Problem Identification and Modelling

Problem Identification and Modelling

Copyright © IFAC Identification and System Parameter Estimation 1985. York. UK . 198:; PROBLEM IDENTIFICATION AND MODELLING J. W. Bryant Departmen...

1MB Sizes 2 Downloads 104 Views

Copyright © IFAC Identification and System Parameter Estimation 1985. York. UK . 198:;

PROBLEM IDENTIFICATION AND MODELLING

J.

W. Bryant

Department of Applied Statistics and Opnational Research, Sheffield City Polytechnic, Pond Street, Sheffield SI IWB, UK

Abstract. Model identification and parameter estimation have an impeccable pedigree in the control systems area, yet the prob lem contexts in which the resulting models are employed are seldom as thoroughly researched. It is self-evident that the issue of problem identification - that is, of negotiating a focus for a study must be a crucial first step in any practical project, often determining irreversibly the direction of its later thrust. While a broad systems stance is necessary, it is not in itself sufficient. There is a strong current in contemporary Operational Research which addresses this subject of problem structuring, and this paper reviews these approaches with the aim of illustrating how they can help us to avoid producing the right answer to the wrong problem. Keywords. Decision-aiding; Game theory.

Operations Research;

Management systems;

110delling;

INTRODUCTION Operational research (O.R.) is concerned with the process of decision-aiding. This concern has been manifested at two levels : firstly, through the development of a body of methods intended to aid in decision situations; and secondly, through the study of the actual process by which decisionaiding may be carried out . Of these two strands, the former has historically been dominant, the process of decision-aiding tending to be handled by the practitioner in a largely intuitive manner. The result has been the accumulation of a sizeable repertoire of techniques, applicable in the main to relatively stereotyped decision situations (and which in some quarters have regrettably even come to be regarded as synonymous with O.R.), but little corresponding growth in understanding the decision-aiding process itself. I n other words, the accrued wisdom of practice has been distilled into a potent if limited technology, whereas the collective memory of its craft has passed unrecorded with each succeeding generation of workers. This steady loss and relearning of expertise occurs in any activity, though the apprenticeship ethos, which appears in many forms, attempts to overcome it.

In this paper, the fundamental stance that problems are the product of a process of negotiation is first introduced and elaborated. A number of decision-aiding methodologies which respect this position are then reviewed, and their potential is indicated. The problem of choosing an appropriate methodology is next discussed and some of the dangers of adopting a mechanistic approach to this issue are highlighted. Finally, current developments in the area of problem identification are mentioned and some cautionary remarks made. PROBLEM NEGOTIATION Just as model identification has been described as 'the process of finding or choosing an appropriate model for a given situation' (Gilchrist , 1984), so problem identification can be defined as the process of finding or choosing an appropriate problem in a given situation. It is this process, which the O.R. worker can be instrumental in facilitating, upon which this paper must first focus.

However, in an area like

O.R., where an overriding concern must always be with the way in which support is given to those in decision situations as well as with the formal structures used as the basis for that support, to rely on the haphazard personal experie nce of individual practitioners to develop their skills of process management is, at the very least unsound, and at worst dangerous for those they are attempting to assist. Fortunately, in recent years, there has been a growing movement within O.R. to redress the earlier imbalance, and to develop a better understanding of what makes for effective practice in decision-aiding. This movement in the development of O.R. methodology already provides some guidance on the way in which studies may be carried out, and in particular sheds light on the crucial issue of problem identification. It is this contribution which will be examined here.

Consider the case of a firm manufacturing high technology products which begins to experience a n increasing number of complaints from customers regarding the performance of units. This situation might be interpreted by the firm's salesforce as being a classic example of the way in which they are being let down by the inadequate quality control procedures of the production function; how can they be expected to maintain customer goodwill and therefore to increase sales in these circumstances?

However, to the

production function, the high rate of returns is seen as an inevitable result of having to keep pace with the demands of the marketplace; how can they maintain such a high manufacturing throughput on obsolescent machinery and within internal budgetary limits. The personnel function blames the level of returns on the lamentable attitude of the work force who appear to take no pride in their

225

J.

226

W. Bryant

work, while the design engineers dispute the claim of the market researchers that the product is not suited to the envi ronments in which it is being used. Probably all concerned would concede that the company has a problem , but there would be some difficulty in establishing precisely what this problem was. Eden (1983) has cogently argued that problems can only meaningfully be discussed in relation to their 'owners'. That is, problems are not objective entities but belong to the individuals who define them. It follow that problem identification is necessarily contingent upon the precise stance adopted. The O.R. worker is professionally engaged to help clients to deal with problems. In fulfilling this role, the operational researcher must rely upon the client being able to communicate the problem, in order that effective support may be provided. However, to expect a c lient to be able to do this implies that the process of problem identification can be carried out unaided by the client. Further, Mason and Mitroff (1981) would argue, that such a formulation of the problem is in any case tantamount to a statement of its solution, and so the role of the helper is reduced to that of providing trivial analyses. The implication is that in order to offer any substantial client support, the O.R. worker must also be able to assist the client in articulating the problem in the first place. One of the chief functions of the approaches which are described in the next section is to aid this artic ulation. However, before proceeding to review these methodologies, it is necessary briefly to dwell o n two . further and related issues: the client-consultant relationship and the nature of the client. Underlying the establishment of a c lientconsultant relationship is the implicit belief that such a relationship will be mutually beneficial, even if for covert rather than for manifest reasons. The consultant's key role is to help the client to shape problems and this may be done in a more or less directive manner. The definition of the problem to be tackled is thus the outcome of a process of negotiation between client and consultant, rather than something imposed unilaterally by either party. Frequently problem-helpers have to work with groups or teams of people as clients rather than with individuals. Indeed in some cases the whole proc ess centres upon the creation of a group around an agreed problem nucleus, and so the problem is the product, rather than the subject of the debate: this is the rationale of the problem-solving workshop (Webb, 1982), used in conflict resolution. In these circumstances the identification of a problem is the outcome of a far more complex social proces s within the group and between the group members and the consultant. However the principle remains valid that the eventual problem is negotiated rather than stated or imposed.

METHODOLOGIES It would be foolish to attempt a comprehensive review of all those methodologies which have be en successfully used in decision-aiding, because of the sheer range of approaches concerned. Inevitably too any review will be a partial one in both senses of the word, both because of the imbalances of reporting and also because of the predilictions of the reviewer. The former factor is a significant one since it has not been common

practice in O.R. case histories to record the approach used in managing the process of decisionaiding; nor indeed to dwell over-much on selfcritical analysis of practical studies, either at the time of their execution or retrospectively, an activity that many have regarded as unhealthily introspective and incestuous. The latter source of imbalance is best handled by making clear at the outset that the following review is centred about a number of methodologies with which the author has been more immediately involved, but at the same time pointing out that these approac hes do represent a substantial stream in contemporary O.R. and that they all fulfil the requirement that problem identification is treated as an explicit issue to be examined. The arbitrary starting-point for consideration of the process of problem identification is the beginning of the dialogue between client and consultant. Inevitably, prior to this the situation concerned will have exercised the client, and some attempts will probably have been made to shape and define the problem. It is to these primordial thoughts that the client will refer during the initial exchanges. An approach which facilitates the expression of these personal appreciations of a situation has been developed by Eden (1979) and is now quite widely applied. This method relies heavily upon a technique known as cognitive mapping, a way of portraying a client's idiosyncratic thinking about a situation. This involves the creation of a map, usually depicted in the form of a directed graph, which sets down the client's beliefs about the causal structure of the situation, and the parallel development of a hierarchical model of the client's value system. Each node within such a map is a concept expressed by the client, and is written in bipolar form, the poles containing psychological opposites; these bipolar elements are termed 'constructs' and are conceived as first proposed by Kelly (1972). Arrows or lines link the nodes and indicate the presence and nature of causal or connotative links respectively between constructs. Overall the map provides a rich picture of a situation and of the problems and their interconnections embedded in it as seen by the client. However, it would be wrong to overemphasise the importance of the cognitive map in Eden's approach, although it does play a major role. The basic purpose of Eden's method is to aid expression of thoughts about a problem and to "access 'theories' which have been developed through experience" (Eden, 1983 ). The map is simply a decide whcih facilitates this process in an empathic manner, and which generates an artefact which can be effectively used in discussion and debate both between cl ient and consultant and between members of a client group. The result is the evolution of a problem expressed in the vocabulary of the problem owner to whic h the client can give full commitment. The cognitive mapping approac h is perhaps unique in its sensitivity to the c lient's view of a situation. Other methods, to a greater or lesser extent, appear to involve using a predetermined external frame of reference to structure the situatio n. Arguably the 'gentlest' of the other approaches which will be discussed here is that of Bowen (1981) devised essentially as a problem formulation process. This uses as its central artefact a diagrammatic repr esentation of the relationships between various parties who may be involved in a situation; spec ifically, purposeful interactions, lines of communication and conflicts are shown. A supporting picture is used by Bowen

Problelll Identification and Modelling to depict the temporal sequence of interact ions between parties in order to capture some of the dynamic aspects of the situation. Bowen employs these artefacts to reflect back to the c lient his interpretation of the Situation, continually amending and updating them until through the process itself, the prob lem is adequately defined. Bowen's approach differs from that o f Eden in its conscious focus upon the relationships between significant actors in a situation. The hypergame approach associated with Bennett (1982) also involve s in its preliminary problem structuring (pPS) phase the elicitation, through a dialogue with the c lient, of a cast list o f actors involved in a situation , and an identifi cat i on of the issues through which they interact . The structured picture whic h emerges from thi s process may t hen be systematically developed to investigate the options available to the various protagonists. Through the methodica l expansion of the description, new insights into the situati on can be generated for the c lient and this may lead to changes in the initial ca st list and a reconsideration of the 'prob lem' in an iterative manner . Strongly related to Bennett's work is that of Howard (1971) and Radford (1980) which have also addressed situations through a conflict framework. The previous two approaches have been characterised by Bennett (1984) as being based on theor ies about the structure o f the socia l world, rather than being based o n any explicit theories about the decision-making process. The method of strategic c hoice (Hickling, 1974; Friend, 1983) lies, with cognitive mapping, firmly in the second of these categories. The essence of this approach is the development of a response to uncertainty. The cyclic process involved explicitly includes a mode termed 'problem shaping' in which a client's areas o f choice are identified. Th ese so-called decision areas, which may relate to many areas and level s of choice form the fundamental units of analysis and their interconnections are shown in a picture called a decision graph. Partic ularly in group work, the decision graph forms a vital fo c us for developing a shared understanding of the shape of problems. The process moves o n to an exploration and analysis o f options, but mdY return to problem reformulation at any point. To the extent that strategic choice imposes a structure upon a situation, it resembles the hypergame-related approach. The methodologies which have been briefly introduced above represent some of the most powerful streams in contemporary ' so ft' O.R. However, it would be wrong to conc lude this review without making reference to the work of Cropper (1984) in the field of comparative methodology. Thi s work is important not just for the light that it throws on the characteristics of the various approaches, but also for the novel concep tual framework which it employs to do so . Cropper suggests that the practice of an approach cannot simply be encapsulated in a statement of the methodology employed, but must take account of other equally significant factors specific to the consultant. Thus, an individual decision-aider's 'way of working' represe nt s an interaction between th e nature of the technologies within the consultant's repertoire, and a personal philosophy of, experience of and rules of thumb for decisionaiding. From these derive a preferred way of using technologies, which in turn define the general span of ways of working which are brought to bea r in the decision-aiding situation. The implication is that to discuss a methodology

227

indepe ndent ly o f the decision-aider who is using it is to present an incomplete picture of the process. A corollary is that a conSUltant's way of working is a unique combination of methodological and personal cha racter istics. Cropper's work injects a useful cautionary note into this methodological review. Certainly it is a substantial step f orward to bring some selfawareness into the reporting of the practice of O.R. but it is as well not to be fooled into thinking that a bland methodological sta tement is enough.

PANDORA'S TOOLBOX The review above demonstrates t he range of approaches which have been developed for problem identification. The inevitable question raised is, 'which methodology shou ld be used in a given situation?' . Jackson and Keys (1983) have produced a taxonomy of systems-based, p roblem-solving methodologies in which they cross-classify problem contexts in a 2x2 matrix, and then attempt to associate particular approaches as being most suited to each type of situation. The implication, stated explicitly by the authors, is that a decisionaider should be able to select between and apply whichever of these off-the-shelf approaches is deemed to be most appropriate in a given case. A secondary problem, that of problem-context identification necessarily arises, but this is perhaps less serious than the concept of methodology selection. The brief summary of Cropper's work in the last section will give a c lue to possible objections to the notion of picking from a methodological toolbox. An approach to problem identification is more than the technology of the approach employed; indeed it is more than the simple product of this and the consultant's personal characteristics, being a systemic property stemming from the two factors. Consultants tend to use and develop expertise in particular approaches which match their personal styles of decision-aiding; to pretend that the choice is as simple and as mec hanisti c as selecting an appropriate 'workshop manual' from a methodological 'library ' is to ignore this practical reality. Further, to pretend that there is a sanitised situation 'out there' to be characterised is to ignore the idiosyncratic meanings which the consultant (and client) will attach to their predicament, and which make problem identification far more than pigeonholing some objectively perceived item. Is the idea of a methodology for methodology identification a chimaera? In the sense that a single consu ltant should be able t o switch modes at will between approac hes and to use whatever seems most promising, then the present answer is probably 'yes'. It is no more reasonable to expect a decision-a i der to be master of the range of approaches developed than it is to expect a comprehensive coverage of the 'hard' mathematical and statistical techniques which may also need to be drawn upon in other phases of a project. It is reasonable, however, to expect in a consultant a fully worked-out philosophy of and approach to decision-aiding, and an appreciation of its relationship to more developed methodologies, so that they can be referred to should it seem appropriate to do so. Further, no methodology exists in a ' pure' form; it on ly takes on shape when it is used by an individual. Thus a decision-aider could certa inly maintain personal

J.

228

W. Bryant

versions of the more potent approaches to be used in a mode-switching manner. The on ly qualification must be that such application should not be purely opportunistic, but should be founded on an understanding of the philosophical bases of the approaches tapped and of their implicit frameworks. Before leaving this discussion it is worth adding two substantial footnotes regarding potential developments in methodology selection. The first is an optimistic one, and concerns work which has been carried out during the past two years aimed at linking approaches to decision-aiding. Bennett's paper (1984) of this title gives a structured review of current work which is intended to support the mutual enrichment, and possibly even the partial integration of approaches, at both the theoretical and practical levels. Most striking in this has been the creation of a working group in which, through demonstrations and discussion, the elusive

process-management aspects of different approaches have been exposed and communicated. Such work holds out hope for the creation of more flexibl~, powerful and accessible methodologies for problem identification. The second footnote is of a more cautious nature, but is perhaps more relevant in the present context. It concerns the exaggerated claims of the expert systems and artificial intelligence gurus. In recent years, both simple expert systems, which reflect kn~wledge bases created by human experts, and intelligent knowledge-based systems (IKBS) which incorporate more sophisticated inference systems, have appeared to offer a route towards decision support systems of novel forms. However, the suggestion that such systems will, even in the medium-term, come to be of central value in problem identification, is seriously to underestimate the real complexity of what may appear to be a straightforward process. Certainly the power of computer technology can be, and has been, usefully applied within decision-aiding methodologies, but to suggest that a methodology could be based just upon such an approach is to ignore the vital human factors which makes practical problem identification the art which it is.

CONCLUSIONS This paper has shown that a lively and relevant explicit craft of decision-aiding is developing, which may be used for problem identification. It is seen to complement existing techniques and methods in O.R. by providing insights which may be used to guide the process of decision-aiding. Currently a variety of disparate approaches exists, stemming from divergent philosophical bases, but work to relate these one with another is in progress. These 'soft' methodologies are partially accessible through the written work of the authors most closely associated with them, but it must be recognised that conventional presentations cannot convey much of what is important in an approach. For this reason, there is no real substitute for becoming involved in practical studies. In the hands of every decision-aider any approach will take on a distinctive personality. This is something to be encouraged as long as the user is fully aware of the rationale underlying the approach, and it is to be hoped that more people will choose to make forays into this field in the future.

Finally, it seems necessary to warn again about the snare and delusion of computer-centred systems. Problem identification is not 'simply' a matter of choosing a suitable set of parameters, or a well-fitting mathematical formulation to be used in a situation. Nor is it a matter of interrogating an intelligent knowledge base for information or guidance. It is rather a process to be entered into with a helper who can provide assistance by reflecting back and developing personal appreCiations of a situation, and who thereby invokes problems to identify themselves.

REFERENCES Bennett, P.G. and C.S. Huxham (1982). Hypergames and what they do: a 'soft O.R.' approach. J. Opl. Res. Soc., 11, 41-50. Bennett, P.G . (1984). On linking approaches to decision-aiding. Operational Research, University of Sussex, Brighton, June 1984. Bowen, K.C. (1981). A conflict approach to the modelling of problems of and in organisations. Paper presented at IFORS Conference, Hamburg, July 1981 . Cropper, S.A. (1984). Ways of working: fine tuning our ideas about O.R. methodology. Operational Research, University of Sussex, Brighton, July 1984. Eden, C . , S. Jones and D. Sims (1979). Thinking in organisations. Macmillan, London. Eden, C., S Jones and D. Sims (1983). Messing about in problems. Pergamon Press, Oxford. Friend, J.K. (1983). On the craft of strategic choice. Paper COOR 107: Tavistock Institute, London, August 1983. Gilchrist, W.G. (1984). Statistical Modelling. Wiley, Chichester. Hickling, A. (1974). Managing decisions: the strategic choice approach. Mantec Publications, Rugby. Howard, N. (1971). Paradoxes of rationality: theory of metagames and political behaviour. M.I.T. Press, Cambridge, Mass. Jackson, M.C., and P. Keys (1983). Towards a system of systems methodologies. Dept. of Operational Research, University of Hull, Hull, May 1983. Kelly, G.A. (1972). A theory of personality. Norton, New York. Mason, R.O., and 1.1. Mitroff (1981). Challenging strategic planning assumptions. Wiley, New York. Radford, K. J. (1980). Strategic planning: an analytical approach. Reston Publishing, Reston, Va. Webb, K. (Ed .) (1982). Honest brokers: experience of mediation. Frances Pinter, London.