 | The Operations Research Revolution Rolls On, |
Operations research (OR) is the use of quantitative techniques (statistics, etc.) to solve problems and help leaders make decisions. OR has been around for centuries, but in the decade before World War II it came to be recognized as a distinct discipline. Operations research was used extensively during World War II to solve numerous problems; everything from how best to use radar or hunt submarines to running factories and getting supplies to the troops efficiently. Following this wartime experience, there were even more numerous peacetime successes. This led to OR being called "Management Science." OR is still widely used in military applications.
No Respect
Why is it that OR never gets any respect? Operations Research is, arguably, the most important scientific development of the 20th century. OR, like the earlier Scientific Method, is basically a management technique. Management is also a 20th century concept that gets little respect. In some respects, OR is the combination of management principles and the Scientific Method. Without the breakthroughs in management techniques, the enormous scientific progress of the 20th century would not have been possible.
No Visualization
Unfortunately, management is, like the air we breathe, something most of us take for granted. But take away the air, or the management, and you notice the absence real quick. Techniques have always had difficulty getting attention. You cannot see them. Objects, on the other hand, are easy to spotlight and promote in the media. Even genetic engineering can generate pictures of new proteins, or the latest crop of clones. But how do you get a picture of OR in action? Someone playing with a slide rule or workstation don't quite cut it. It's so bad that the most common representation of engineers at work is Dilbert. And no one has ever asked me to talk about OR on TV or radio.
OR The Ancient Art
Part of the problem is that OR has actually been around for thousands of years, it just wasn't codified until the 20th century. From Alexander the Great to Thomas Edison, we have records of great men applying OR methods to problems. They just didn't call it OR. But if you examine what they did, it was. Consider some of the examples. Alexander the Great, and his father Philip, had a firm grasp of politics, finance and mathematics. There was no magic involved in how Philip came out of nowhere and dominated the more powerful Greeks. His son was equally astute, coming up with one clever solution after another. Alexander was always in the company of mathematicians and scientists, something which Greece had an abundance of at the time. Military, scientific and political problems were all carefully thought out and solutions adroitly implemented. Take a close look at the period and you'll recognize a lot of OR at work. The Romans were equally adept, and much OR can be seen as they built the largest empire of the ancient period. Napoleon, who was educated as a mathematician, again used OR tools to innovate and accomplish his goals. And then there was Thomas Edison, the most prolific inventor ever. He gave a splendid example of military OR in action two decades before OR was recognized as a discipline. Asked by the US Navy to help in dealing with German subs and the threat to US shipping, Edison analyzed the situation and came up with the convoy system and much of what we now think of as convoy protection and anti-submarine warfare.
OR The Packaged Art
The reason OR had to be reinvented time and again was because no one in the past established OR as a distinct discipline. This was common in the ancient world, where many modern devices, like the steam engine, were invented, but there were no trained engineers to bring such devices to production and wide spread use, or to record and preserve the technology. Until the last two centuries, knowledge was acquired via apprenticeship, not formal education. But the achievements of those few who reinvented OR were always considered individual genius, not something you could package and reuse. Packaging knowledge is another 20th century movement that propelled everything else. While books have been around for thousands of years, and the modern university education was developed in the 19th century, the mass production of scientists and engineers is a 20th century innovation.
OR Roots
Roger Bacon invented, published and propagated the Scientific Method in the 13th century. This led to a steady and growing progress in science and engineering. OR was the next logical development in the systematic application of knowledge to problems. But we should take heed of the experience with the scientific method. This technique was used successfully by scientists, but also misused or ignored by opportunists, politicians and charlatans of all types.
OR's Few Practitioners
Because OR requires you to think clearly and methodically, there aren’t many practitioners. Despite strenuous efforts, only about five percent of US army officers are familiar with OR techniques (based on an estimate I did with Army OR instructors at Ft. Lee.) And the army has cut back on the training of officers in operations research techniques. The Scientific Method is rather simpler to explain and implement than OR. The Scientific Method was a relatively simple recipe for getting to the truth. It is taught in high school science courses. OR involves a lot more imagination and heavy duty math. Coming out of World War II, U.S. flag officers were true believers in OR. Soon there were several new agencies set up to do OR type work. Outfits like the Army's Operation Research Office (ORO), the Army-Air Corps' Operations Analysis Division (OAD) and the Navy's Operations Evaluation Group (OEG after ASWORG). In the beginning, OR was considered such a broad approach to problem solving that many different disciplines were accepted as part of the process. But eventually OR practitioners like Koopman, Leach, Morse, Kimball, Newman, Solant etc. came up with a generally accepted curricula. Initially, the non-military universities providing OR instruction taught the fundamental academic subjects and some applications. This approach was less pure OR and more systems engineering and business decision/quantitative sciences. Officers obtained OR training at the Naval Post Graduate School, the Air Force Institute of Technology and the civilian schools the Army used for such training (Georgia Tech, Colorado School of Mines, Florida Institute of Technology, and MIT). Other civilian schools producing OR practitioners were; Purdue, Texas Tech, UT Austin, Ohio State, and Texas A&M. By the 1980s, the Army had created MOS 49 (Operations Research/Systems Analyst) and established a school at the Logistics Management College at Ft Lee, VA. Graduates could go right to work, or transfer to a Masters degree program with 15 credits already taken care of by the Ft Lee course. There were between two and four classes a year, with about thirty officers per class. During the same period other military schools were running two classes a year with varying number of students. Naval Postgraduate School had 40 per class, Colorado School of Mines had ten, Georgia Tech had five and AFIT 25 students per class. Graduates were sent off to places where analytical skills were needed, especially in staffs and research operations. Results were forthcoming, as a lot of the smart moves made during the 1980s were done with OR operators guidance. There were enough OR practitioners that OR cells could be set up in many key organizations. This all changed when the Cold War ended. Staffs were cut and the smart guys were the first ones to move out for greener pastures. The OR operators knew they were in demand outside the military, and also realized that someone in a narrow specialty like OR was not going to make a lot of rank. Another "peace dividend" cutback was in training for military OR specialists. The service schools for teaching OR were shut down and officers sent to civilian OR courses. This was not the same, as the military used OR differently than civilian organizations. However, sending the students to the civilian schools made it easier for these officers to get jobs when they decided to bail out.
Robotic OR
But the very complexity of OR makes it possible to encapsulate OR as a distinct tool. This is what is happening. A black box on steroids. Going into the 21st century, we are beginning to mass produce robotic scientists and engineers. Increasingly, control devices use computers and OR techniques to run everything from automobile engines to stock portfolios. We think nothing of using powerful microprocessors and sensors to do, automatically, what once took a team of highly trained people to do. OR appliances or (ORAs) are an outgrowth of the development of expert systems and more powerful microprocessors. We already have ORAs in the form of powerful diagnostic systems on PCs and in automobiles and other vehicles. We tend to overlook the increasing amount of problem solving AI being used in machines and large systems. Diagnostic software, in particular, is making great strides.
OR and Wargames
OR has not served us as well as it could in wargaming and policy studies. Many lost opportunities have resulted. Take the Rawanda situation. The myth forming is that if troops had been sent in immediately, there would have been no genocide. In part, this is another case of amateurs studying tactics while professionals pay attention to logistics. But where are the OR studies of this situation? It's not a difficult one to do. Calculating the logistics is easy, working out the impact of peacekeeping troops on the killing is a bit more of a challenge.
More Process Than Problem Solving
Wargames are another area that could use more OR. As far as I can tell, OR shows up most often in professional wargame development as more process than problem solver. You can make a case that is how it should be, but my personal experience was that OR was the primary tool for developing a simulation of a historical conflict. OR techniques were used to solve the problem of how to develop a system that would generate reproducible results. Those of you who have played manual wargames long enough to absorb those games design techniques, and have designed your own, know what I mean. Wargame designers have abundant personal experience in making this work. I first encountered this while inadvertently predicting the outcome of the 1973 Arab-Israeli war, and more deliberately predicting the process by which the 1991 Gulf War was fought. A few years later, the lead designer on the Gulf War game found himself tasked, on very short notice, to create an accurate game of a conflict between Ecuador and Peru. His CINC gave him a medal for that effort, for the CINC considered COL Bay's overnight effort superior to what was being sent down from Washington. Even with the new wargaming MOS in the army, you still have tension between OR practitioners and wargamers. As OR types are heard saying, when you've tried everything else, try simulation. This destructive attitude was picked up in the civilian schools now teaching officers OR and is another example of how inadequate such schools are for training military OR practitioners.
Fear of Trying
Some subjects are difficult to even touch in professional wargames. And these are often issues that any straight-ahead OR analysis would encounter and deal with. But many OR operators shy away from the soft factors (morale, interpersonal relationships, fog of war and the like.) For example, one of my games (NATO Division Commander, or NDC) was adopted by the CIA as a training device in the early 1980s because it went after items CIA analysts felt were crucial, but most wargames, especially DoD wargames, avoided. Namely personnel issues among the senior leadership. NDC was part wargame, part role playing game and double blind as well. I don't think anyone ever did a game on how division staffs operate, but it was a worthy exercise. But it wasn't just the CIA that found wargames like this useful. I have continually heard from officers, with both peacetime and combat experience, who find that wargames give them an edge. The users don't have taboos about the simulation being near perfection. Like professional gamblers, the troops know that anything that puts their odds of success over 50 percent provides a tremendous advantage. Such an approach will be essential to handle things like Information Warfare (IW) or Revolution in Military Affairs (RMA). IW, for example, deals with shaping both friendly and enemy perceptions of what is happening. This is very difficult to model because there are many subjective and soft elements to be quantified. This makes it tough for traditional OR practitioners that try to deal with combat as hard science. Warfare is anything but and things like IW even less so. But historical game designers have dealt with this sort of thing successfully for decades. While traditional OR tends to focus on attrition, which is easier to model, but run these models by a military historian and they will provide numerous examples of where battles were won or lost not because of attrition but because of troop morale or one commander simply deciding he had been beaten.
BOAs
One solution to the problem of making OR more useful for the troops is the concept of "Battlefield OR Appliances. (BOAs)" The business and financial community already uses such beasts (less the "Battlefield" tag) for doing complex analysis in real time. Neural nets and genetic algorithms are attractive for the business "appliances." The idea is to create apps that think quickly and accurately, far more rapidly than any human practitioner. Program trading for financial markets is based on such concepts and, although few will admit it, these trading droids are often turned loose with little human supervision (mainly because there are situations where the action is so fast that slowing the droid down so humans to keep up would cripple the usefulness of the operation.) The Air Force has been talking about BOAs in the cockpit (pilot's assistant.) The zoomies are thinking about a BOA that would wrestle with things like compensating for battle damage, other equipment problems or EW situations while the pilot continued the battle. Of course, air combat is so complex that pilots could use a little coaching in things like the maneuvers most likely to bring success in a particular operation. Ground units could also use BOAs, especially in conjunction with digitalized maps. Setting up optimal defensive positions, patrol patterns or how best to conduct a tactical move. Sailors have similar needs (and one of the first OR successes was the development of optimal search patterns for ASW.)
Imagination versus Knowledge
We live in an age of unprecedented knowledge production. Part of OR is finding the right knowledge and applying it as a solution. One of the better Japanese work habits was their diligent collection of new knowledge. This is one reason why, for several decades, they have been one jump ahead of the "more imaginative" Americans in developing popular new products for the American market. In DoD, only the Marines consistently cast a wide net for new knowledge. When commercial wargaming became popular with the military in the 1970s, it was the marines that went after it most aggressively. When the marines recently showed up on Wall Street to study the workings of financial markets, they were really on to something. The manner in which these volatile, and quite huge, markets have moved from all manual to man-machine trading (program trading, etc) has direct application for the military. And the manner in which the man-machine concepts were implemented were classic OR exercises.
Future Trends
Items that can be expected to happen in the future, either because they are likely, or because we can only hope.
Wizards
Long a part, often an annoying one, of commercial software, these apps are constantly being beefed up to engage in more complex troubleshooting dialogs with users. There is potential here to obtain technology that can be used for battlefield OR appliances. The development work on Wizards draws heavily from decades of work on AI and Expert Systems. Much exciting OR work is going on in this area, and I believe there are already a number of military applications in development.
Troops Rolling Their Own
This has been going on since the early 1980s and the results are becoming more and more impressive. As the off-the shelf development tools become more powerful, more OR type military and wargaming apps will come from the field. These apps will co-opt the official wargames and sims in shops that want to get the job done rather than just perform the official drill. Some folks may not like this, but you won't be able to stop it.
Commercial Games
Much of the technology for these products has long been available off the shelf. Not a lot has been taken up by the DoD crowd, at least not in peacetime. Even slower to cross over have been the commercial development standards, which put wargames to realistic testing routines and quickly modify as needed. This does get done in wartime, as witness some of the rapid development that occurred during the Gulf War. The only military organizations making use of commercial gaming technology are those outside the DoD wargaming mainstream. This is largely a political (commercial stuff is "not invented here") and contractor (doing it from scratch makes for larger contracts) one.
Process Control and Program Trading Technology
Much of this is proprietary and you'd probably need an Act of Congress to extract a lot of this technology from the firms that developed it. However, much can be obtained from trade journals and a little (legal) competitive intelligence. Most of what is being done is no secret, it's the details of the execution that are closely held. And for good reason. In the financial markets, any edge is usually small and short lived. But this is what makes this technology so valuable. The manufacturing and financial markets are "at war" all the time. They thrive or go bankrupt based on how well their "weapons" perform. And much of the technology is transferable to military uses. Many of the components of commercial apps are available as off the shelf toolboxes or widely know concepts. One of the more useful of these is known as "fuzzy logic." This item addresses many of the problems DoD wargamers have with dealing with soft factors. Civilian practitioners face the same problems and they have come up with many working solutions using fuzzy logic. In my experience, nothing is fuzzier than modeling combat.
What Percent Solution?
There may eventually be more acceptance of OR solutions that are sufficient rather than perfect. When modeling weapons or equipment performance, you can get a 100% solution (or close to it.) Many OR practitioners are more comfortable with this than those elements that involve more people and less machinery. It is currently impossible to get a 100% (or 90%, or often even 50%) solution for things like how an infantryman performs in certain situations. In peacetime, there is tendency to gold plate things. Errors of any sort are threatening to careers. In wartime, you can make mistakes. Everyone else is and the honors go to those makes the fewest. But the peacetime zero defects attitude hampers innovation and performance. You get a lot of stuff that is perfect, but doesn't work. The latest example of this was seen in the Gulf War. When a more powerful bunker buster bomb was needed, the weapon was designed, developed and delivered in less than two months and used at the end of the war. The air force also improvised a mission planning simulation in record time (using everything from spreadsheets to existing models) and used it. The army CENTCOM wargames shop also improvised (although less successfully, but this was only discovered after the war, a common occurrence in wartime.) We should also remember that during World War II, OR practitioners recognized that their calculations could not cover all critical factors. They had to work with fuzzy situations and before "fuzzy logic" became a recognized tool, the World War II operators managed to work with the problem and not just walk away muttering that "it can't be done." The current rigidity in can be traced to the relative lack of operational experience. And when operational experience does become available, it is often the case that battlefield calculations that ignored those pesky soft factors were way off. A good example of this was the analysis TRADOC did of NTC engagements. They found that the ammunition expenditure data for NTC was much different than the OR predictions, and closer to the expenditures in earlier wars. You cannot ignore dealing with the soft factors, for eventually they will bite you in the ass.
Hired Guns
The end of the Cold War coincided with a growing demand for OR skills in the civilian sector. So many of the uniformed OR people left, and their numbers are still dwindling. Where there used to be OR groups in HQs and schools, you don't see much of this anymore. This is bad. Part of the problem is that military OR people can make more money, and have fewer unaccompanied overseas tours, as a civilian. But the numbers of uniformed OR operators is shrinking so much that the military is having a hard time properly supervising the civilian hired guns. It is also important that military OR operators be warriors (or military practitioners) first. Otherwise, you often encounter the syndrome of "if the only tool you have is a hammer, all problems look like nails." The same syndrome is noted in the civilian sector, and the solution is often to take a banker, plant manager, structural engineer or whatever and train them in OR techniques (or programming) and then turn them lose on the problems.
Putting Operations Back into OR
Originally, OR operators researched operations first hand and then devised solutions. There has been a trend away from this and towards an emphasis on technique: linear programming, dynamic programming, queuing theory, chaos theory, neural networks and so on. Knowing these techniques is a good thing, and can even be useful if you collect valid operational data. But it is not real operations research. If you can't get an experienced infantry officer as an OR analyst, then maybe you should consider sending your analysts through boot camp. Most fighter pilots have technical degrees and can pick up OR techniques quickly. Using OR trained pilots to work on combat aviation problems is an enormous advantage, for the operator with practical experience will catch things that the researcher trained only in OR will miss. Happens all the time, and the military is noticing the lack of military experience among the OR practitioners now working on military projects.
Putting the R back into OR
The "R" -research- has been largely replaced by "A"; analysis. The Navy now talks more of "OA" (Operational Analysis) than "OR" with "OA" often being considered an adjunct to Modeling and Simulation. The result has been an emphasis on quantification and metrics at the expense of understanding the problem. Like so many other debilitating trends, this one developed largely in response to what "decision-makers" have demanded. What we often have now is "advocacy analysis," where much time and effort is spent to provide justification of a position or decision based on having more and "better" numbers and metrics than your critics. This often occurs by focusing on a very narrow slice through a problem that is often far removed from the true context of the overarching problem.
Crunching Numbers versus Getting Results
There has long been a split among OR practitioners, especially in peacetime, over how best to achieve results. On one side you have the "physics" crowd, who insist on reducing every element of combat to unequivocal data and algorithms. On the other side you have the "whatever works" crew. The "physics" bunch are basically engaged in CYA (Cover Your Ass) operations, because there are many soft factors in combat that are not reducible mathematically the same way weapons effects can be.
Better OR Tools
You can't have too much of this. I'll never forget the first time I did a regression analysis. I did it manually. Try and get students to do that today and you'll get arrested for child abuse. By the 1980s we had spreadsheet plug-ins for Monte Carlo, Linear Programming and so on. I thought I'd died and gone to heaven when I first got to use that stuff. Then came MathCad, SQL on analytic steroids and more. Yes, we want more. We need more. We deserve more. If we can't get any respect, at least we can get more neat tools. Warning; too much of this stuff appears to contribute to overemphasis on analysis at the expense of getting something useful done.
Put OR Back in Uniform
A combination of vastly increased demand for civilian quants in the 1990s, reduced promotion opportunities after the Cold War, and the usual problems with having a non-mainstream MOS saw a steady decline in the number of uniformed OR operators. Norm Schwartzkopf was one of these, but he's retired, as are many other OR qualified officers. Either that or out making a lot more money as a civilian quant. I don't know how you're going to get quants back in uniform. It will take a decision at the top. In times of crises and resource shortages, a lot of really important things get shortchanged because they are difficult to understand.
Please note: This article is revised and expanded from notes used for a talk at a November, 2000 INFORMS meeting I'll be gradually editing this into a more useful format. One of the people at the talk was Mike Garrambone, a long time wargamer and former Major of Engineers (and instructor in wargaming and the AFIT.) Mike has been working on a history of OR and we are going to work together to integrate that into this document.
About the Author
If you have any comments or observations, you can contact Jim Dunnigan by e-mail at jfdunnigan@aol.com. Jim is an author (over 20 books), wargame designer (over 100 designed and publisher of over 500), defense advisor (since the 1970s), pundit (since the 1970s) and "general troublemaker". Dunnigan graduated from Columbia University in 1970. He has been involved in developing wargames since 1966. His first game with Avalon Hill (now a part of Hasbro), "Jutland" came out in 1967. He subsequently developed another classic game, "1914", which came out in 1968. A year later he began his own game publishing company (Simulations Publications, Inc, or SPI). In 1979 he wrote a book on wargames (The Complete Wargames Handbook). In 1980 he began a book on warfare (How to Make War). In 1982 he accepted an invitation from Georgia Tech "to come down and lecture at the annual course they gave on wargaming. Been doing that ever since."
In 1985, he was asked to develop a tactical combat model to see how robotic mines would work. In 1989, he got involved editing a military history magazine (Strategy & Tactics) "which was the one I ran while at SPI." In 1989, he got involved in developing online games, and that continues. Jim edits strategypage.com.
Citation
Dunnigan, James F., "The Operations Research Revolution Rolls On, To Where?", DSSResources.COM, 05/28/2004.
James F. Dunnigan provided permission to archive this article and feature it at DSSResources.COM on Monday, December 7, 2003. This article was posted at DSSResources.COM on May 28, 2004.