Dr Frank H. Gregory & LAU Siu Pong

Series No. WP96/05 Editor: Dr. Matthew Lee

July, 1996 Assoc. Editor : Robert Davison

CONTENTS

7. References
8. Figures

It has been argued that the Conceptual Models used in Soft Systems Methodology (SSM) are logically inadequate for some of the purposes, notably information system design, for which they are used. In a series of theoretical papers it has been argued that this shortcoming could be rectified by producing logically enhanced versions of the SSM models. The paper describes the use of these models for the identification of information requirements to support planning in a marketing department of a Hong Kong telecommunications company. The project demonstrates that the models are not, as has been suggested, too complex to be used in client led modelling.

Key Words: Information Requirements Analysis, Soft Systems Methodology, Modal Logic, Logico-linguistic Models.

The information systems literature, with its emphasis on repetitive data processing, pays scant attention to the information needed to support irregular planning procedures. In this context the tools employed by traditional information system design methodologies, such as data flow diagrams, are of little use. Of much greater relevance in the context are the broad based models (SSM Conceptual Models, Cognitive Mapping, Strategic Choice, meta-game and hyper-game models) developed in Soft Operational Research.

The present paper describes the use of logico-linguistic models, logically enhanced versions of SSM Conceptual Models, to determine the sources of information needed to support the planning activities for new products marketed by a Hong Kong telecommunications company. The term "Information Source Analysis" is used to describe the process in order to distinguish it from Wilson's [1] Soft Systems "Information Requirements Analysis (IRA)" and from the narrowly defined "User Requirements Analysis" found in software engineering. Both of these are founded on an input-process-output model while Information Source Analysis is founded on a cause and effect model.

In the present case logico-linguistic models were used to determine what information was needed to produce successful marketing plans. The project was conducted in an SSM action research mode. A member of the planning staff acted as analyst/facilitator and the preliminary models were constructed by him and other people in the planning team in a typical SSM iterative debate. However, logico-linguistic models differ from standard SSM models, which are purely notional, in that they contain empirical elements as well as purely notional ones. The latter stages of the model building required real world research into what actually causes a plan to be successful.

The final model of the information source requirements was compared with the sources of information that were actually available to the planning department. Information that was currently unavailable or inadequately provided was thereby identified and recommendations for improvement followed.

The project demonstrates the application of a broad based and management orientated method of information analysis that can produce results of which the narrowly focused models used in traditional methodologies would be incapable. It also shows that logico-linguistic models can be understood and constructed in a real problem situation by people with no prior training or technical background.

The study was conducted along similar lines to projects undertaken by the Department of Systems at Lancaster University. It was particularly close to the modus operandi of projects conducted on the Department's M.Phil. program where members of a client organization conducted their own SSM projects and organized their own model building. This in-house approach can provide considerable economy in terms of time and money as it dispenses with the need for an outside analyst/facilitator.

Lau Siu Pong, a market analyst with the telecommunications company, was undertaking a part-time Masters degree in Information Systems at the City University of Hong Kong. He was interested in developing an Executive Information System (EIS) to support the activities for the Business Market Business Unit which is charged with planning the marketing of new business products.

The company had previously enjoyed a franchise for the provision of all telephone services in Hong Kong. Recently, in July 1995, that privileged position ceased and for the first time the company was faced with competition. It is, however, a successful and well managed company and economic pundits do not expect that competition will do it any serious harm. Nevertheless, the market for communication technology and services is extremely fast moving in Hong Kong, and good planning for new products will be vital for the company's future success.

It was not anticipated that the project would immediately result in a fully computerized EIS. Rather it was intended to undertake analysis to determine if an EIS would be suitable and if so to broadly determine its configuration. In this context SSM and logico-linguistic modelling were appropriate tools.

The theoretical foundations of logico-linguistic modelling have been laid out in previous papers. However, there is one element of the theory that is particularly pertinent to the present case. Fundamental to Checkland's work is the belief that, while a physical system can exist [2], most systems are in the mind. Conceptual models of human activity systems are not intended to represent actual physical systems or even to represent systems that could physically exist.

Despite Checkland's continued emphasis on this, it is something that seems to be missed by many students and even by many SSM practitioners, the conceptual models in the Multiview methodology are an example [3]. This mistake is easy to understand. In Wilson's method there is a move, apparently a priori, from a conceptual model to an information systems model that is capable of representing physical events. Given this students may reason that if the model representing real world events has been obtained from a conceptual model without a posteriori (empirical) input, then the conceptual model must have represented real world affairs.

This point can be brought out by the distinction between models of concepts and models that are conceptual [4]. A model of a bridge can be physical, as in a scale model for wind tunnel testing. This will be a good model if it behaves like the full scale bridge would. We can also have a conceptual model of a bridge: this might consist entirely of mathematical formulas. This will be a good model if a physical instantiation of the model gives the predicted results. However, a model of the concept "bridge" would be quite different. It would be a model of what a bridge is; it would give criteria for membership in the class "bridge"; in linguistic philosophy it would be an analysis of what the term "bridge" means. Checkland's models are models of concepts but the models built by many SSM practitioners are models (that are conceptual) of real world objects and events.

The theory behind logico-linguistic modelling explains the Checkland model building process as a Wittgensteinian language game [5] in which the stakeholders agree to a set of definitions for the discussion of the problem situation. Agreement about the definition of key terms is essential if agreement about events in the real world is to be reached. Definitions (rules of language), tacit or explicit, are necessary for any description of real world events. Definitions, however, are not sufficient for a meaningful account of real world events; such an account also requires factual rules and factual particulars. Logico-linguistic modelling provides a facility for adding factual (empirically discovered) rules to the Checkland style definitional rules. The definitional rules are marked with the "L" modal operator (standing for logically true) while the factual rules are marked with the "M" modal operator (standing for factually true). A number of logical connectives (see Appendix) are also added to the models to give then the power to adequately represent causation.

The project proceeded in a manner very close to the conventional seven-stage model of SSM [6]. However, this was by coincidence rather than design. There was no attempt to rigidly adhere to the seven stage model.

Stage 1 is entry into the situation considered problematical. This did not need to be undertaken in this project because the analyst was already part of the problem situation.

Stage 2 is the expression of the problem situation. This was fairly perfunctory. It comprised discussions between Lau acting as analyst and Gregory. Because Lau was part of the problem situation there was no need for rich pictures or a debate with colleages and superiors.

With an analyst/facilitator from outside the organization stages 1 & 2 are crucial and time consuming in terms of both man hours and elapsed time. The interviews, discussions and presentation in these stages consume the time of the analyst and, probably more importantly, take up the time of members of the client organization. Having an insider as analyst/facilitator saves this expense. The down-side to this arrangement is that an insider fails to bring the fresh perspective on the problem that might be expected from an outside analyst. In the present project this was compensated by the fact that Gregory constantly challenged the analyst's assumptions and criticized his evaluation of the problem.

Stage 3 is the formulation of the root definition of relevant systems of purposeful activity. Stage 4 is the building of Conceptual models of the systems named in the root definitions. These stages proceeded as they normally do in an SSM project. The situation was slightly unusual in that the analyst was himself one of the "Actors" defined in the CATWOE analysis. The analyst produced a number of root definitions and conceptual models. These were discussed with the supervisor, colleages and managers. The root definition and conceptual model shown in Figure 1 were the ones selected as most appropriate. The Conceptual Model was then decomposed and each of the activities taken to a higher resolution level. This expansion of the model followed traditional SSM modelling techniques.

At this point the project began to diverge from the normal SSM path as the analyst began to use the addition arrows and logical constants to develop the logico-linguistic model. The logico-linguistic model developed gradually through iterations and debate between the supervisor, analyst, the analyst's colleages and managers. The final model is given in Figure 2. There are a number of steps in the development of an SSM conceptual model into a logico-linguistic model. These must now be explained.

SSM models are expressed in the language of commands. The words in the bubbles in SSM models are imperatives; they instruct the reader to do something, viz. "obtain this", "develop that", "study this". Commands do not have truth values (they are neither true nor false). However, tacit truth bearing statements do underlie the models. Take bubbles 1 and 2 and the arrow between them from Figure 1. This is underpinned by a truth bearing statement of the form "In order to obtain the existing profile of target customers, target customers must be identified". If bubbles 1 and 2 were not underpinned by such a truth bearing statement there would be no need to perform the activity specified in bubble 1.

The first step towards a logico-linguistic model is to convert the commands into truth bearing propositions. This is easily accomplished. For example "Identify target customers" in Figure 1 becomes "Customer/market need is identified" in Figure 2. It needs to be pointed out that few of the bubbles from the original model (Figure 1) survived intact to the final model (Figure 2). During debate and more detailed analysis most of the original elements were refined and revised. The second step is to introduce the additional connectives. These are: broken arrow, double headed arrow, AND box, OR box, ANDOR box. The solid single headed arrow remains the same (see Appendix).

The object of these steps is to increase the logical power of the models. This can be useful for a number of reasons. If, like Wilson, we are going to use the models as a foundation for the design of information systems intended to be informative about real world events, then the models will need to have the power to represent causal sequences; models containing these additional connectives can, traditional SSM models cannot. In the present project, traditional SSM models were enough to build a consensus about the main business and goals of the unit. It was when the analysis proceeded to consider how the desired physical events could be brought into effect that the Logico-linguistic models started to be useful.

The bubble diagram format of the model is the easiest to comprehend but there are direct equivalents in predicate logic, PROLOG or simple English [7]. For example, the broken arrow going into element 4 represents a sufficient condition and the box containing elements 41, 42, 43, represent conjunction. In ordinary English this could be expressed as "IF the characteristics of target customers is studied AND the cost of sales estimated and compared among different channels AND agreement is reached between a competent sales manager and marketing manager THEN an effective sales channel will be selected". The solid arrow leading out of element 9 represents a necessary condition. It states that it is not possible for all of the elements 2, 3, 4, 5, 6, 8 to be true unless accurate profitability analysis is carried out.

The introduction of the modal operators did not proceed in the manner suggested in the theoretical exposition of logico-linguistic modelling [7]. This called for the construction of the definitional model and then the addition of factual rules. This is not how it proceeded in this case. It was clear that with the conceptual model Lau had produced a model of how to develop a marketing plan rather than a definition of a marketing plan. In a conventional SSM project this would have been a serious error but, as logico-linguistic models can clearly distinguish between definitions and factual rules, it was possible to formulate the definitions at a later stage. To make this perfectly clear, traditional SSM calls for consideration of "what" is to be achieved and then consideration of "how" to achieve it. In the terminology of logico-linguistic modelling the call is for a definition of a desirable state of affairs and then a model of what events will cause that state of affairs to come into existence.

In this case Lau's model was initially of what causes a marketing plan (Figure 1). Later on, consideration was given to how a marketing plan should be defined. At this point it was realized that producing marketing plans was not very interesting and that the desirable state of affairs was actually "a successful marketing plan". A successful marketing plan could be defined in terms of one or more of: the achievement of the stimulation target (subscription, usage or revenue); meeting the market share; achieving the customer satisfaction index target. With this definition in place, the factual elements in the model were revised in the subsequent iterations. Other definitions were then added to the model. The solid arrow between element 4.1 and element 4 represents a logically true necessary condition. In simple English it says "by definition, the sales channel cannot be effective if the sales target is not achieved".

Stage 5 consists of the comparison of models and the real world. In the present case this was achieved by producing a table. The table listed the elements from the logico-linguistic model, the information required to make the element true (bring it into effect), the information currently available, the distribution channel and the inadequacy or improvement needed.

This stage was rather similar to the information category stage in Wilson's method. With Wilson, activities from a conceptual model are listed and the information inputs to the activities and the information outputs from the activities are identified. There was, however, a very important difference. In Wilson's method the information categories are constructed below the line between the real world and system thinking [1]; this indicates that they are constructed a priori. In the present case the information requirements were obtained empirically by research in the real world. Lau produced the table by looking at the records of past planning projects and questioning the people that had been involved in them. The relationship between the entries is a matter of factual truth rather than logical truth.

Elements that had an "L" connection to element 7 were not included in the table as they played no role in the causal account of how the desirable state of affairs can be achieved. Only the elements at the periphery of Figure 1, those with no arrows going into them, needed to be included because the internal elements, those with arrows going in, would, by implication, be true if the external elements were true.

In most cases the information requirement could be matched with some sort of information provision. But in many cases this provision was inadequate. Five areas of inadequacy were identified. Three of these were concerned with the internal organization of the company but two were of a general nature and can be described here.

One was newspaper clippings which were cited as the source of information needed to support market trend identification, the study of competing products/services and competitor’s offers. However, the clippings were only circulated on a daily basis; they were not catalogued and filed for future reference. The existing situation might help to keep staff up to date but they could not be called upon as an integral part of the planning procedure.

The second, which was changes in the market situation e.g. demographics, economics, consumer behavior, was identified as information required for market trend identification. The cited source was the Hong Kong Monthly Digest of Statistics. While this provides a wealth of information it had not been determined which figures were relevant to the market for the company's products.

Stage 6 is the identification of changes that are systematically desirable and culturally feasible. In the present case this consisted of a set of recommendations. None of the recommendations involved any cultural or political feasibility problems. The only problems that would be involved in the implementation of the recommendation would be the normal ones of time and money.

There were four recommendations for action. These included:

The distinctive feature of models that differentiate between logical and factual universals is that they facilitate testing and amendment. They not only include hypotheses about the real world but also provide criteria by which the hypotheses can be substantiated or falsified. The main hypothesis in the case model was that given a quality product with a competitive price and effective sales channel and market communication plan and back-end support and good planning management then the marketing plan would be successful. The criterion for success is that the product marketed reaches stimulation or market share or customer satisfaction targets.

If the next planning exercise conditions in the hypothesis are met and the criterion for success is met then the model will be substantiated and we can have greater confidence in it. If, on the other hand, the conditions of the hypothesis are met but the criterion for success is not met the hypothesis will have been falsified. In this case the model will need to be revised. Such revisions will consist of additions to the model - conditions that were not thought of when the model was produced. There is no reason why the model could not be maintained indefinitely and survive through radical changes in the organization and its environment.

This process conforms to common sense ideas. We formulate rules to explain events in the world, we find exceptions to the rules and then formulate new rules to accommodate the exceptions. While this framework is common in the physical sciences it is comparatively rare in many areas of management science. In information system design methodologies, for example, the notion of falsification simply does not exist. In the computerized information systems built by these methodologies the implicit factual rules cannot be shown to be false by particular facts; if the system goes wrong there is no mechanism to detect the defective rule. Error detection mechanisms are concerned only with logical consistency not with consistency with real world facts.

Information requirements as it tends to be understood in the information systems journals and in the established information system design methods is very narrowly focused. It assumes that tasks have been clearly defined and that the users know exactly, and can articulate exactly, what information they need to support these tasks. This assumption may be appropriate for people concerned only with transaction processing systems, such as order processing and stock control, but it is not appropriate to executive activities such as planning.

SSM offers a much broader approach towards understanding what needs to be accomplished but when it comes to the fine detail, as in Wilson's information categories, we again find the assumption that the users know what information is needed to accomplish the required tasks. Information source analysis does not make this assumption. It uses soft methods to define the desirable state of affairs but uses empirical methods to find out what information is needed to bring about this desirable state of affairs. It is, in fact, doubly empirical. The model is built empirically and also remains permanently open to falsification and revision by real world events.

Another outcome of the project is its contribution to the dispute between Klein and Gregory [8, 9, 10]. Klein has suggested that logico-linguistic models are too complex to be used for stakeholder driven modelling. His last words on the subject were "Both Gregory's argument and my own are theoretical ones... The practical implications of human information processing needs to be examined in the context of action research programmes...".

This has now been done. The project described here was a classic action research project. The findings are that the stakeholders did not have any serious problem understanding the logico-linguistic models. In spite of the fact that they had no training or previously familiarity with logic, they seem to have found them no more difficult than the traditional SSM models. This demonstrates that logico-linguistic models are not too complex to be used in stakeholder driven modelling.

Of course it would be foolhardy to infer from this that it is generally the case that these models can be built by non-professionals. Logico-linguistic modelling is only in the early stages of practical application. Nevertheless, preliminary results are encouraging.

"AND" Box. Represents conjunction. If "AND" is true, all the elements in the box must be true.

"ANDOR" Box. Represents inclusive disjunction. If "ANDOR" is true one or more of the elements in the box must be true.

"OR" Box. Represents exclusive disjunction. If "OR" is true one and only one of the elements in the box must be true.

Broken Arrow. Represents implication or a sufficient condition. A broken arrow going from p to q means that if p is true then q must be true.

Solid Arrow. Represents implication or a necessary condition, however, the direction of implication is the opposite to that of the broken arrow. A solid arrow going from p to q means that if q is true then p must be true.

Double Headed Arrow. Represents mutual implication. A double headed arrow between two elements means that if one is true then the other must be true.

"L" Modal Operator. Indicates that the relation is true as a matter of logic.

"M" Modal Operator. Indicates that the relation is true as a matter of fact.

1. Wilson, B. (1990) Systems: Concepts, Methodologies and Applications, John Wiley. Chichester.

2. Checkland, P.B. (1981) Systems Thinking, Systems Practice, John Wiley. Chichester.

3. Gregory, F.H. (1993) Logic and Meaning in Conceptual Models: Implications for Information System Design, Systemist, 15, 1, 28-43.

4. Gregory, F.H. (1993) Cause, Effect, Efficiency and Soft Systems Models, Journal of the Operational Research Society, 44, 333-344.

5. Gregory, F.H. (1993) SSM to Information Systems: A Wittgensteinian Approach, Journal of Information Systems, 3, 149-168.

6. Checkland, P.B. and Scholes, J. (1990) Soft Systems Methodology in Action, John Wiley. Chichester.

7. Gregory, F.H. (1995) Soft Systems Models for Knowledge Elicitation and Representation, Journal of the Operational Research Society, 46, 562-578.

8. Klein, J.H. (1994) Cognitive Processes and Operational Research: A Human Information Processing Perspective, Journal of the Operational Research Society, 45, 855-866.

9. Gregory, F.H. (1995) Over Simplistic Cognitive Science, Journal of the Operational Research Society, 46, 274-275.

10. Klein, J.H. (1995) Over-Simplistic Cognitive Science: A Response, Journal of the Operational Research Society, 46, 275-276.