In Search of Techne

Ch 3.6: Complex Systems and Operations Research

(c) Roy Johnston 1999

(comments to rjtechne@iol.ie)

This chapter includes, as well as the Irish Times material, some of the writer's contribution to the 1972 Dublin IFORS conference, as well as some contributions to a philosophical discussion in the ORSI Bulletin in 1981-83.

December 29 1971

I had the pleasant job of inaugurating the first national conference of the Operations Research Society of Ireland at Dundalk on November 25-27. This conference arose as the culmination of my Presidential year of office. It resulted from the fact that at the start of the year there were no papers forthcoming from the members.

It being the President's job to conjure the papers out of the members, some new departure somewhere was necessary if ORSI was to survive. The conference did the trick. 'We have no papers, lets call a conference' I submit as the magic formula for beating the papers out of the bushes. There is a dynamic here which is worthy of study. No-one will displace themselves to write a paper to read to a dozen people on a damp Tuesday night in February; they will, however, deliver the goods if offered an audience of 60 people for a half-week or weekend, with the possibility of interacting with a dozen other paper-writers and various critics over beer and meals in a congenial environment.

The ORSI conference this year had the added attraction that it involved a competition, with a prize of £60, for the honour of presenting a paper to the IFORS(1) conference scheduled for Dublin in August 1972.

....Enquiries about participating in the 'workshop groups' should be directed to the writer of this column, who happens to be on the IFORS programme committee, and has undertaken to produce some people from the Irish scene with whom the working groups can interact.

The ORSI conference produced sixteen papers, of a high standard of originality. Text-book solutions were the exception. I have no space to summarise the papers here, except to mention the winning entry: Brian Linehan, of the Department of Finance, produced a model for the deployment of an ambulance service, and used it for making quantitative predictions of the level of service in the Western Health Board area....

The head of the Department of Finance group is John Cantwell, who did his OR training at Lancaster. The group has recently been joined by Aoileann ni h-Eigeartaigh, one of the first crop of TCD graduates of the MSc programme in Statistics and OR.....

August 9 1972

I attended on May 30 a seminar in TCD organised by Professor David Spearman of the Applied mathematics Department. This was a conscious attempt to bring applied mathematics into biology, especially genetics and ecology.

Professor Spearman opened with the classical two-species model, for the 'prey-predator' and 'competing for resources' cases. The former, which has an oscillatory solution, is capable of explaining quantitatively phenomena occurring in nature: cycles of population growth and decay are usual in species such as the lemming and the anchovy.

Dr David Jeffrey produced ideas for an ecological model of the oak-caterpillar system, where a 70% pesticide kill brings a 30% reduction in next years population, while a 99% kill gives a 200% increase, due to the fact that you have also killed the predators.

Dr Paul Dowding described work on an energy balance model for soil organic matter. Subsequently there were a number of short 'previews' (this is a procedure borrowed from the DIAS mathematical seminars) from DJ McConnell (who sought, rather prematurely, I thought, an information-theoretic approach to differentiation), RE Moore (mice on an island), EP Cunningham (animal breeding policy), RE Blackith (trigger effects, limiting factors, what determines the percentage of the biomass devoted to the production of the young?) and the present writer (on the solution of some differential equations governing the growth of yeast, the absorbtion of nitrogen and sugar and the production of alcohol, using an analogue computer).

Professor Andrew Young (NUU) came and listened; although he didn't contribute he regarded it as worthwhile having come. I mentioned his work with the Agricultural Institute at Glenamoy on the ecology of the liver-fluke(2).....

On the same day I heard Vincent Watts, of Arthur Anderson and Co ltd, outlining his ideas on 'the measurement of hospital output' at a meeting of the ORSI.... His theme was the problem of defining the term 'better' in quantitative terms. He considered whether it was possible to invent a measure dependent on the disutility of being dead, and to assume a discount rate for morbidity in the future. The term 'better' is then defined in terms of weighted sums of discounted morbitities for treated and untreated cases. Such a measure gives being dead after admission when sick as 'better' than being dead after admission if not so sick, so clearly it is not good enough. So he drew in and looked for measures of disability, pain, distress etc which would be reproducible when estimated by various different types of medical specialists. He was able to relate these to High Court judgments on accident awards, so that the value of the improvement of a patient's condition passing through a hospital system could in some way be quantified.

Value-judgments of this nature are being made all the time in hospital investment programmes. What fraction of the total hospital expenditire should be in the intensive care unit? The methodology developed by Vincent Watts, and others in the OR field, is likely to be called upon increasingly as public expenditure on hospital services increases. There is a 'workshop group' on this area in connection with the coming IFORS conference...... Some Regional Health Board people are involved.

August 23 1972(3)

Dr Naylor's paper puts the philosophy of simulation and validation before us in a comprehensive yet concise manner.

In response, I make three assertions which complement it, and attempt to defend them:

1. The ratio of simulation to analysis in a model can have any value between zero and one; it is not a binary variable.

2. An experimental model of a real system ought to be designed to as to produce a big signal, well above the noise-level, if insights are to be usefully gained.

3. There is a need for some objective common measure in 'multiple-response' situations; subjective measures,such as utility, are of doubtful value.

As regards the first assertion, Dr Naylor gives the impression that he regards simulation and analysis as two separate camps. I suggest that it is possible to build hybrid models, containing elements described analytically, interacting in response to signals one or more of which are simulated.

If I may take an example from my own experience, in the early days of airline real-time reservations experience (1963-64), little or nothing was understood about the statistics of the response of random-access computer systems to an environment which manifested itself as a stochastic demand for service of randomly-varying type. Serious errors were made in the design of real-time systems as a result. Simulation experiments were done subsequently to try to discover the nature of the design faults, and valuable experience was gained.

At this time the writer was working for the Irish national airline and was able to produce an 'analytical simulation' of a prospective real-time system, in which the results of queue-theory were used to predict waiting-times consequent on the levels of demand for services by the central processor, for the 'channel' serving the files, and for the files themselves. The demand for services of each of these elements was calculated from a knowledge of the number and type of 'messages' which were 'in the system'. 'Messages' were specified by patterns of demend for services by the central processor and files. The 'next message' was chosen by the one Monte Carlo procedure in the model(4).

This 'analytical simulation' gave results in broad agreement with the traditional simulation developed by the manufacturers; it predicted the same main bottleneck, namely the file-access channel, which in the next generation of equipment was multiplexed. The manufacturer's simulation took some man-years to develop, and hours to run. The analytical simulation took man-months, and minutes. This work is on record in the proceedings of the 1965 AGIFORS meeting at Chicago.

The writer was convinced by this experience of the wisdom of practising the maximum economy in the use of random variables..

Turning now to the second assertion (on the need for a big signal): Nature has provided an abundance of noises with which the signals describing the states of our systems are masked. The philosophy of experimentation which the writer imbibed when working as a physicist consisted in constructing an experimental model of the system which reduced to the minimum all extraneous noise and focussed on those signals which were considered to be the key to the essential dynamics.

I get the impression that the type of simulation philosophy expounded by Dr Naylor is at variance with this: he feels it necessary to make the noise-level of the model system comparable to that in the real system, so that he is in the happy position of needing to use all the tricks in the statistician's bag to pick the signal out. In physics, the lore used to be that if someone had to resort to sophisticated statistics, his experiment was suspect....

Fortunately, in business systems nature has provided a filter with which the experimenter can pick out a relatively small number of significant variables from the mass of background. This filter is the judgment and experience of the manager; those who have not developed a feel for the significant variables from the experience of working the system are unlikely to have survived. It therefore becomes possible to make meaningful simulations of complex systems, giving large and clear signals, provided we allow ourselves to use judgment and experience in the suppression of irrelevant noise. The writer prefers this road to that mapped out by Dr Naylor.....

..On the question of multiple responses and the utilities approach....(there is) an alternative approach in which the sole measure is the survival probability of the system as a viable organism. This approach leans on insights from current work on aging in biology, which depends on information-theoretic concepts such as garbling of coded data, entropy levels etc.

Consider a 'thermodynamic' model, with temperature and entropy defined in information-theoretic terms. One can envisage an economic organism (a firm) ingesting nutrient from a disordered environment, ordering the ingested material into finished products, which are placed with precision (using intelligence) in a disordered market. Each step involves entropy reduction, which costs money. It is useful to reflect that the product of entropy and temperature has the dimensions of energy, which is equivalent to money. Each entropy-change step in the process must occur at an associated 'temperature', which may be conceived as a measure of management ability. A 'hot' management can reduce entropy rapidly. (Think of Maxwells Demon: the temperature gradient he can maintain across his barrier is the measure of his ability to measure molecualr velocities, sort them, and react in such a way as to deflect or accept them.)

An economic organism is viable if its revenue from sales exceeds its total costs; the cost function includes 'volume times unit-cost' terms and 'temperature times entropy' terms. This approach therefore contains the embryo of a theory of management costs or 'overheads', as the core of the analysis of the viability of the economic organism. I commend this approach to the theoreticians. It is related to a body of theory which is developing in biology, with sound roots in physics, thermodynamics and information theory. It represents a radically different approach to that of utility-theory, although both relate to measures of value. It substitutes for a multiplicity of subjective sub-goals a single over-riding goal, survival....

..May I finally make a plea that the theoreticians should pay more attention to the challenges thrown up by the practice of OR. The theory/practice ratio is out of proportion to that which obtains in other branches of applied science. Because theoreticians are solving problems posed by each other and neglecting the experimenters' world, important areas such as that outlined by Dr Naylor are neglected......

February 21 1973

I attended a seminar at Abbotstown (at the Department of Agriculture Veterinary Research Station) on January 9, at which a number of people who had been attempting to quantify some of the complexities of biological systems got together and changed ideas. This had been organised by Ken Strickland of Abbotstown.. Professor RP Lee (TCD Veterinary Department) presided..

Much time was spent on the life-cycle of the liver-fluke; this parasite causes a continuous leakage of good protein back to Mother Earth from our sheep and cattle. The extent of this leak, if one includes general loss of condition as well as acute cases resulting in death, is estimated to be of the order of 10M pounds annually.

The background was outlined by MJ Hope-Cawdrey, of the Agricultural Institute, Creagh, Co Mayo.

The parasite undergoes an asexual phase of reproduction inside the bodies of a particular species of snail. The drained bog at Glenamoy, Co Mayo, which now carries good grass, unfortunately provides an ideal habitat for this mollusc; it has been said that the chief outcome of the Glenamoy work has been the development of good field conditions for the study of the liver-fluke, rather than economic reclaimed land!

The purpose of any mathematical model of this system must be to show how the parasite responds to treatments at the various sensitive points in its life-cycle: does one (a)reduce the viability of the eggs by reducing the moisture level (b)apply molluscicide or (c)dose the animal with a specific flukicidal drug?

As background to ecological model-building, Professor Andrew Young (NUU, Coleraine) developed the consequences of variability in population statistics. By manipulating simple propability concepts, he showed that for a particular species of bird on Rathlin Island to have a 90% survival probability after 40 years, you had to adjust the survival rate per generation in such a way that there was a 5% chance of a 'population explosion'. In other words, with small populations, you get a compounding of the variability effects.

Professor Young also described a mathematical model of an inland fisheries system, involving pike, trout and five or six interfering species. He was able to adjust the pike cycle parameter until the catch records were matched. Then changes were introduced and the effects of various conservation policies predicted. One fact which emerged was that the trout population was improved by the presence of some pike, due to the positive effect of culling.

Professor JNR Grainger (TCD, Zoology) described some work on the hatch-rate of liver-fluke eggs as a function of temperature. This is clearly an important co-efficient in the life-cycle model, providing a link with the meteorological records and a possible basis for prediction of an optimal dose-time.

Michael Connaughton, of the Meteorological Service, outlined some attempts that had been made to develop a predictor by Dr Ollenshaw in Anglesea. One assumes that between May and October temperatures are acceptable and that the rate of growth is moisture-dominated. What seemed to me to be a rather poor forecaster emerged; the final forecast was, in fact, based on informed judgment.

It struck me that it would not be too difficult to improve the model by building in the Grainger results: allow the growth to take place at a rate determined by the temperature, when 'switched on' by the presence of a threshold level of moisture.

Finally, George Gettingby, of NUU, outlined his attempts to build a multi-stage mathematical model of the liver-fluke system, along lines similar to Professor Young's trout/pike system.

This is good work, representing the vanguard of the contemporary movement to cross the disciplinary boundaries. The Abbotstown meeting, however, I thought was marred by a tendency for the mathematicians to talk shop to each other, to the exclusion of the applied biologists.

March 28 1973

The output of the Dublin Transportation Study has been discussed in the press; no doubt it will get continuing attention as the various phases impinge on the public consciousness.

It deserves notice in this column because of the way it was done: an ad-hoc team, including statistical and model-building expertise, using mainly local talent, was set up under an Foras Forbartha.

This team included G Declan McIlraith, a TCD graduate with post-graduate experience in Transportation Planning from Northwestern University, Evanston, Illinois; he is now working with the Department of Local Government.

Declan McIlraith outlined the methodology behind the plan at a meeting of ORSI on March 22, giving a masterly account of what stands up as a competent piece of statistical analysis and OR model-building.

The key features are well known: the outer ring motorway, the Macken St bridge, reserved priority routes for scheduled public transport and an upgraded rail network.

The model starts from a basis of 300-odd communities of about 1000 households, each having a defined set of socio-economic characteristics established by a sampling procedure, generating a transport demand which gives a predictable aggregate.

One can quarrel with the socio-economic assumptions (eg increasing inequality of wealth, too low a % of local authority housing), or the national macro-economic assumptions (eg assumed availability of oil at a stable price in vastly increased quantities(5)), but this does not alter the basic soundness of the methodology. A different government might make different assumptions and come out with a different public/private mix as the 1991 output from the model.

The interesting thing is that despite the conservative political assumptions, the public transport network and the general amenity situation comes out strengthened. I find this encouraging.

I also find interesting the use of the 1000-house community unit. This is a statistical convenience, but it is also food for thought. Does it not, perhaps, reflect a reality which should be the basic unit of local government?

Most important of all: here is a piece of expert team-work, carried out by an ad-hoc group which is now dispersed. While it was together, it developed an expertise which was well up to international standards. Is it too late to ask that AFF consider recruiting so as to set up a stable core of urban transportation planning expertise, which would then be available, not only to update the Dublin plan continuously, in a changing situation(6), and produce plans for other Irish urban situations, but also to sell urban transportation planning on the export market?

Possibly the first step into the 'export' market might be to make a bid to take on the quantification of the implications of the Copcutt(7) plan for Derry and the North-West which has been presented recently to the Brussels Commission. It would be a pity if this opportunity to demonstrate the essential Irishness of the area were lost to an English firm of consultants who would probably be insensitive to the positive political opportunities for promoting a peaceful all-Ireland settlement.

Recruiting staff, in this case, need not be a cost to the AFF budget. The staff concerned would be revenue-earners.

December 12 1973

I have to hand a transportation study of the CIE bus system, produced jointly by J Markham of the CIE Corporate Planning and Management Services and PN O'Farrell of the Institute of Science and Technology, Cardiff. It is published in 'Transportation Planning and Technology'.

The study consists of a survey of bus journey times on various routes at various times of the day over a four-week period. Some statistical analysis was done, which proves, among other things, that Friday is the worst day and that buses go more slowly in the centre of the city.

I seems to me that this work is incomplete. Some measures have been provided of a base-line situation. We now need measures of a changed situation, namely the present one, with the car population cut by 20%(8).

We have been handed on a plate an 'unplanned experiment' from which some real conclusions can be drawn, enabling CIE and the long-suffering consumers of of public transport to discredit for once and for all the position of a certain gentleman who purports to represent the interests of the private motorists in the correspondence columns.

We need a measure of the total (ie car and bus) passenger miles per hour travelled at peak, as a function of car density. All perceptive users of transport in cities are agreed in a conjecture that if the latter comes down the former will go up. A suitably planned experiment along the lines of that carried out by Markham and O'Farrell, spanning the period of development of the oil crisis, would enable this important differential effect to be measured definitively.

If this differential indeed proves to be negative, it would then be necessary to publish, not only in the learned journals with statistical jargon, but also in the newspapers and on RTE, in the form of a publicity campaign, backed up by a guaranteed high-frequency service all day in the city-centre. To do this CIE would need bus lanes and further restrictions of traffic in the city, backed by the necessary public education campaign.

In continental cities, where they very sensibly kept the trams, a separate independent signalling system gives the trams priority, and no-one would be foolish enough to stop a vehicle across the tram-tracks. By this means the vast majority of the travelling public are transported.

Dublin, in contrast, is just about the most individualistic city in Europe, as measured by ratio of 'own vehicle' to public transport for the journey to work.

Public transport users in Dublin should, I suggest, use the opportunity of the oil crisis to unite, organise and ally themselves with CIE in a broad campaign to rescue Dublin from the self-defeating individualism of the private car-owner.....

March 14 1974

Readers of this column will be aware of my continual search for the ideal public transport vehicle. One, to which I have previously referred (it is in use in Germany), is powered by a small, quiet engine matched to the vehicle average load (i e about 10 H P for a bus; most of the power in conventional vehicles is used to give acceleration at start-up). This charges batteries, which take up the acceleration load, and can be charged regeneratively during braking.

According the Novosti (the Soviet news agency) the Kursk Polytechnic and the Lvov bus design people have got together and come up with a system analogous to this, except that instead of a battery they use a flywheel. Braking energy is used to spin up the flywheel. A 50% fuel saving is claimed.

The Swiss have a flywheel-driven bus which connects to an electric power supply at stops. The flywheel is spun up while the passengers are getting on and off.

Perhaps for a small percentage of CIEs current fuel bill a development project could be got going? There are plenty of people around who would make mincemeat of the design and production work. This is not a mass- production system, nor a high-technology device development problem. It is system-engineering, with standard, readily-obtainable parts.

February 7 1975

I make no apology for coming back to the topic of public transport, as the rapidly rising price of petrol is making many people pull out of city motoring, and indeed car ownership. The problem is seen by the aspirant public transport-user at the service-level: he is faced with 15 or 20 minute waits at bus stops. So, naturally, he (or she) becomes resigned to going back to the car.

Some time ago, in a letter to CIE, I proposed that the system be nodalised. In other words, that instead of long routes, with A and B versions related to low-frequency dog-legs, there should be two types of services: a feeder service, with small buses, feeding nodes connected by large express buses. Such nodes might be Dun Laoire, Blackrock, Rathmines, Terenure, the old city-centre; where the railway-system permitted they would be based on a station, so that the railway would serve, in effect, as one of the express routes in the mesh(9).

The CIE response was to the effect that people do not like to change vehicles; they prefer an infrequent dogs-leg leaving them near to their homes at one go. They had survey results to prove it(10).

I wonder if survey results from an uninformed market are in fact valid? Is there not scope for introducing what could be a better system, along with a creative advertising campaign such as to enlist peoples' support for an imaginative new concept of urban transportations?

I am made optimistic on this score by an article in the 'New Scientist' on January 6 about the bus system in Delhi, India.

(Transport planning is a major problem area where technology and urban society interact; it has been receiving increasing attention from those who concern themselves with using scientific methods to match an appropriate technology to deliver an acceptable service level at minimum cost. Whence the status of the New Scientist, and indeed this column, as the place for remarks on buses in Delhi.)

Delhi, under the old scheme, had 69 routes and 199 sub-routes (what I have called low-frequency dods-legs above). Under one version of the new scheme, eight interchange nodes connected by 28 express routes, were proposed. A compromise centralist version had one central node and seven off-centre nodes connected by a ring route. The final version adopted, under the influence of the topography of the street system, had nine peripheral stations; these were connected by a rapid service at 10-minute intervals from 7 am to 10 pm.

The feeder system was not yet developed; the old system was simply left there, in the hope that it would act provisionally as a feeder system. It was intended to reconstruct it specifically for that purpose over a period.

The statistics show that for quite a marginal re-allocation of resources there is a very striking improvement in the economics of the operation, and apparently the people have responded with enthusiasm. Indeed, surveys show an increasing acceptance of the new public system among the majority of urban motorists. It will be of interest to see how the economics and service levels look when the full system is implemented, and when the target 25% reduction of city traffic has been achieved.

The Delhi system is a perfect example of a 'software solution' to a problem, instead of the hardware solution preferred by bus operators to date (ie 'bigger and better buses', ignoring the frequency problem).

A final point in favour of the nodal system: a passenger does not have to wait miserably in the rain for one of the 10 or 15 buses going west; he or she gets on any bus going west and changes at the appropriate node on to the feeder system. I think, in spite of what CIE says, that people would buy the idea if the overall service level were improved.

May 18 1976

I recently came across a paper published by the Inland Waterways Association, published in May 1975, which urges the need for a quantitative evaluation of the costs and benefits associated with the removal of the Limerick bottleneck from the Shannon navigation. Recent development of barge-carrying ocean-going vessels (an obvious extension of the container principle) has changed the economics of inland waterway transport favourably. There are 8 vessels of this type plying between the Continent and Britain; the smallest and most flexible is the BACAT-LASH system in which the parent vessel carries 10 BACAT 140-tonne capacity barges and 3 LASH 370-tonners.

With some relatively minor improvements to Thomas Mulvaney's visionary Shannon Navigation system(11), it should be possible to develop the Shannon towns up to Carrick as container ports for trade to and from the continent, thereby avoiding the long road-haul to Rosslare, and reducing congestion in Dublin. With a factor of five in ton-miles per unit of fuel consumption as between water and road, and with the present and likely future escalations in fuel cost, water-based transport systems are due for a positive re-appraisal.

June 8 1976

The Chloride battery-driven bus ('Silent Rider')... needs no introduction to the public. There are however some technical features which make it interesting and deserving of success.

The principal technical snag found in the course of the development of a battery-driven car in UCD(12) was overheating in the batteries under conditions of rapid acceleration. This lowered drastically the overall energy conversion efficiency, to the extent that the project was abandoned. Standard car-batteries were used.

Chloride, however, are in the battery business, supplying the existing milk-float and fork-lift truck market. They have consciously developed the design of the batteries to cope efficiently with the work-load presented by a 'traffic-compatible' vehicle. Add in a control system which allows power to flow smoothly at a pre-set rate (without stepping-switches, resistive loads etc as in the old trams) and enables the braking energy to be recaptured, and you have a very attractive vehicle.

The range is 40 miles and the top speed 40 mph; this is stated to be adequate for 90% of city-centre bus-route situations. The Silent Rider has been in use in Manchester since March 1974 under conventional conditions.

Its introduction to Dublin might possibly be linked with a nodalisation of the Dublin system: use the existing double-deckers as expresses between the main nodes, the latter being fed by 'silent riders' (which are single-decker; they could be single-manned at a flat-rate fare) feeder routes....

September 21 1976

I have just completed reading the PhD thesis of Michael J Walsh, who has been awarded his doctorate by Trinity College for work done in connection with Professor W G Foster in the Department of Statistics..

Up to 1970 Michael Walsh was chief engineer and computer services manager for the Irish Sugar Co. While there, he developed a theoretical approach to data-processing in industrial systems which he called 'functional cost analysis'. He read a paper on this at the Internationa Federation of Information Processing (IFIP) at Ljubljana in 1971, which aroused some international interest. Realising that he was in a theoretically unstructured area, he followed up by registering for a PhD with Professor Foster. In the meantime he had left the Sugar Co and gone into consultancy.....; in the interstices of systems analysis jobs for Cement ltd, Euronet, NET(13) etc he managed to get the necessary original theoretical work done and a thesis written. The latter in my opinion should be published as a text of interest to systems analysts and management accountants.

I do not want to go into the technicalities, but....I can state that one of Michael Walsh's tricks is to start off with the accountants' conventions regarding variances of price and usage, and to generalise this into a finite-difference calculus which in turn can be generalised over a hierarchy of variables of diferent rank. The accountancy convention then re-emerges as a more basic, and philosophically sound, rule: 'no variable should be credited or penalised with any element of variance associated with a variable of higher rank'.

I look forward to what I can call the 'Walsh procedure' (whereby mature technogists crystallise their theoretical thinking derived from practical problem-solving into a PhD thesis) becoming more the norm in postgraduate work. It would enrich academic life no end, especially if funding were available to enable someone to participate fully in College life. There are at present few such opportunities; the Player-Wills Fellowship is one such. I look forward to seeing the present incumbent, Padraig O Hailpin, complete and publish his history of the State-sponsored bodies(14).

July 1981(15)

Without wishing to disparage those of the OR fraternity who are working at the currently very active 'microcomputer applications' frontier, bringing easily-accessible decision-models to various productive decision-points, may I make a plea for some of us at least to pick up the mantle of the classical OR people and give consideration to global matters such as the political economy of nuclear warfare.

Classical OR, as it developed in World War II, was concerned to take laboratory research (such as led to the production of magnetrons etc) 'downstream' into the operational field, by developing (eg) tactics for the use of radar systems, with the aid of models of battle situations (deterministic or statistical).

In the post-war environment, the 'statistical models' aspect took off into the business field, with some success, though there are some who say that it was oversold and has retreated into academicism, especially in the US.

The deterministic aspect of model-building took off via Forester's 'industrial dynamics' and 'global dynamics' in a direction which culminated in the Club of Rome work.

The 'evaluation of new technology' aspect, however, I feel has been somewhat neglected, except among the war-gamers, where the estimation of projected post-nuclear-war scenarios has become an implausibly precise art (Herman Kahn etc).

It would be instructive to survey what those who originated the concept of OR have done since the war. My knowledge is incomplete, but let us take two, J D Bernal and Stafford Beer.

Bernal's involvement with classical OR was in the planning of the Normandy landings in 1944. His way into OR was typical of the epoch; he was primarily a physicist specialising in crystallography; he had an unusually acute feel for measures of performance of complex systems. He did not stick with the post-war OR scene, as it developed. He divided himself in two; he went back to his laboratory in Birkbeck, and simulataneously devoted himself to the problem of geopolitics and the prevention of World War III. By and large his work in the latter field has been neglected and forgotten, at least in the West, though it was highly regarded in the East. The problem is still with us, and again becoming acute, after a period of pseudo-stability that some have attributed to the 'mutually assured destruction' (MAD) principle. This pseudo-stability has however now been undermined by technological 'advances' leading to increased estimations of the probability of success of a 'pre-emptive strike'.

What are the OR fraternity doing about this? Nothing, to my knowledge, except to give technical support to the military planners. No-one has inherited the mantle of Bernal, or stood up and said 'divert the resources on both sides from military to civil applications, and everyone will be better off'. Is there not a theoretical approach to this via game-theory, communications theory, marketing theory etc, in some new synthesis leading to global 'detente'?

Stafford Beer, after a period as a successful international consultant in management information systems, has pulled out of the rat-race and retired to a mountain fastness in Wales. His recent book 'The Heart of Enterprise', along with his re-issued 'Brain of the Firm', constitute readable distillations of his wisdom. The new edition of the 'Brain' places on record the work he did for the Chilean Government under Allende (1971-73); the fruition of this interesting attempt to apply cybernetic principles to socialist planning in a developing country was unfortunately pre-empted by the CIA-inspired military coup. The main emphasis of his current thinking remains on the problem of resource-allocation in managed 'third-world' economies.

We need a focus for what can be labelled the 'Bernal-Beer' tradition in OR, which would be socially responsible, outside the multi-national dominated rat-race, and rejecting utterly the nuclear war scenarios so beloved on the US military OR establishment. A case might perhaps be made for the establishment of such a focus in Ireland, Bernals native country.

October 1981

I feel I must continue to develop the theme I initiated at the start of this series, when I called for a return to 'classical OR', invoking the names of J D Bernal and Stafford Beer. I return to this theme with fresh insights, derived from a first-hand look at the development problem in Africa. I had occasion to attend the UN conference at Nairobi on 'New and Renewable Sources of Energy', as a non-gevernmental observer acting on behalf of renewable energy workers in Ireland(16).

The central problem which emerged was that of diversion of global resources from military R and D, and competitive military expenditure generally, towards a co-operative approach to the global energy problem.

The energy problem in the Kenya context (and in Africa generally) takes the form of a high and increasing proportion of export earnings being taken up with the purchase of oil, coupled with the denuding of the forests for firewood, to the extent that by the year 2000 at the present rate there will be none left. Yet under tropical conditions a tree will grow to 30 feet in 3 or 4 years; there is no intrinsic fuel problem but there is a forestry management problem, with socio-political dimensions. In fact, the technology exists for the replacement of oil imports by a developing biomass-based energy production system, producing (eg) liquid fuels with wood pyrolysis gas as feedstock. The problem is in establishing socio-political structures capable of managing the transition to renewable energy technology, both in developed and developing countries.

There is a school of thought which says that the developed world should be leading the way in the transition to renewable energy technologies, leaving an increasing fraction of the declining oil stocks to fuel the the industrial development of the 'south' (shorthand for the 'third world'). This view is held by the existing southern industrial elites, whose training is in the tradition of northern conventional wisdom, and who lack scientific insights. Such a policy is predicted by environmentalists to lead to famine-driven crisis-urbanisation of the rural subsistence populations, and to irreversible decline of the capacity of the denuded and eroded land to grow food.

The main objective of the UN conference was to agree a global action programme whereby the energy transition could be managed. Such a programme was in fact agreed, but largely on paper, as few, if any, additional resources were allocated to implementing it. The Action Programme, however, is a useful guide to bilateral aid in this field, so that the excercise was not entirely a waste of time and effort.

There are other dimension to the global energy problem which affect us all. For example, if dependence on coal increases, levels of CO2 and SO2 will increase, leading to climatic warming (possibly even to the extent of icecap melting), acid rain (already a serious problem in Canada, Scandinavia and Germany) and various dire related consequences. If forest denudadion continues, the global CO2-sink will become constricted, increasing the rate of onset of climatic change.

The net effect of present trends by the end of the century, in terms of soil degradation and climatic change, is measurable in nuclear-war terminology (mega-deaths etc).

The reversal of these adverse trends is not a problem of technology; all the necessary technologies are available to us. It is a management problem with a socio-political dimension.

An important resource is that tied up in the arms race. The force motivating this is 'east-west tension'; the 'south' clamours but is largely ignored.

Is there not a role for OR in formulating credibly the east/west/south problem as a non-zero-sum game?

If there is nuclear war, all lose.

If the arms race continues without war, east and west survive, though with increasing internal stresses; the south loses; global real wealth declines.

If the arms race can be stopped and the resources diverted towards (eg) managing the energy transition, east, west and south all win.

The focus for promotion of this global re-think could credibly be in Ireland, building on its neutral status in the east-west military confrontation.

January 1982

Robert Machol's article on OR in Ireland is of interest in Ireland for a variety of reasons, primarily perhaps for the light it throws on the world-view of the US OR establishment. For example, he seems surprised at our academic/practitioner ratio, contrasting it with that in the US which is dominated by academics who produce large volumes of paper solutions to problems which are posed, in the main, by other academic papers. Machol may regard this as normal; in the opinion of most practitioners in Britain and Ireland, it is pathological. There is a clue to the pathology in the auspices under which Machol writes his survey: the Office of Naval Research. This, presumably, is the paymaster. The funding of academic OR in the US is dominated by the MADmen(17) who live in a lunatic imagined world of global nuclear duels. The present criminal rate of world military spending must be attributed at least in part to this blinkered abstracted analysis of artificial problems by those of the US OR establishment who have sold themselves to the military.

He is surprised to find group MSc projects the norm.... he attributes this to the difficulty in getting problems, so that they have to be shared, reluctantly..... if he had enquired further he would have found that the group MSc is explicitly cultivated as a mode of training in team-work, this being a classical OR principle this side of the Atlantic. He also finds it surprising that College of Commerce students have to submit project theses.. I find it surprising that he finds it surprising: is the project as a training device really so unknown in the US?

If 'important theoretical advances' start coming out of Ireland as a result of Machol's admonitions, I will then assert that OR in Ireland has joined the US pathology. Yet what we are doing is by no means 'technician-level', any more than is the work of the medical general practitioners. We are general-purpose scientific problem-solvers; we know who our back-up specialists are, and we know how to call on them when we need them. We are not a branch of academic applied mathematics, as OR appears to be in the US.

Machol went back through the past Presidents of ORSI for interview material. I'm sorry he didn't get as far as me; I would have liked to tease this out with him.

I found some of his interview material interesting; it hadn't surfaced in our local OR workshops, particularly the Bank of Ireland material.. I now know whom to blame for their overdraft/term loan policy, which as a consumer of bank services I find iniquitous. I can however understand the bank's unwillingness to expose their OR staff to the flak, and find it somewhat impertinent for a foreign publication to expose what was no doubt volunteered informally, 'off the record'. The message of Machol's article is clear: be careful what you say to visiting experts from the US; they may be 'anthropologists looking for eskimos'.

It is appropriate to ask Machol to do a similar study in a State of the US having demographic, economic and geographic factors broadly similar to Ireland, and let us know how we compare. We are not in the super-power rat-race; we can survive quite well in the undergrowth, provided the MADmen can be brought out of the MADhouse and set to work dismantling their nuclear arsenals.

May 1982

I am tempted by the provocation kindly supplied by Robert Machol to develop further the theme of diversion from military R&D, to which a handful of socially responsible scientists (including the writer) are giving their attention via the Pugwash Group, the Irish Campaign for Nuclear Disarmamant, the Irish Futures Society and other such groups. However, I will refrain until there is sufficient material for a significant interim report......

Let me instead develop a concept which surfaced about half-way through my riposte to Machol: OR people as the 'general practitioners' of the science-technology-economy process. I introduced this as a counter-argument to his accusation that because we did not produce papers we were technician-grade people. It is, I think, a valid concept, and it needs elaboration.

Consider the medical system. The front line is occupied by the GP, whose strength is in diagnosis. He/she knows when it is necessary to call in a specialist; the latter are organised in an extensive and well-structured back-up system, strongly linked with the medical schools and research centres where the state of the art is continuously being advanced, often by non-medicals working in academic departments of physiology, immunology or whatever.

The Irish medical system is close-knit, productive and cost-effective; it is highly regarded throughout the world, and does a big export business in training services.

In contrast, the science-engineering system in Ireland is fragmented and lacks the world status of its medical analogue (as a system; some islands of specialist expertise have, to their credit, established export markets). It needs to look to its cross-links, referral procedures, structures etc with the medical parallel in mind.

The analogue of the GP could be said to be the engineer; this is a valid parallel for hardware. Economic life, however, is an organism having hardware and information components. The GP-analogue on the information ('software') side is the management consultant; most OR people are in this general area. There are back-up specialists, like soil-testing for the Civils, SEM(18) analysis of fracture surfaces for the Mechanicals and computer software producers for the management service people. What is lacking, however, is any real bridge into the world of the science faculties; these live in their own worlds closely linked to the world scientific networks, mostly avoiding applied-science (and any consequent link with engineering) like the plague. As a consequence they lack status and recognition among the engineering fraternity as possible sources of specialist knowledge with a role in upgrading the 'state-of-the-art'.

Many of the products of the science-faculties, after a period in the academic rat-race, drop out, sometimes in a state of disillusionment with basic research. Some, like the writer, trickle into OR by chance. Such people who have crossed the 'cultural gap' between science and engineering are acutely aware of the dichotomy, and uncomfortable with it.

My own evaluation of Irish science is that it is world-class in spots, but mostly grossly underutilised, and reserving of more recognition as a potential source of specialist support from the GP fraternity.

On March 30-31 I attended a conference, sponsored jointly by NBST and the Allied Irish Bank, which reported progress on a working group of Deans of faculties of Science, Engineering and Business Studies. This was an interesting and welcome event, providing a structure for broadening peoples' understanding of the innovation process. The input from the science faculties however was weak, aggressively defensive, promoting the thesis that pure-scientific postgraduate research teaches 'creativity', so that the existing system is OK and just needs more money if it is not to collapse. This thesis is open to question. Any follow-up of this conference which leads the scientist into areas of applied research will help them to evolve into a learning situation, in which each department can actively monitor its 'pure/applied ratio' as optimising parameter.

Here is a nice field for OR at the national level.

January 1983

I am stimulated to pursue some of the philosophical paths suggested by Richard Gault(19) on October 7 last year in his TCD seminar. The need for such a pursuit has suddenly become urgent, if I read correctly the message suggested by the quotation from the IFORS Bulletin 6/1982....

The message is one of utter philosophical confusion, overlaid with crises of confidence and identity. The following questions are asked:

*Can OR models be part of a decision support system?

*Are OR models friendly to microcomputers?

*Can the man-model-computer system be made to work?

*Should OR methodology be adapted to the requirements of decision-support, or vice versa?

I assert that the apparent need to ask these self-doubting questions stems from the domination of OR thinking by a false paradigm, in the sense hinted at by Richard Gault but not developed.

Gault asked another self-doubting question: 'Is OR a science?' He went on to outline a Kuhn-type philosophical analysis of science, with the three levels: the scientific experiment, the general law and the 'paradigm'; the latter implies the acceptance in the common currency of a complex of general laws within a field of activity, within the constraints of which one tends to plan the next experiment and interpret its results. (Examples of 'paradigm shift' include the switch from Newtonian to Einsteinian gravitation, the general acceptance of 'continental drift' within a framework labelled 'plate tectonics' etc.)

Gault searched for OR embodiments of this Kuhnian philosophical structure, suggesting that the first level was the development and application of a decision-model in a specific problem situation, while the elaboration of families of such solutions having a 'standard technique' component in common (eg queue theory or linear programming) constituted a second level. Gault however failed to find a third level, and consequently questioned the scientific status of OR.

I now take up where Gault left off, with the following suggestions for the definition of a philosophical structure within which it should be possible to deal with the IFORS self-doubt syndrome.

(a) OR is scientific; its main thrust is the synthesis and analysis of complex systems involving human beings.

(b) There is no such thing as an 'OR model' as distinct from other theoretical models used by scientists to describe systems and to help predict the results of experiments.

(c)There is a paradigm level in OR, governed at present by a false concept deserving overthrow, namely, that there exists a class of models called 'OR models' which when applied to problems give so-called 'optimal solutions', to which the 'decision-maker' (assumed to have an independent external identity) is expected to bow.

(d) The new paradigm which will overthrow the above is one which has been consciously developing in the writer's mind since about 1962, when he first started doing work which was recognisably within the OR canon. It is that there exists a triad: human being/model/system. The model exists in the mind of the human as a perception of the way that the system works.. The model is usually step-wise refined, as regards the quality of its prediction of the system performance, by a series of experiments, often unplanned, sometimes planned. The model itself does not 'optimise'; the human/model/system complex strives to achieve a goal, which is survival (see Stafford Beer). Any attempt to 'optimise' a particular parameter is likely to divert attention from the main goal (survival) and undermine the robustness of the system (this aspect has been developed with the label 'catastrophe theory').

Let us now return to the IFORS questions. Drop the prefix 'OR' from the word 'model'.. The answers are:

1. Not only 'can' models be part of a decision-support system, but every decision-process must involve a model at some level of approximation to a valid description of the system.

2. Microcomputers enable predictions based on models of systems to be produced rapidly, providing the model is well structured and the inputs readily available. 'Friendly' is an irrelevant concept at this level; it refers merely to the quality of the programming.

3. The triad is wrong; it should be 'man/model/system'. The computer has no standing except as an information-processor, for predicting with the model, and comparing the predictions with the actual performance of the system. Of course the (correctly specified) triad will work; it is working all the time. The 'wrong triad' will work if and only if it correctly reflects and interacts with some system outside itself, of which it constitutes the controlling interface.

4. Of course OR methodology must be adapted to the needs of the decision support process, but it can only so adapt if it throws out the concept that 'OR is about optimising' and that 'OR models' are in some way special. It must shift its paradigm as suggested above.

The above deserves expansion in scholarly mode, by someone with the time to develop an approach to OR within the canon of the philosophy of science. The writer would be interested in interacting critically with such a person, with a view to injecting these concepts into the international literature, which, if IFORS is anything to go by, appears to be badly in need of philosophy.

NOTES

1. International Federation of Operations Research Societies. For more on this conference, see Chapter 5.2 (Scientific and Technological Information) on 23/4/70 and 6/9/72.

2. See also 21/2/73 below.

3. This was the writer's contribution, at the 1972 IFORS conference, to the discussion on one of the 'state-of-the-art' papers, that by Dr T H Naylor, of Duke University, North Carolina. Dr Naylor reviewed the various approaches to simulation, with emphasis on techniques for validation, particularly in situations where background noise effects necessitated the use of refined statistical procedures.

4. This area was subsequently worked over by Harry Perros (now at the University of Chicago) in the course of PhD work in TCD under Professor F G Foster. He used the (by then well-matured) Aer Lingus real-time reservations system as data-source. The approach pioneered somewhat pragmatically by the writer, and reported at the 1965 AGIFORS meeting, is now on a sound theoretical basis.

5. The writer in continually amazed by the number of situations in which the November 1973 energy crisis came as a surprise. See Chapter 4.2 (Energy) on 11/7/73.

6. Like, for example, the analysis of the impact of the electric rail system and its associated feeder routes on the overall urban transport scene; there is here considerable potential for urban regeneration.

7. Geoffrey Copcutt, an enlightened English architect and town planner, had produced a plan for urban development in Northern Ireland, the rejection of which by the Unionist authorities contributed fuel to the start-up of the present period of political unrest.

8. The writer made an attempt, without success, to interest AFF staff in the opportunity presented by this 'unplanned experiment'. They were however constrained by a rigid experimental programme, and those in charge at the time apparently lacked the imagination necessary to appreciate the importance of this opportunity to determine definitively a key planning coefficient for the Dublin transportation system.

9. Other elements in the reformed system would be (a) a system map at each node and halt (b) Times in minutes past the hour for the basic schedule on display (c) reserve capacity such as to enable the basic schedule to be doubled up in key sectors during the peak hours (d) location of nodes at actual or potential business development centres throughout the city, thus spreading 'centre-city' business over a wider area.

10. Based no doubt on responses from users of existing routes, rather than frustrated aspirant users of an improved system condemned to a life of own-vehicle frustration.

11. Thomas Mulvaney designed the Shannon Navigation in the 1850s to take ships of 500 tons. He got into political trouble in connection with the Fenian rising of 1857, as a result of which he went to Germany, where he was responsible for the canalisations of the Ruhr. There is a statue to him in Essen. On of the by-products of McLaughlin's Shannon Hydro-electric Scheme was the breaking of the connection between the Shannon Navigation and the sea at Limerick, in order to reduce the level of the tail-race. Some civil works would be necessary to restore this link, but in the opinion of the author, Terence Mullagh, this should be feasible at an acceptable cost.

12. See Chapter 3.1 (Engineering and Manufacturing) on 25/8/71.

13. Nitrigin Eireann Teo, the State fertiliser factory.

14. Alas this was pre-empted by his untimely death in 1978.

15. These and the following essays are part of a sequence requested by the editor of the ORSI Bulletin, for circulation among members. As they are of general applied-scientific interest, and in the tradition of the Irish Times column, I feel that it is appropriate to include them.

16. A Report covering the Irish background to the UN Nairobi conference, and a summary of the outcome of the latter, was produced by the writer and published by the Irish United Nations Association.

17. The acronym 'MAD' for 'mutually assured destruction' may no doubt be attributed to some Pentagon staff OR sour-wit.

18. Scanning Electron Microscope.

19. Lecturer in Operations Research in University College, Galway.

[To Irish Times Column Index] [1970s Overview]

Some navigational notes:

A highlighted number brings up a footnote or a reference. A highlighted word hotlinks to another document (chapter, appendix, table of contents, whatever). In general, if you click on the 'Back' button it will bring to to the point of departure in the document from which you came.

Copyright Dr Roy Johnston 1999