These developments led the above-mentioned Andy Marshall, as the high priest of wargaming at the Pentagon, to call for games of a different kind. They had to be capable of comparing various force structures, allowing a much richer set of operational factors to be included. Not only strategic nuclear warfare but the much more complex type waged by conventional forces, which might result in escalation, had to be simulated. Another important objective was to find ways of incorporating Soviet points of view on these matters, which Marshall, following discussions he had had with the Soviets years before, thought were likely to differ considerably from American ones.22 For example, the Soviets might not attribute the same importance to warhead accuracy, silo hardness, and the like. Their methodology for assessing effectiveness, and consequently their plans for achieving it, might also differ considerably. Even if the doctrine was clear, it might not necessarily be understood or adhered to at all command levels. In other words, problems of command, control, and communications had to be brought in as well. The ultimate objective, Marshall wrote, was to design computerized games that would meet the needs of three distinct groups of people. They were, first, those interested in assessing military balances; second, those involved in attempts to change those balances by evaluating the merits of alternative force programs; and, third, those involved in devising and evaluating operational plans.23
As usual, the principal contractor was RAND which had been working on this kind of problem for over a quarter-century and where Marshall himself had worked for a time. The objective was to automate “all features of political-military gaming from the force calculations to the decision making participants in the traditional wargame,” no less. Specific problems to be explored included what might happen in case the Soviets attacked Washington on Inauguration Day, what the effect of the pre-delegation of authority might be, what might happen in case early warning satellites were lost, and the like.24 Speaking of games, one can imagine the fun the RAND personnel had dreaming up these and other scenarios and trying to find ways of making their computers simulate them as closely as possible. A preliminary report came out in 1982 even as Marshall made the project’s existence public.25 The analysts in charge duly noted how demanding the research was, given that it forced them to make every assumption explicit, find a way to assign to it a numerical value, and create an algorithm to link it to the rest so that the computer could “understand” it. All this implied a major effort to locate the necessary information and using one’s imagination where, particularly on the Soviet side, it was not available. Next, no fewer than four hundred different situations had to be created in painstaking detail and decision-making trees for each one devised to enable the computer to “crunch” its way through them. Even that was only the beginning, since care had to be taken to make the program expandable so that even more situations could be incorporated later on.
The analysts also noted that, under this method of gaming, the scenario, instead of being presented before the beginning of play as in traditional wargames, emerged from the interaction of the computer programs themselves. Last but not least, even though the “primitive automated agents are not presently programmed to be very sophisticated, we have been impressed with the complexity of play that evolves from the interactions of several simple automatons.” In other words, simply by interacting with each other, programs, though they could not of course violate the assumptions (rules) that had been written into them, often produced highly unpredictable results. “Even at this rudimentary stage of development,” the analysts added, “some of us have found that the horizons of our own strategic thinking are expanding very rapidly.” As so often, perhaps the best way to benefit from a wargame is to design one.
Four years later a more detailed report was published.26 By this time the scope of the project had been widened so that it comprised not just “Sam versus Ivan” games but some third countries known as Greens as well. The report specifically mentions Saudi Arabia, Poland, Belgium, and France, though it is not clear which and how many of these were included in any specific game. As before, for each country the computer had to be fed not just with data concerning its own forces and behavior but also with the image it had formed, or was supposed to have formed, of the rest. In the case of the superpowers, those images played a major role in the decision to escalate or not to escalate. In the case of the Greens, the images might affect − perhaps determine − their decision as to which side to join, if any. As the number of countries involved in any particular game increased, the resulting level of complexity went up exponentially.
The report also referred to a number of sophisticated features that its predecessor, whether because of its summary nature or because it only dealt with preliminary work, did not mention. One was artificial intelligence, meaning the ability, for example, of Ivan to compare his doctrine in respect, say, of rates of advance or the consumption of fuel and ammunition to the “actual” results that his interaction with Blue produced. The doctrine having been duly adjusted, the effect of the changes might make themselves felt as the game continued. To that extent, it is possible to speak of true wargaming between thinking opponents. An attempt was also made to enable the programs to look ahead towards certain predetermined outcomes and adjust their “play” accordingly. Superficially such a capability was needed in order to simulate the way human decision-makers think and act. However, it quickly led to the problem of what mathematicians call infinite recursion. Infinite recursion is what happens when trying to guess the intentions of an opponent in a strategic-type encounter: I think that he thinks that I think that he thinks that I think . . . Since the series has no logical end, in theory it may last forever. As the analysts noted, under such circumstances the ability to distinguish Sam from Ivan, and the other way around, is quickly lost. That, incidentally, is as true in the real world, especially that of intelligence and espionage, as it is in games, computerized or otherwise.27
Interestingly enough, looking forward so and so many moves – currently not even the most advanced supercomputers can work through all the moves in a game, and they probably never will – determining the most favorable interim position, and continuing from there is how modern chess-playing programs work.28 To be sure, even in a game as circumscribed and as structured as chess, short of meeting the victory condition – mate to the opposing king – defining the exact meaning of a desirable outcome is very hard. Indeed it is precisely in this respect that the differences between various programs are most pronounced.29 As the RAND analysts admitted, they did not know what outcomes decision-makers might be looking for, much less how they might evaluate those outcomes if and when they did in fact take place. Not to mention the very likely possibility that, in the “real” world, the decision-makers might not know either but would feel their way forward step by step as sleepwalkers are supposed to do. Hence it was all the analysts could do to use their imaginations instead.
Summing up what appears to have been some seven years of work by experts from fields as different as artificial intelligence, military science, psychology, and the history and culture of the countries being simulated, they said that “currently the prototype NCL (National Command Level) models do not contain the richness of detail that is possible and essential for applications work.” Nevertheless they considered “prototype development to have been very successful.” They also made specific recommendations for further research, suggesting, for example, that the laws of war should be brought in and that greater attention should be paid to problems of command and control as well as “asymmetries” between the ways Sam and Ivan understood the world and each other. The last-named problem, of course was exactly the one they had been asked to solve in the first place. Neither the 1982 report nor the 1986 one wasted a single word on the most important question, namely whether the project produced anything of any use to any of the three groups Marshall had mentioned.
To return from these Pentagon-inspired attempts to build a computerized model capable of simulating the way
the world works to the more limited issue of gaming strategic nuclear warfare, it would appear that this kind of gaming never became very popular. Almost certainly one reason for this is that such warfare does not make any great intellectual demands on either strategists or gamesters. To be sure, there may be a first nuclear strike and even a second one. However, a third one is almost inconceivable, as is proved by the fact that practically nothing has been written on it. The rarely used term “broken back warfare,” meaning warfare that involves whatever forces may survive a first and second strike, merely confirms this point. A good analogy is presented by some high-level tennis games in which one player serves one ace after another, stunning his opponent and leaving him helpless. In such a game the give and take that both represents the essence of strategy and accounts for much of men’s fascination with it is almost entirely absent. This was especially true after ballistic missiles largely took the place of manned bombers from about 1965 on. Even today, though a few anti-ballistic missile defense systems have been deployed by various countries and are supposedly operational, the situation remains substantially the same.
Throughout the ages, war has always been primarily a question of character attributes such as steadfastness, determination, stamina, and, above all, the readiness to kill and be killed if necessary – to commit, and cope with, bloody slaughter. Many wargames reflected this fact: there are even a few cases, such as the late nineteenth-century duels fought with rifled pistols in which the parties are expected to maintain a stony mien while doing nothing to defend themselves, when courage in particular is almost the only relevant quality. By contrast, in nuclear warfare these and other human qualities count for little if anything. At the top, everything is done to ensure that decision-makers and operatives at all levels should act on the basis of rational considerations alone. At the bottom, such is the scale of destruction that there is no coping with it.
Both those who made the decision and many of the rest of us are very likely to end up dead, leaving behind neither leaders nor anybody who may obey them and admire their leadership. In the words of French President Charles de Gaulle, “[in the wake of a nuclear war] the two sides will have neither powers, nor laws, nor cities, nor cultures, nor cradles, nor tombs.”30 The heroic narrative, which, beginning at the time when Homer praised andreia (manly courage) as the highest virtue, has always provided much of the motivation behind war and, perhaps even more so, many kinds of wargames, is unable to develop. As one well-known Cold War-vintage Soviet joke put the matter, the citizens of Moscow, when coming under nuclear attack are called upon to don shrouds and crawl slowly to the cemetery: slowly, in order to avoid a panic.
Above all, even the most conservative estimates, whether produced by wargames or by other methods, put the number of US casualties resulting from just ten 0.5 megaton weapons over ten urban areas at just under 20 million.31 This calculation ignored the fact that the real number of available warheads ran into the thousands. Furthermore, it included neither the millions who would die later as a result of radiation-induced cancer nor those who would succumb to the political, economic, social, and psychological conditions in an industrial society abruptly sent back into the Stone Age.32 As if to highlight the madness, one article on the subject carries the title, “Victory is Possible.”33 Taking a saner approach, most people spoke of “the end of civilization as we know it.” To quote a character in the 1983 movie WarGames, the only way to win the game was not to play it.34 In the movie’s opening scene the Department of Defense holds an exercise involving the simulated launch of ICBMs with their nuclear warheads, only to discover that one-third of the personnel refuse to turn their keys as told. The movie does not say whether this was because they could not believe that the orders were for real or because, knowing them to be so, they shrank from the task. Coming to the rescue, the BYMs replace the operators with computers, automating the system and opening a nest of nuclear troubles that need not concern us here. Among the consultants who helped create the movie was Peter Schwartz, an engineer/scenario writer/futurist who, as it happened, later worked for Marshall too.
The secrecy that surrounds the issue notwithstanding, there is some evidence that the fear that some humans in the loop might refuse to play ball presented a real concern. Several sources suggest that, in 1980s wargames, players willing to cross the nuclear threshold were hard to find. Explanations for this phenomenon differ. The players may have preferred the devil they knew (conventional war) to the one they did not know; or else they simply did not want to blow up the world. Be that as it may, in game after game they explored every alternative, desperately signaling, dodging, and negotiating with their opponents. This might reach the point where Control found it necessary to push them in that direction even at the cost of producing increasingly implausible scenarios. Should they fail to respond, it might even accuse them of “ruining” the exercise.35 I personally have heard a slightly inebriated British officer explain how, when the alarm in a nuclear storage installation where he used to work was sounded, those present could not at first make themselves believe it was real and that an accident, involving a bomb that had fallen off its trolley, had taken place.
Marshall himself must have shared those concerns, or else it is hard to see why one of the problems that the RAND analysts he commissioned set out to “solve” for him was the frequent reluctance of human players to unleash such a conflict. They even argued, as Kahn (who happened to die in the same year in which the movie came out) might have done, that, “if gaming . . . the initiation of a nuclear conflict holds any analytical value, it might only [my emphasis] be possible with a programmed agent.”36 To that extent, the oft-noted tendency to look at nuclear wargamers as a bunch of Dr. Strangeloves, blame them for some of government’s more outlandish policies, and accuse them of making the unthinkable a little less so appears to be fully justified.37
Finally, this argument may also be turned upside down. The reluctance of wargamers responsible for simulating the behavior of senior decision-makers in various countries to cross the nuclear threshold may be a cause for concern to a handful of experts who want to know, as closely as possible, at what point a nuclear war might break out, how to prepare for it, what it might look like, how best to wage it, and how to end it short of blowing up the world. Outlandish and disturbing though those concerns may seem to many people, it is hard to question their importance. Yet there is some reason to believe that the reluctance to escalate may faithfully represent the games’ unexpected success in capturing the way the real world behaves.38 If so, then the news is encouraging indeed.
Onscreen war
As the Cold War started winding down during the late 1980s the probability of a Soviet “strike out of the blue,” never very likely, declined. So, as a result, did gaming “strategic” nuclear exchanges. On the one hand, as the 1991 Gulf War was soon to demonstrate, the US advantage in waging conventional warfare was so overwhelming as to make resort to nuclear weapons unnecessary. On the other, such was the fear of nuclear weapons that it was, and remains, difficult to imagine a war directed against an enemy who possesses, or is strongly suspected of possessing, even a single such weapon. Indeed some countries, notably India and China, seem to have based their entire strategy on that assumption.39 As far back as 1965, speaking to the French writer André Malraux, Mao Zedong himself said as much.40 Increasingly, nuclear wargames at the Pentagon tended to deal with conflicts between third parties such as India and Pakistan or else in the Middle East. Nuclear warfare in Korea must also have been gamed countless times. In an ironic reversal of previous fears that when it came to unleashing a nuclear war human operators might not cooperate as they should, some games focused on the problems of unauthorized war and accidental war.41
Having grown from modest beginnings in the 1950s, forty years later wargaming, whether carried out in the think tanks or in-house by various headquarters with the Joint Chiefs of Staff at their head, had acquired a large following. It consisted of highly qualified and ex
tremely specialized experts who, as well as trying to reduce everything to numbers, programmed the computers, and understood the highly esoteric language in which many reports were written (e.g. “FSS runs on the Synchronous Parallel Environment for Emulation and Discrete Event Simulation (SPEEDES) framework [which] helps exploit available high performance computational resources and provides much needed functionality”).42 The cost must have run into the hundreds of millions of dollars each year. No wonder the community in question sometimes seemed to take on a life of its own, dreaming up new kinds of games each time changing circumstances made the old ones appear less relevant.
Wargames Page 39