the
S.G.S.R.
46 P. B. Checkland (1972) Towards a systems-based methodology for real-world problem-solving. J. Systems
Engng 3, No. 2.
47 Sir Geoffrey Vickers (1978) Practice and research in managing human systems?four problems of
relationship. Policy Sci. 9.
48 K. Walsh et al. (1981) Power and advantage in organisations. Organisation Studies 2, No. 2.
49 A. R. Thomas and M. Lockett (1977) Marxism and systems research: values in practical action. Proceedings
of the Silver Anniversary Meeting of the S.G.S.R.
50 S. Lukes (1974) Power: A Radical View. Macmillan, London.
51 S. Eilon (1982) Silver Medal Address. J. Opl Res. Soc. 33, 1090-1098.
52 G. Burrell and G. Morgan (1979) Sociological Paradigms and Organisational Analysis. Heinemann, London.
53P. Keys (1982) Introducing behaviour into O.R. models. Mimeo, Department of Operational Research,
University of Hull.
541. Hoos (1972) Systems Analysis in Public Policy. University of California Press.
551. Hoos (1979) Engineers as analysts of social systems: a critical enquiry. J. Systems Engng 4, 81-88.
56 J. Rosenhead (1978) Operational research in health services planning. Eur. J. Opl Res. 2, 75-85.
57 J. Rosenhead (1981) O.R. in urban planning. Omega 9, No. 4.
58 C. W. Churchman (1979) The Systems Approach and Its Enemies. Basic Books, New York.
486
This content downloaded from 5.198.35.247 on Tue, 05 Feb 2019 13:35:10 UTC
All use subject to https://about.jstor.org/terms
