ORSNZ Newsletter, September 1997
The time has sadly come to pass when we must bid farewell to Hans Daellenbach as Editor of the ORSNZ Newsletter. The Council is in the process of arranging a new editorial team, the members of which will be announced at the APORS 97 conference in Melbourne at the end of this year. I think that we would all agree that the members of this team have a hard act to follow. When I offered to write an editorial, Hans modestly asked that there be no eulogies lamenting his departure, so I shall oblige with his request. However I would like to dwell here on some parts of the philosophy of OR which Hans has espoused in his editorials over the years, and in particular talk about implementation.
I think that it would be fair to say that Hans and his disciples advocate that OR/MS should make a difference in terms of the practical improvement of some situation faced by an individual or organization. This is what I would call implementation. Sometimes implementation refers to the fact that a client has paid for some OR/MS software. But if the software is not used, and its answers used to influence the behaviour of the agents, then OR/MS cannot be said to have made a difference, or to have been implemented in the proper sense of the word.
Since I am currently on sabbatical leave I have decided to clean up my files on one of the departmental disk drives. There are optimization problems on this drive of every imaginable flavour, but only a handful of them have been used by an actual decision maker apart from me and my students. (Of course we have gained insights from the models, which may have an indirect influence on other decision makers through consulting advice or conference presentations, but the direct coupling to solving a problem is missing.) I bet that similar graveyards of unimplemented optimization models exist on disk drives all over the world. In some cases these models might have been inappropriate or made invalid assumptions, but in many others they were realistic representations of at least some aspect of the decision problem faced. These models failed to make a difference because there was not sufficient support in the organization for the methodology, nor sufficient time and energy on the part of the author to try and create this support. The approach was "here's the optimal solution, take it or leave it", and they left it.
There are at least two ways in which one can improve the chances that OR/MS will make a difference. The first is to ensure that the models that the OR/MS practitioner develops are used to solve the real decision problem, and not a similar but ineffectual model which has some elegant mathematical properties. Of course the techniques used must be guaranteed to get the correct solution to a given model. But if the model is solely an exercise in mathematics, with only a passing relevance to a real problem, then even the ?correct' solutions will be of little use. This is an issue which Hans has discussed at some length in previous issues of this publication.
The second condition necessary for the successful implementation of OR/MS is that the user of the model buys in to the solution. By this I mean that the user feels some affinity or ownership of the model. This might be because they contribute to its development, or have learned about what makes it tick. Ensuring this transfer of expertise is an equally challenging task to developing the right model and methodology to tackle the decision problem.
So what can be done to help the user buy in? Though ultimately it depends on the circumstances and personalities involved, there are some obvious strategies which, I am sure, most OR/MS practitioners will recognize. The first rule is to start with a simple model before a complicated one. If the end user can interpret, or even guess in advance, the solution that the model delivers, then it gains credibility, and makes the user feel confident with the approach. If the model that the OR/MS practitioner seeks to implement for a complex problem can be used with slight modifications on a small subproblem, then applying it to this is a good way to get the support of the owners of that subproblem. In mathematical programming the modelling languages like GAMS and AMPL are ideal for this sort of prototyping. Some would advocate simple spreadsheet models, which have many advantages for proving a modelling concept to management. (A discussion of the merits and drawbacks of these for mathematical programming should be left for another editorial.)
The second rule is to proceed gently and patiently. The people who need to accept and use the new model probably have methodologies of their own that they are unwilling to abandon or modify. If one can encourage them to contribute to the development in the model by incorporating some of the features in which they have some expertise, then they will treat the model as one of their own. (A useful way of presenting computer models is to sell the model as a solving engine, where the user interface is the domain of the end-user, who can design it to look and feel how they like. As an aside, it is my observation that when buying a computer model, users are typically much more impressed by a slick user interface delivering an approximate answer than a complicated, cumbersome tool delivering the exact answer. This is not to say that we should provide them with the former, but it is a useful first step on the way to a comprehensive computer model.)
The third rule is "do not blind with science". Most end-users of OR/MS models are smart people, but they are usually not mathematicians, so they are not going to be impressed by the mathematical sophistication of the techniques that the OR/MS practitioner uses unless they can either understand the mathematics or understand the solutions that the techniques generate. Either the methodology one uses must be simple, or there is an education and empowerment function that must be performed by the OR/MS practitioner to help the user accept the techniques.
One of the difficulties with following the rules above is that it takes a lot of time and energy. If it is done on a commercial basis then this makes it expensive. The corollary is that such an approach is unlikely to be adopted by a (profitable) commercial operations research consultancy. It might be argued that (at least in the short term) the economically rational position for such a company is to sell just a solution, perhaps in a software package or a weighty report to the board of directors. This strategy ensures that the expertise remains with the consultant, and not the client. A dependent client who is likely to come back to the consultant for more solutions is a useful source of revenue. However, it is not a strategy that is likely to result in a deep acceptance of OR/MS by the client. This comes with education and ownership.
The issue of implementation is thus one involving education and empowerment of the client. Many academics who claim to work in OR/MS do not spend large amounts of their time performing this role. It is a shame that empowerment of the OR/MS user is not a common academic pursuit, because as practised educators, academics are ideally suited to educate users of OR/MS, leading to the successful implementation of OR/MS models and methods.
The following is the second installment of James Corner discussion regarding the teaching of decision analysis at the University of Waikato. As promised, he keeps this installment short! Started in the last newsletter, these snippets eventually will culminate in a paper for Interfaces. He thanks all those readers who responded to last time's discussion.
Consistent with the general teaching philosophy of my department, elements of experiential, or autonomous, learning are always used when I teach decision analysis (see Scott and Buchanan [1992] for a deeper discussion) . This is in part due to a recognition of differing learning styles across students, but also because I see decision analysis as a topic which must be applied in order to be mastered properly. No doubt this is a common feeling among educators.
Three elements of autonomous learning typically are used: learning contracts, buzz groups, and outside applications. Learning contracts are contracts negotiated between students and the lecturer to establish what the student is to learn, how it will be learned, what is to be produced as a result of the learning, and how and by whom the work is to be assessed [Knowles 1986]. These documents are a wonderful aid in building motivation to learn and they help shift the responsibility for learning back to the student. They are used during those portions of the course which are not lecture-oriented, as students typically will be working in groups and on topics different from one another. They especially are useful for getting students motivated to begin their applied projects as discussed below.
The buzz group is a tool for use during classroom sessions. The idea is to periodically (say, every 15-20 minutes) break up the class into groups of 3-5 students to discuss and reflect on what has been said during lecture or other discussion. The lecturer can walk around to help seed the buzz group discussions or usually just gather feedback from the groups. This feedback is invaluable in that it allows the more timid students a chance to interact with the lecturer and others in the class. It also is a wonderful way for the lecturer to focus in on what the class is actually learning, so it aids in mid-class clarification and direction-setting.
Buzz groups are examples of what are more commonly referred to as cooperative learning teams (CLT's). Levasseur [1996] provides a useful review of how to launch such teams. Just like CLT's, buzz groups can be extended for use outside of class. That is, if students are asked to learn in the group setting and outside of class, classroom sessions can be spent on building on this knowledge, rather than on the initial material to be learned.
Finally, as mentioned earlier, I feel that much of the actual learning of decision analysis occurs when students are asked to apply it. I therefore have mandatory semester-long projects where students help real organizations tackle some hard decision problem currently faced by them. At the undergraduate level, example projects include: how to get more hydraulic head on a local hydro-electric dam, whether to change a nearby town's sewage effluent from an ocean dumping to a land-based scheme, and options for choice of a mixing vat in a local laboratory. In our shorter MBA classes, we have students break into pairs, one to act as analyst and one as decision maker, to deal with a real-world decision faced by the student decision maker, such as which job to take, what car to buy, etc. The experiences of the students make for a wonderful late course discussion which can easily tie back into the descriptive/prescriptive discussions which began the course (as discussed in the last newsletter). That is, students typically will be able to discuss their prescriptive experiences with the projects, but also will have a word or two to say about the more descriptive issues they saw during their projects.
Knowles, M. 1986, Using Learning Contracts, Jossey Bass, London.
Levasseur, R. E. 1996, "People Skills: Launching a Cooperative Learning Team," Interfaces, 26/6: 112-116.
Scott, J. and Buchanan, J. 1992, "Teaching MS: Hold the Lectures," OR/MS Today, October: 46-50.
This is another one of thoseOR Newsletters that threatened to be rather thin. In desperation, I dug out a paper I presented at the TIMS Anchorage Meeting in 1994. Andy Philpott's choice of topic for his editorial is a rather interesting coincidence. So you get a double dose of it!
The aim of undertaking an operations research (OR) analysis of a decision situation is to improve the quality of decision making. This is achieved by providing new insights into the problem situation that could not be derived by other means. Viewing OR in this light, its aim thus goes beyond simple optimization. Furthermore, OR will only achieve this aim, if its findings are used as important inputs into the decision making process. Such use I refer to as `implementation'. This is not equivalent with adopting the recommended solution. It is conceivable that the insights gained will, in fact, lead to a decision which may not or only partially follows the recommendations. The following example demonstrates this dramatically. A study done by D. C. McNickle [1994, pp 423-43] for a large wood processing company showed that the installation of a second service facility would save close to one million dollars net annually in waiting costs alone for an initial investment of $400,000. However, sensitivity analysis revealed that a reduction of as little as 5% of the rate of arrivals would make waiting times almost disappear. The insight gained from that led to the discovery of ways to divert about 5% of the arrivals elsewhere at practically no additional cost. This was the solution implemented.
It is tacitly accepted that implementation is facilitated if the model or suite of models and the modeling process satisfy certain desirable properties. (For convenience, I will assume the term ?model' to also include ?multiple models'.). Every student of OR should thus be fully aware of what these properties are, and even experienced operations researcher will do well to remind themselves occasionally. Therefore, it is rather surprising to discover the real paucity in the OR literature on what makes a good OR model. The best-known paper is the one by J.D.C. Little [1970].
Little lists five desirable properties for a mathematical model to meet the needs of the user of such models. I briefly list them with some comments:
(1) Simple: Simple models are more easily understood by the problem owner or decision maker, who is often mathematically untrained. The decision maker will more easily follow the logic of a spreadsheet than of a complicated set of equations, which may do little more than the computations performed in a spreadsheet � admittedly more elegantly. To get simple models, the analyst may have to make suitable approximations to the real situation or even delete certain significant aspects, which may later have to be taken into account in different ways.
(2) Complete. The model should include all significant aspects of the problem situation that affect the measure of effectiveness. The problem here is to know whether an aspect is likely to affect the optimal solution in a significant way before the model is built. Using a systems approach, i.e., exploring all systemic relationships within the context of the total problem situation — not simply the narrow definition of the problem formulation — will go a long way toward establishing which aspects are likely to be significant and which ones may have only negligible effects. Obviously, extensive experience in modeling will help. However, in many situations only by building several models, one without these aspects, the others with various combinations of them included, and then comparing their answers can a confident judgment be made as to the significance of particular aspects. Few operations researchers ever do this.
(3) Easy to manipulate. It should be possible to obtain answers from the model, such as the best solution, with a reasonable amount of computational effort. I am reminded here of the situation faced by the U.S. meteorological services in the 70s and 80s, where they could produce accurate 7-day weather forecast only by having very fast mainframe computers churn away for 5 days!
(4) Adaptive. A model should not be parameter dependent, i.e., invalidated by sufficiently wide changes in the input parameters. However, even reasonable changes in the structure of the problem situation should be able to be accommodated by alternative options in the model. If changes invalidate the model, it should be possible to adapt it to the new situation with relatively minor model modifications only. This is more likely, if the model consists of a sequence of small modules that each perform a reasonably separable task or set of computations. Any structural changes in the problem situation may then only require modifications to or additions of one or a few modules of the model. An adaptive model is often referred to as a robust model.
(5) Easy to communicate with. It should be easy for the analyst and/or the user to prepare, update, and change the inputs and get answers quickly. In today's world of interactive user-friendly computer programs and software, such as spreadsheets (see, e.g., , ) and the new generation of mathematical programming interfaces, e.g., GAMS and AMPL, this property has become one of the standard selling points.
Little also states that the model user should become the `model owner'. His paper then demonstrates how a model meeting these properties was built for dealing with a marketing mix decision problem.
Note that some of these properties put conflicting demands on the modeling process. In particular, a simple model may be unable to capture all significant aspects of the problem situation. In such instances, the analyst may have no choice but to build a complicated mathematical model. In such cases, the decision maker will gain confidence in the model if he or she has the opportunity to experiment with it, e.g., by exploring if changes in the input parameters produce intuitively reasonable changes in the best solution and, if not, whether counter-intuitive results can be explained convincingly. A robust model may not be simple. A model that includes all significant aspects may not be easy to manipulate. The model builder will have to balance these conflicting demands and come up with a suitable compromise. This compromise will by necessity reflect not only the training of the analyst, but also the amount of resources in terms of time and funds available for the analysis. It should also take into account the likely benefits that can be achieved. It may be economically more advantageous to use simple quick-and-dirty rules that only capture 50 to 80% of the potential benefits, rather than develop a sophisticated and expensive models that may capture 95%. The cost of developing a mathematical model, collecting the required input data, computing the best solution, implementing the model, and finally operating and maintaining it all increase much more than proportionately as the sophistication of the model increases, while the additional benefits go up less than propor tionately — anothr case of increasing marginal cost and decreasing marginal returns. All mathematical models are thus to varying degrees approximations to the real situation as perceived by the analyst.
Few other papers explicitly address this question. The ORSA guidelines for the practice of OR [Caywood et al., 1971] cover some aspects, particularly those dealing with the professional and ethical accountability of the modeler, and S. Gass [1981] introduces the notion of model credibility. However, it is the literature on simulation, and model validation and verification that extensively contributes to the topic of desirable model properties in an indirect way, by studying what facilitates validation of models.
Fishman and Kiviat [1968] define validation as the process of assessing the agreement between the behaviour of the model and the real world system being modeled. The difficulty with this definition is the meaning of `real world system'. Since the `real world' cannot really be observed, all the modeler can do is to compare the model with his or her perception of the `real world'. Furthermore, that perception may be different from the one of the decision maker or problem owner. Such differences in perception need to be resolved during the problem formulation stage. The basic premise must be that the problem owner is the expert on the working of the system, and it would be rather presumptious to assume that the modeller's perception is the correct one. So the onus is on the modeller to show beyond a reasonable doubt that her or his interpretation is the more useful one for modelling purposes. With this difficulty in mind, let us now see what we can infer from the literature on validation.
The April 1993 issue on Model Validation of the European Journal of Operational Research contains a number of highly relevant papers that contributes significantly to the topic of desirable properties of models [Landry and Oral, 1993; D�ry et al., 1993; Oral and Kettani, 1993; Gass, 1993]. These papers echo the theme appearing in the simulation literature already in the mid-eighties [Carson, 1986], namely that whether or not a model will be used rests largely on the credibility and confidence the problem owner and user has in its ability to produce useful information. Unfortunately, neither these papers, nor the simulation literature, offer much practical guidance on how to foster credibility and confidence. For example, Law and Kelton [1991] imply that `establishing credibility' is equivalent to `selling results to management' (Fig. 5.1, p. 299, opus cited). But, as Gass [1993] points out, credibility and confidence are not attributes of the model, but of its user. Little's paper lists desirable properties of the form of the model. They may help, but are only incidental to the perception of the problem owner's confidence in and credibility attributed to the model. Satisfying Little's criteria for a good model may thus be far from sufficient for a model to be implemented and used.
Confidence and credibility add a new dimension to the notion of desirable properties of models. In fact, it may not be useful to talk about desirable properties of models, but of desirable properties of the modeling process, since user credibility and confidence are more related to the form of that process and the interactions with the modeler, than to the model itself. Naturally, Little already hints at this when he states that the decision maker should `own' the model. This is confirmed by the research on which factors enhance the likelihood of implementation. The literature on implementation discusses such aspects [e.g., see A. Reisman and C. A. de Kluyver, 1975]. Fostering the feeling of `ownership' of the model and modeling process may not be a easy. It requires appropriate people skills of the modeler and a substantial commitment from the problem owner. The former may not have the necessary skills training, while the latter may not be willing or able to devote both the required effort and time. Furthermore, in many instances the model's sophistication precludes active involvement of the problem owner in the modeling process. Hence `ownership' has to be brought about by other means.
It is also clear that with this we enter the murky area of the psychology of modeling and its process — another indication that the practice of OR cannot be disassociated from the people and the modeling environment, including the social, cultural, and political facets in which the process of modeling occurs. Denying or ignoring these aspects is the major cause for implementation failure. Unfortunately, with minor exceptions, the literature on OR concentrates almost exclusively on the mathematics of OR and the form of the models. Even the reports of successful implementations, such as the accounts of the Franz Edelman Award winners, published in January issue of Interfaces, or the Application Papers, regularly featured in each issue of Operations Research, are mostly silent on this aspect.
It is beyond the scope of this short paper to go into the extensive literature on implementation and address the modeling process. Instead I will limit myself to draw some conclusions from the points raised in the validation literature in the form of two additional desirable properties for models that will facilitate adoption and use of models.
First, I will take for granted that the model actually deals with the problem situation as perceived by the decision maker. Looking beyond the process of modeling, what are some of the crucial aspects that help building up credibility and confidence of the problem owner that the model is useful? [Daellenbach, 1994]
(6) The model is appropriate for the situation studied. By this is meant that the model has the appropriate level of resolution for producing the relevant outputs at the lowest possible cost and in the time frame required for effective decision making. For example, a simple financial spreadsheet may well be the appropriate choice of model if our objective is to provide a sufficiently accurate estimate of the company's profits for the next quarter quickly and with minimal effort, whereas a simulation study which models the movement of every single widget along the production line will not. Even if it also produces suitable financial variables, its level of detail and resolution will be excessive and take too much time. Hence it will not be appropriate for the situation studied. On the other hand, if our objective is to estimate the maximum possible rate of production, the location of bottlenecks, or the size of buffer needed between consecutive production stages, the simulation model will be able to mimic the appropriate level of resolution, whereas the financial spreadsheet will not.
To derive the outputs relevant for decision making, a `good' OR/MS model may not necessarily show details of or resemble the physical system we are studying. What is important is that the model enables the analyst or problem owner to measure how well the stated objectives have been achieved by the proposed solution, and that this information is provided cheaply and in a timely fashion. Such output can often be more efficiently be achieved by a black box approach, rather than a detailed representation of the system's transformation process of inputs to outputs.
(7) The model produces information that is relevant and appropriate for decision making. This means that the output of the model has to bear directly on the decision process, has to be useful for decision making, and has to be in a form that it can be used directly as input for decision making, without the need for further extensive translation or manipulation. This does not imply that the problem owner may not have to use judgment in interpreting the information provided. But the information should lead to new insights into the problem situation that the problem owner could not easily obtain by other means. Such insight is often obtained through sensitivity analysis in terms of the effect of deviations from the proposed solution and of changes in various inputs on the optimal solution, as well as the effect of errors in various inputs, particularly those that are subject to uncertainty or are costly to ascertain at a sufficient degree of accuracy.
For example, referring back to the waiting line study mentioned earlier, the model output provided not only an estimate of the potential net savings due to the addition of a second service facility, but also information about waiting times as a function of the traffic intensity. It was the latter information that gave the problem owner the insight necessary to find a much cheaper solution to the original problem.
If the model satisfies these two properties and the analyst can demonstrate this to the problem owner and the intended user of the model, then this will considerably increase the likelihood that the latter will judge the model as usable and useful. This will enhance the problem owner's confidence in the model and her or his willingness to implement its findings.
As mentioned earlier, confidence in and credibility of the model do not necessarily require an thorough understanding of how the model works. It may be largely intuitive, based on a demonstration that the model gives usable, sensible, expected, and explainable answers in a timely fashion and with a reasonable expense. Furthermore, it will be strongly influenced by the working relationship between the modeler and the problem owner/user and the latter's involvement in the modeling process itself. This is particularly true and important if the working relationship between the modeler and problem owner is new. Once a sufficient degree of trust has been established between the two, then it will be much easier in subsequent projects to gain the problem owner and user's confidence in the model and its findings.
Carson, J.S. (1986). "Convincing users of model's validity is challenging aspect of modeler's job," Ind. Eng., June, 18: 75-85.
Caywood, T.E., Berger, H.M., Engel, J.H., Magee, J.F., Miser, H.J., and Thrall, R.M. (1971). "Guidelines for the Practice of Operations Research," OR, 19/6: 1123-1258.
Daellenbach, H.G. (1994). Systems and Decision Making, Chichester: Wiley.
D�ry, R., Landry, M., and Banville, C. (1993). "Revisiting the issue of model validation in OR: An epistemological view," EJOR, 66/2: 168-83.
Fishman, G.S., and Kiviat, P.J. (1968). "The Statistics of Discrete Event Simulation," Simulation, April, 10: 185-95.
Gass, S.I., and Parikh, S.C. (1981). "Concepts of Model Confidence," Computers and OR, 8/4: 341-346.
Gass, S.I. (1993). "Model accreditation: A rationale and process for determining a numerical rating," EJOR, 66/2: 250-8.
Landry, M., and Oral, M. (1993). "In search of a valid view of model validation for OR," EJOR, 66/2: 161-7.
Law, A.M., and Kelton, W.D. (1991). Simulation Modeling and Analysis, 2nd ed., N.Y.: McGraw-Hill.
Little, J.D.C. (1970). "Models and Managers: Concepts of Decision Calculus," Mgt Sci, 16/8: B-466-485.
McNickle, D.C. [1994]. Chapter 16 in H.G. Daellenbach, opus cited.
Oral, M., and Kettani, O. (1993). "The facets of the modeling and validation process in OR," EJOR, 66/2: 216-34.
Reisman, A., and de Kluyver, C.A. (1975). "Strategies for implementing systems studies', in R. L. Schultz and D. P. Slevin (eds), Implementation of OR/MS, N.Y.: Elsevier.
One Sunday a few months ago, I made myself a large pot of tea in preparation for grading my third-year Operations Research students' papers. As part of their assessment, I require each student to write a three to four page review of a journal article (a practice I brought over from my teaching in the States). Halfway through the first article, I was dismayed at its weak content and improper construction of ideas. I was further surprised when I realized that the paper was written by one of my better students. Disappointed, I abandoned the first paper and randomly selected another but I was presented with the same inadequate level of quality. In total, I read eleven papers that day with only two that I would marginally label as well written technical papers. The other nine were filled with circular arguments, illogical conclusions, and poor sentence construction that conveyed the authors' poor analytical skills, as well as lack of breadth in their educational background.
The Monday after, I visited a few colleagues in other departments to complain (a gran 'ol academic tradition) and I was mildly surprised to hear similar complaints from those colleagues (even those in the Humanities and Social Sciences). I am convinced that we need to consider introducing "Breadth Requirements" for our undergraduate students. In our current program we are producing students who have have a narrowly focused education based on fulfilling a single major (or, for some, two related majors) without any structured general education. I firmly believe that all students would benefit from an undergraduate programme that requires them to take courses from other departments or fields. A general education will give students the opportunity to become acquainted with intellectual, social and aesthetic perspective that can, from the basis for an expanded plan for lifelong learning and enjoyment as well, assist students with their programme at the university.
I propose a programme where a semester of mathematics (the language of sciences) and English composition would be required in the first year of each student's career. Both these courses will help students to develop their communication and analytical skills which will facilitate and enrich further studies. I would further require an additional four courses outside the Field of Students' Major. These courses would be taken one from each of the Humanities, Social Sciences, Physical Sciences, and Biological Sciences. This requirement will only take up about 1/6 of a typical student's programme and will not hinder double majoring, but at the same time give a great opportunity for students to learn some basic ideas from fields outside their own major. Departments can also benefit from such requirement by developing courses that will cater to a general audience.
The Ministry of Research, Science and Technology (MoRST) have contracted me, through my role as Chairman of the Royal Society of New Zealand Standing Committee on Mathematical and Information Sciences, to organize a Review of Mathematical Sciences in New Zealand.
The first phase was to prepare a report on the underpinning requirements of mathematics and its associated disciplines in relation to the socio-economic framework used by Government Science in the PGSF and the associated disciplines and technology requirements. This was subcontracted to Malcolm Menzies of Victoria Link Limited.
The second phase consisted of identifying a time line and action plan for the exercise; the development of survey instruments for assessing research and user requirements to meet gaps and opportunities for the sectors identified and to develop a discussion paper to consider issues identified in the Terms of Reference. This paper was presented at a session during the Australasian Mathematics Convention and NZ Statistical Association Annual Conference held recently in Auckland. (Copies of the paper are available from Jeff Hunter). The results of this exercise will feed into the final report.
The final phase centres around the following Terms of Reference:
Through a foresight exercise, prepare a report on future likely developments in mathematical sciences in New Zealand and internationally, and assess their impacts on 1) other science disciplines, 2) socio-economically driven science (i.e. PGSF outputs), and 3) the socio-economic sectors of the New Zealand economy and society, including supporting mathematical services for these sectors over the next 25 years. The report should use the Knowledge Base and other reports on mathematical sciences and be able to contribute to any future priority-setting for science, particularly for the PGSF.
For the field of mathematics identify: knowledge trends, and likely developments (where is the strength, where is it developing internationally and in New Zealand); performance outside and inside New Zealand in identified "gaps"; "breakthrough" areas of mathematics and its applications; the implications of new technologies in computing, information and communications in the use of mathematics; the opportunities for socio-economic sectors within New Zealand in mathematical developments and the supporting needs for mathematical services; implications for PGSF priorities including any shift(s) in socio-economic science priorities; the enabling science capability required to meet identified opportunities including infrastructure, human and other resources.)
Issues such as links with interdisciplinary science and international linkages should be considered. Development of the report should involve consultation with key providers, users and funders. The study should also fully assess currently available information and analysis relating to mathematics foresight in order to avoid duplicating effort. Bibliometric and other quantitative analysis should be used where appropriate.
The review will be modelled along the lines of the very extensive review carried out in Australia which culminated in the widely publicized document "Mathematical Sciences: Adding to Australia". A small Secretariat has been established to coordinate the review and subject area coordinators are being appointed to facilitate input into the review. Towards the end of the year, survey results will be presented in a series of regional meetings (Auckland, Hamilton, Palmerston North, Wellington, Christchurch and Dunedin). From these meetings a set of findings and a series of recommendations will be compiled and presented to a Workshop, to be held in Wellington late in the year. This will then be followed by the writing of the final report which is due for submission to MoRST in April 1998.
Although the review has been commissioned by MoRST, the mathematical community needs to take full advantage of this window of opportunity to inform government of the problems being faced not only by researchers but also the difficulties being experienced across all sectors of the mathematical and statistical disciplines.
Your interest and involvement in this exercise will be welcome. Submissions on any items of the Terms of Reference can be made at any time to Professor Jeffrey Hunter, by mail (Massey University, Private Bag 11-222, Palmerston North), by email, or by fax (06 350 2258).
On Friday, 27 June 1997, Professor James Ho presented a seminar through the Centre for Continuing Education (and in association with the ORSNZ) at the University of Auckland. The seminar was titled "Internet Strategies: Beyond Web Sites and Home Pages" and was a full 1-day program starting at 9am.
Jmes Ho, professor of information and decision sciences at the University of Illinois at Chicago, visited NZ earlier this year as an Erskine Fellow at the University of Canterbury. He is author of "Prosperity in the information age: Creating value with technology - from mail rooms to boardrooms" (1994). His latest work on evaluation of the World Wide Web has gained international recognition and is featured as an business resource by numerous business organizations and international media.
Business on the Internet holds tremendous promises of opening up global markets and streamlining transactions. Professor Ho's seminar was aimed at examining the critical issue of value creation over the Internet, assessing current practice worldwide and offering practical ideas and suggestions for those involved in creating strategies for their business on the Internet.
The seminar was advertised as being aimed at Senior executives and managers. Regrettably, it was not well attended with approximately 8 people external to the University with an additional 4 - 5 people being made up of interested ORSNZ and university staff.
It was interesting to see the results of the application of the purpose-value framework that Professor Ho had developed. It analyses Web sites from the customer's perspective of value-added. He applied it to a sample of 1000 North American web sites (see "Evaluating the World Wide Web: A Global Study of Commercial Sites", J. of Computer-Mediated Communication, 3/1, 1997; http://207.201.161.120/jcmc/vol3/issue1/ho.html) and then followed this up with comparative studies (with smaller samples in 20 selected industries) in 8 other countries. It was clear from these studies that the use of the Web to process business transactions is largely undeveloped ( as at the time of the study - 1996) and this is a key area where Professor Ho felt competitive advantages would arise.
This study and the issues that he highlighted would be useful in defining a strategic approach to creating a presence on the Web for businesses. However, I felt the majority of the seminar was pitched at a level below the participants knowledge base. I believe most participants were aware of the Web's presence (and hence did not need a long introduction) and were more focussed on how they could use the Internet to gain a competitive advantage and whether it was worth investing the time and money into having a presence. Not enough time was spent discussing how businesses could develop strategies to make use of the information gleaned from the study. Later in the day, Professor Ho dived into the Web and presented some examples of innovative Web sites. This was fun, but I think that participants would have liked to have seen more examples besides monster.com, and those with links to Professor Ho's own site.
The afternoon session was aimed at giving the seminar participants time to use the insights they had gained from the course to build a "live case". Unfortunately, many of the participants had decided to leave before this session, and so the session amounted to a tutorial led by Professor Ho based on how one might design a Web site for a school or university. The workshop concluded with an enjoyable social hour.
Massey University, like everyone else in this universe, is going through some changes, and it will not be an exaggeration if I report to you that the former faculties now constituting the College of Sciences (The Mathematics Department's future home) are going through rather extraordinary metamorphoses. As of the first day of January 1998, all departments cease to exist and faculties will be replaced by "Institutes". The Department of Mathematics has "decided" to join the Institute of Fundamental Sciences (i.e., to be with Physics and Chemistry). The Operations Research Group seems to have more choices for its future home, but due to the lack of information we have yet to arrive to an optimal solution. Presently we have four options:
If we want to keep the Stochastic and the Deterministic parts of OR together our choices will be limited to options 2 and 3 with the latter being a rather risky proposition. Will keep you posted . . . . .
Of course, I cannot write this column without mentioning the Journal of Applied mathematics and Decision Sciences. The first issue of JAMDS is out, and I have made it available, free of charge, on our Web page (http://fims-www.massey.ac.nz/~maths/jamds/). This is a limited time offer and I hope that it will give you an opportunity to learn more about JAMDS.
By the way, John Giffin is still in hibernation and will not come out 'till Lois & Clark is back on first-run TV.
The big news is obviously that we have Professor Fred Glover of the University of Colorado, Boulder, as a Distinguished Erskine Visitor until early December. Fred Glover needs no introduction. He is Mr Tabu-Search (a nice hyphenated name) in person. He will offer a number of seminars and do some joint research with John George and Ross James. He and John have organized a Tutorial on 'Tabu Search in Heuristic and Exact Methods for Integer Programming' at the Melbourne APORS Conference. So those of you who will not make a pilgrimage to Christchurch, you will have your chance to meet him in Melbourne.
After almost 7 years with us, Bruce Lamar has decided to return to the US and taken up a research position with the Mitre Corporation in Boston ([email protected]). He made a valuable contribution to the department during this time beyond his single-minded determination to improve the world of networks and talk to us through NETSPEAK. True to his traits, he packed the marked assignments of our graduate class and shipped it to the US. He promised to airmail them back to us!
James W. Bryant from the Sheffield Hallam University ([email protected]) will visit Canterbury for a few days on his way to the APORS Conference. He will give a workshop on Drama Theory, which will also be the focus of his three papers in the Problem Structuring Stream at APORS.
The Management Science group in our department has repackaged our current second- and third-year courses into half-year courses (worth 3 points). In part, it simply meant splitting existing courses into two parts. For others, the material covered was repackaged, with a slight shift towards a fuller coverage of production and operations management by the introduction of courses in supply chain management and in project management. On the OR side, we have a new case course in modelling at the second-year level with plans to introduce another case course at the third-year level.
They have all taken cover for the imminent explosion of ?NZ First', or should they be renamed ?NZ Once'?
this time!
MoRST review: Andy Philpott has been asked to assume responsibility for the Operations Research component of the MoRST review of mathematical sciences in NZ, being conducted by Professor Jeff Hunter on behalf of the Royal Society. Selected members of ORSNZ may be contacted by Andy in due course with respect to this exercise.
Financial position: Andrew Mason reported that we are in a good financial position with unaudited Net profit before adjustments of approximately $4700. Our current membership consists of approximately 160 members.
Election of new editor for OR Newsletter: Hans Daellenbach has tendered his resignation as Editor of the OR Newsletter. Council would like to thank him for his huge contribution. If anyone could propose a new editor, please contact Andy Philpott.
ORSNZ Visiting Lecturer: One application was received for the ORSNZ visiting lecturer. He was John Ranyard ( see note below).
Representative for Manila conference: The next IFORS conference is in Manila. A representative from the ORSNZ has been invited to speak to them about OR in NZ. No funding has been catered for this, but Council will give a grant in aid of $200.
Other business: Bruce Lamar has resigned from his position as Senior Lecturer in OR at Canterbury University and will be moving back to the USA. Council wishes Bruce well. Steve Butt has resigned from the Auckland University and has taken up a position at Michigan. Council would like to thank Steve for his invaluable assistance on affairs of the ORSNZ and wishes him well in his new position.
Membership on Internet: Do not forget that, if you agreed to publication of your name, it is now available on the web (http://www.esc.auckland.ac.nz/Organizations/ORSNZ/)
This will be my last pre-APORS item, as by the time the next newsletter comes out, APORS will be already over! With this newsletter, you will be receiving an APORS'97 Brochure, containing program information and a registration/accommodation booking form. As you will see, there is an exciting program arranged. There are around 470 invited and contributed papers, 19 invited streams, 10 tutorials, and 2 keynote speakers, including our very own Dr Grant Read, plus 2 post-conference workshops. There are also two further post-conference workshops, to be run by the Avraham Y. Goldratt Institute to complement the Theory of Constraints stream at APORS, for which you should also find a flyer enclosed.
In addition to the fine conference program, the APORS team have also assembled an array of enticing social events for participants and partners, including a welcome cocktail reception, dinners, shopping, and Aboriginal arts. Additionally, there are no papers on the Wednesday afternoon: instead there is a choice of two trips, either to the penguins or the sanctuary and winery, or golf! I gather the fairy penguin parade is a "must" for tourists from many parts of the world, though for me the wildlife sanctuary and winery tour will no doubt win the day, and I know many of my colleagues will choose the golf at RMGC.
If you have not already done so, just take a look now at the brochure. And if you needed an added incentive, our one-and-only ORSNZ Annual General Meeting will be held on Monday 1st December from 6 - 7:30pm. There'll be a chance to unwind after the day's papers, catch up with old friends and make new ones, before the serious business of the AGM, and dinner at Melbourne's Southbank complex. Sounds too good to miss?
The organisers estimate that around 400 people will attend. Many leading researchers from all over the world will be attending, including some high-profile people. For example, I have just heard that: the current and immediate past presidents of IFORS will be there, the inaugural president of APORS will also attend, and there are also quite a few high profile people from Europe. And the nice thing about APORS is that you will be on exotic yet familiar territory and it will be easy for you to meet these people.
You may like to check the data-base to check on abstracts, streams, attendees, etc. The web site address is:
http://www.maths.mu.oz.au/~worms/apors/apors.html
We are also likely to have many delegates come to NZ en route, and we have arranged a local travel agent to assist overseas visitors plan their visits to NZ. You may wish to expand your networks, and invite some of them to visit you. If you are interested in hosting visitors, let me know and I'll see what I can do.
Any other questions or comments, please feel free to contact me, and I look forward to seeing as many as possible in Melbourne in November.
Early bird Registrations close on 30 September.
Accommodation can be booked up through PRCC till 14 November, but they advise to book early to avoid missing out on your preferred choice.
Conference dates: 30 November - 4 December
P.S. The annual conference of the Australian and New Zealand Academy of Management is also on in Melbourne, during the latter part of the same week, and may wish to take advantage to attend parts of both. I can provide details for anyone interested.
The ORSNZ Council has awarded the newly-created ORSNZ Visiting Lecturer award to Dr John Ranyard for 1997. John will be visiting us in December, following APORS'97. He will be giving talks in Wellington and Auckland, on 8 December and 12 December respectively.
John is the Editor of the prestigious Journal of the Operational Research Society, the official journal of the OR Society in the UK. He will be talking on trends in OR practice in the UK. John is well qualified to speak on this subject, having been a practitioner in industry for nearly 30 years, a manager of a 25-person OR group at British Coal for 20 years, and the co-author of a recent study, sponsored by the OR Society. That study looked at the success and survival of OR groups in the UK, and noted the changing pattern of OR practice, including the dispersal of some central OR groups and the growth of external OR consultancy. The study also came up with some lessons for the continued success of OR practitioners and OR groups, which he will share with us.
John also chairs the OR Society's membership committee, and he would be keen to discuss the results of a recent survey of overseas members, and talk to ORSNZ members about future developments of the JORS and OR Society's services to remote members.
We hope as many members as possible will be able to attend the meetings which, given the time of year, are likely to incorporate a bit of festive cheer, as well as an opportunity to discuss the state of OR here and in the UK (… and elsewhere if you fancy).
Any enquiries or suggestions about Dr Ranyard's visit can be sent to Andy Philpott, or directly to Vicky Mabin, who is coordinating his itinerary.
Harrah's and Hampton Inn at Harrah's, Reno, Nevada, USA
Pacific Rim Track submissions: Prof. Miles G. Nicholls, Grad.
School of
Management,
Swinburne University of Technology, POB 218, Hawthorn, Vic.
3122, Australia
[email protected]
Deadline for papers: 1 Octobder 1997
Universit� Libre de Bruxelles, Brussels, Belgium
Information: Martine Labb�
[email protected]
http://smg.ulb.ac.be/co98
Deadline for papers: 30 November 1997
Queen Elizabeth Bonaventura Hilton, Montreal, Canada
General Chair: Paul Mireault, �cole des Hautes �tudes
Commerciales,
5255 Avenue Decelles, Montreal, Quebec
[email protected]
Quebec City, Canada
Information: Jean-Marc Martel, Universit� Laval, Sainte-Foy,
Qu�bec,
G1K 7P4, Canada
[email protected]
University of Virginia, Charlottesville
Abstract deadline: 5 Januaruy 1998
Chair: Yacov Y. Haimes, University of Virginia
[email protected]
http://www.virginia.edu/~risk/mdcm98.html
Chair: Jacob Hornik, Tel Aviv University, Recanati Grad. School of Mgt., Ramat Aviv 69978, Israel
Brussels, Belgium
Contact: Jaques Tegham
[email protected]
Deadline for paper submission is Dec. 15, 1997
Lancaster, UK
Information: Chris Barrett, Operational Research Society, 12
Edward Street,
Birmingham B1 2RX, UK.
[email protected]
Seattle, Washington
Chair: Marisa Altchuler, Boeing Computer Services, P.O.Box
24346 M/S 7A TH,
Seattle WA 98124-0346
[email protected]
Chair: David F. Rogers, University of Cincinnati, Ohio,
45221-0130, USA
[email protected]
Friendship Hotel, Beijing, China
Chair: Prof. Zhang Xiang-Sun, Institute of Applied
Mathematics, Academy of
Sciences
[email protected]
(Note 11 is eleven)