The decade following Dulles’s massive retaliation speech has been described as a ‘golden age’ of strategic studies.1 It saw the growing importance of civilians in framing the great policy debates of the nuclear age and then in influencing their outcome, whether as advisers from the think tanks or officials in government. The annual conference of the Institute of Strategic Studies, established in London in 1958 and later rebranded as ‘International’, provided a place where those engaged in this new and controversial field of study, including non-Americans, could meet and discuss new ideas. Because this was such an intellectually fertile time it is possible to mythologise this period, assuming a greater homogeneity of approach and clarity of purpose than was the case, for example by exaggerating the influence of new techniques such as game theory or the more celebrated theorists such as Tom Schelling or Herman Kahn to explain policy.
The most important feature of this new group was that it was largely civilian and was confident enough to challenge military perspectives. The armed services found that they could no longer set the terms of the debate on how best to prepare for war. With the missile age destruction was becoming automated and irresistible and the links with traditional military practices tenuous. The bomber pilot, weaving through air defences, locating his target and releasing his bombs, was to be replaced by a technician, situated in some underground silo, protected by concrete and ready to launch his pre-targeted missile only on the receipt of a signal from a distant commander. The human element was being concentrated into the decision to employ nuclear weapons and the unambiguous communication of the relevant commands. Technical skill and ingenuity were devoted to the design and production of offensive weapons, reducing the opportunities for enemy defences, but in the process also reducing the demands on professional military talents. The implementation of an order to attack would begin with the President and thereafter require a strict adherence to orders. Loyalty and discipline were in fact the only traditional military virtues relevant to the missile age; heroism, leadership, quick wits and other combat virtues had little place in this sort of warfare. Soldiers were not expected to take the initiative. Every move had to wait authorization from political masters. The last vestige of a contest with the armed forces of the other side, rather than the destruction of its society, would be lost. When remote-controlled missiles dominated the force structure of both sides, the human element of war would be lost altogether. War itself would become less relevant when it was geared to an orgy of destruction so that the strategic objective had to involve deterrence.
The military resisted taking such a passive role and operational nuclear planning assumed considerable continuity with past practice. Yet they now had to cope with new types of experts intruding into their professional domains. The problems of national defence were those of the management of technical innovation, large-scale engineering projects and far-flung organizations, and of the formulation of doctrine for the employment of means of unprecedented destruction, intended to persuade potential adversaries not to test its credibility.
departed his role as scientific adviser to exercise highly persuasive influence in matters in which his convictions were not necessarily a reflection of technical judgment, and also not necessarily related to the protection of the strongest offensive military interests of the country.4
To Oppenheimer’s many supporters a failure to comment on the moral and political implications of their scientific work was an abdication of responsibility, for they could point to possibilities that non-scientists could not be expected to see. In addition the move against Oppenheimer was seen to be part of a general drive to dismiss scientific commentary in areas where it could be embarrassing. In particular the readiness of the AEC to deny the harmful effects of radioactive fallout from nuclear tests, including possible genetic mutations, generated a major campaign by scientists to bring this issue to the public notice.5
A July 1955 manifesto put out in the name of the philosopher Bertrand Russell and Albert Einstein urged scientists to come together to ‘appraise the perils that have arisen as a result of the development of weapons of mass destruction’ and find a way to avoid world war. The first conference on Science and World Affairs took place in Pugwash, Nova Scotia in July 1957, with scientists from 22 countries, including the Soviet Union.6 Pugwash was one of the few organisations to bring representatives from East and West together.
Objections to government policy encouraged both groups to put their critiques into the public domain. A variation on the scientific approach that was less accessible to the attentive public sought to develop an original analytical method appropriate to the special demands of the nuclear age.
How this might be done had been demonstrated during the Second World War as opportunities arose for scientists to advise on future operations, based on intensive study of the successes and failures of those of the past. The practical nature of the work was captured by its description: ‘Operational Research’. One of its founders, Patrick Blackett, differentiated it from other areas where the scientific method was being used to study social phenomena because of ‘the direct relation of the work to executive action’.7 It began in Britain with studies for the RAF on how to make best use of radar. Some of its early success can be seen in anti-submarine warfare. Other studies led to recommendations on the most appropriate size for convoys and on the most effective targets for air raids on Germany.
A number of features of this early work need to be emphasized in order to contrast it with post-war military applications of the scientific method. The approach was inductive drawing on accumulated data of comparable cases which could be analysed systematically often by drawing on mathematical techniques. This could be translated into tactical advice for future combat. It was therefore a complement to traditional forms of military planning and not a substitute for it. The operations most likely to benefit were those in which there were large numbers of comparable engagements on which to base analysis and in which sufficient factors were constant to prevent the analysis being overwhelmed by too many variables. Sir Solly Zuckerman noted that field warfare presented problems. It moved too fast for any results to be applied and because ‘the situations in field warfare were infinitely more varied than in either air or sea war’.8 John von Neumann, a polymathic Hungarian émigré to the United States, worked with a British operations research team during the war and became an enthusiastic promoter of this type of analysis in the US after 1945.9
Neumann was excited by the possibility of a new science of warfare. Once into the nuclear age comparable research had to address itself to possible operations that were completely without precedent. The relevant weapons had often not been built when the studies were done and would never be tested in combat. If war came there would be no opportunity for learning through trial and error. There would be one trial in which no margin of error would be tolerable. Exchanges of nuclear firepower defied human imagination and often reason. What was known was largely about the effect of a wide variety of types of nuclear explosions based on the aftermath of the detonations over Hiroshima and Nagasaki, subsequent nuclear tests,10 and estimates about the properties of the alternative means of delivery and means of defence. With the improvement of methods for intelligence collection increasing amounts of information were deduced about the capabilities of the potential adversary.
The RAND Corporation, located at Santa Monica, was the proto-typical ‘think-tank’. This was set up under an Air Force grant to develop operational research. It soon became an independent non-profit corporation addressing defense issues and other aspects of public policy using advanced analytical techniques. RAND began by recruiting natural scientists and engineers expecting to deal with hardware. It also hired economists and other social scientists.11 Because Yale decided to end funding for its Institute for International Studies in 1950, on the grounds that policy research contradicted the University’s liberal arts tradition, RAND could recruit individuals such as Brodie, then the leading civilian thinker on nuclear strategy, and William Kaufmann.12
The steady improvements in computational power made mathematical approaches to complex problems more practical. Even economics up to this point had been more literate than numerate. Now quantitative analyses grew in strength and credibility. RAND analysts saw them as supplanting rather than supplementing traditional patterns of thought. Simple forms of cause and effect could be left behind as it became possible to explore the character of dynamic systems, with the constantly changing interaction between component parts. The models of systems, more or less orderly and stable, that had started to become fashionable before the war could take on new meanings. And even in areas where intense computation was not required there was a growing comfort in scientific circles, both natural and social, with models that were formal and abstract, not just based on direct observations of a narrow segment of accessible reality but on explorations of something that approximated to a much larger and otherwise inaccessible reality. Types of systems and relationships could be analyzed in ways in which the human mind, left on its own, could not begin to manage.
There was therefore still a certain amount of material with which to continue the operational research tradition. Analysts attached to the armed forces worked on practical questions, such as the optimum altitude of bomb release or the best way to penetrate air defences. Soon the range of issues widened to provide guidance on crucial policy decisions concerning the design, procurement, basing, control and operation of new weapons. RAND attracted a degree of awe and respect that they did little to discourage. This was a special place offering optimal conditions for innovation, It was well-resourced, with access to computers, and unconstrained by the normal routines of organisational or academic life, combining intensity with informality. The fascination with this group meant that the personal rivalries and intellectual achievements of those pioneering years have been well recorded.13
The term ‘think tank’ had been used colloquially to refer to the brain since the nineteenth century and then adopted for the secure rooms where military planners could discuss coming operations. The first actual use of the term in connection with civilian research institutions (at least in the New York Times) was with reference to the Center for Advanced Study in the Behavioral Sciences (CASBS) at Palo Alto, California in 1958, but it got applied most frequently to RAND.14 The stress on the cleverness of those involved and the novelty of their methods gave their work a certain aura. As it was published in books and journals, the outside world was provided with a glimpse of the relevant analytical techniques and, on a non-classified basis, samples of the fruits of their application. No comparable organizations emerged in Europe. The necessary information and the devices for processing this information remained scarce and monopolized by governments.
The approach was reinforced by the development of computers capable of supporting advanced statistical techniques, hitherto either impossible or incredibly time-consuming. The lone scholar in the library began to be displaced by teams of researchers working with vast amounts of data. Such projects were extremely expensive and could not be pursued without substantial funding. By the early 1960s the US government was spending approximately $10 billion annually on Research and Development, of which half came from the Pentagon and the Atomic Energy Commission. Of this about ten percent went to universities or university-research centres.15 Within the military itself, there were efforts to strengthen ‘in-house’ research, albeit often drawing on outside expertise. The Weapons Systems Evaluation Group in the Pentagon, set up in 1948, was charged in 1954 with providing ‘comprehensive, objective, and independent evaluations under projected conditions of war,’ to include present and future weapons systems, their influence on strategy, organization, and tactics, and their comparative effectiveness and costs. In this role it worked closely with leading universities.16 The Office of Naval Research (ONR) was an especially generous funder, with many projects supported without any obvious connection to maritime matters.
The output of the operational researchers became influential. Academic strategists, such as Henry Kissinger, were exposed to it during visits to the RAND Corporation or at conferences and seminars where the two groups mixed. Nevertheless, much of this core technical work was either classified or highly technical. Albert Wohlstetter described such analysts as ‘the largely unsung heroes of strategy in the nuclear age’.17 He argued that the credit for developing the first-strike/second-strike distinction and recognizing the difficulties of maintaining stable deterrence should go to those working in this empirical tradition, and not to the strategic writings of the natural and social scientists. These, he argued, were largely unfamiliar to the RAND teams working on the basing studies in the early 1950s.18 This encouraged the view that nuclear strategy was an unusually complicated business that only special minds manipulating special techniques could manage.19 Wohlstetter was suggesting that this could require subtle judgements, depending on the technical analysis, taking in such variables as the range, yield and accuracy of weapons and the hardness and mobility of targets, along with issues of warning and sequencing.
The sort of work undertaken at RAND was described as ‘systems analysis’. Some sort of systems approach appeared in attempts to deal with all manner of questions from the trivial to the fundamental. The word ‘system’ came to be a vogue adjective, indicating a capacity to view problems as a whole, rather than in fragmented components, and to master complexity. The term was adopted in a variety of distinct occupations, each with their own literature and specialist jargon. It could be used in connection with complicated engineering projects, so that it became necessary to talk of ‘weapons systems’ rather than a mere weapon (to belie the notion that a single item of military equipment could perform any useful task by itself). Meanwhile, systems theory described an academic aspiration for an all-embracing conceptual framework.
[T]he explicit outline and study of alternative systems of inter-dependent parts where the comparative performance of a system is affected not only by the machines and the men who are elements in the system but also by the opposing behaviour of men and machines outside.20
In this manner it was possible to make the best of the available information and even to work out how to generate some more, but it was not possible to compensate fully for the lack of hard information. The limits of this method were determined by the number and range of the variables and the extent of what could not be known. One result was to encourage a tendency, which gradually became more acute, to place an extremely sophisticated technical analysis within a crude political framework. We noted earlier Wohlstetter’s assumption that the massive losses suffered by the Soviet Union during the Second World War would inure them to equivalent losses as a result of nuclear strikes. It was as likely to lead them to do everything possible to avoid a comparable catastrophe.
Wohlstetter spoke of a ‘conflict-system design’, although he acknowledged that use of this terminology suggested that ‘the goal of study is to generate conflict’. One alternative was ‘opposed system’, for opposition was an essential feature.21 The focus was on the relationship between the capabilities of the potential belligerents and the physical consequences of their hostile interaction, varying assumptions within parameters that could be empirically determined. Where the analysis became far less effective and satisfactory was when considering the ‘conflict system’ as a relationship between men and political structures, rather than machines. Here most of the ‘evidence’ was soft—drawn from inadequate historical parallels, from limited experience of crises and from untutored intuition.
Little could be known about the likely responses of human beings to any of the situations that they might find themselves in during a nuclear war—either in deciding whether to launch a nuclear attack, or in implementing this decision, or in anticipating and suffering the results. With no precedents how could a nuclear war follow recognised conventions and ground rules? Many of the intangible social and political factors that would shape choices lay outside the scope of any manageable ‘conflict system’. Or else temporary phenomena of a sort that could not be anticipated in models would shape events. Some ‘scenarios’ of East-West crises, even if no more than doubtful hypotheses, could if used regularly to provide a context for war-gaming become so familiar that they turned into plausible contingencies. For example, the ‘Hamburg Grab’ had great value in analysing how one might cope with a limited, probing action against a significant piece of territory, but it was chosen because of Hamburg’s proximity to the Eastern border rather than its intrinsic value to the Soviet Union. The interests that might lead one nation to engage in provocative adventures with a high risk of disaster became less relevant in the analyses than the question of how it might succeed. It was as if police officers were being taught the art of homicide detection in terms of opportunity and murder weapon, but never motive.
[I]t is based upon assumptions about human behaviour which seem totally unreal. It neither constitutes scientific analysis nor scientific theorising, but is a non-science of untestable speculations about Western and Soviet bloc behaviour.23
Brodie who had argued in 1945 the need for more rigorous study asking whether ‘a science like economics’ could show the way to a ‘genuine analytical method’24 came to question whether this sort of reasoning was so helpful. The practitioners showed little understanding of the history or politics of the situations being addressed. They showed disdain for other social sciences as ‘primitive in their techniques and intellectually unworthy’.25
Whereas the inclination of the historians and political scientists was to draw on the past as well as what was known about the new technologies to identify the urgent decisions that policy-makers needed to make, this approach presumed that the emerging situation was wholly novel, forgot the past, returned to first principles and attempted to build up a new theory appropriate to the new age.26 The interest here was in a pure theory of strategy with premises, logic, and conclusions relevant to East-West relations but not derived from them. Those who developed their judgments and prescriptions through this sort of disciplined, formalized reasoning can be described as the ‘formal strategists’. The exemplary formal strategist was Thomas Schelling, with Morton Kaplan, Glenn Snyder, Daniel Ellsberg, Malcolm Hoag and Oskar Morgenstern also having a right to inclusion. The approach assumed an underlying rationality informing human behaviour. If behaviour was random then a true science would be impossible. The exemplary methodology for the formal strategists was provided by game theory.
Although the new approach was castigated for assuming rationality in key actors part of its inspiration reflected dismay about past irrationality, including the rise of totalitarianism, as well as alarm about the implications of future irrationality. Without confidence in individuals as naturally rational beings the best hope seemed to be to steer them towards more rational decisions. According to Heyck, this explained why the new approach assigned responsibility to systems instead of individuals, considered choices rather than choosers, and framed ‘the old question of reason in a new language, a language of choices, decisions, revealed preferences, games, moves, payoff functions, subjective expected utilities, uncertainties, information flows, information costs, strategies, heuristics, programs, and the structures of cognitive processes’.27 The objective was to understand how decision-making might be structured to get optimum outcomes. The Cold War gave urgency to this process—it was a ‘bad time for bad decisions’.
The Cold War offered the sort of situation to which the approach seemed particularly suited. Just as it pushed geopolitical logic to a bipolar extreme, with two superpowers leading a substantial bloc of states, representing the opposing ideologies of state socialism and liberal capitalism, so did the logic of the nuclear age warn of the futility of attempts to resolve this deep conflict through war in the expectation of a decisive victory. Conundrums were posed for which there were no obvious military answers. Unusually the situation was so stark that it could be illuminated using abstract theory. Gaming the complex geopolitics of 1914 would have been too much of a challenge, but a conflict dominated by two ideologically-opposed alliances, each led by a superpower and armed with the most destructive weapons could be portrayed in a 2×2 matrix.
Game Theory was invented by John von Neumann in the 1920s and developed in a joint work with Oskar Morgenstern in 1944.28 Inspired by observations on poker, it demonstrated how to proceed rationally in an irrational situation, bluffing for both offensive and defensive purposes, and how the occasional random move could make it difficult for an opponent to discern a pattern of play, thereby adding to the uncertainty.29 At first, under von Neumann’s influence, it appeared as a branch of mathematics special to Operations Research. He saw RAND as an institution that could showcase the new techniques and that is where it began to be picked up by social scientists.30
This was not the only type of refined theory of conflict available to formal strategists, nor was it always employed in a pure form. For a time, until the mid-1960s, the employment of matrices was the sine qua non of a serious strategist, though the models and mathematics were often applied in a somewhat desultory manner. The conclusions and prescriptions encouraged by Game Theory could often be reached by less exotic forms of reasoning. Some leading strategists such as Kahn used it only sparingly. Others, such as Morton Kaplan, had brief flirtations with Game Theory only to abandon it later. Oskar Morgenstern, one of the founders of game theory, explained in his book The Question of National Defense that the theory had ‘clarified the conceptual problem of decision-making, established the necessary theorems and shown the methods for selecting and computing the optimal strategies’, yet for much of the book its influence was hidden. It was more an accusation against the RAND alumni, put forcefully by their critics, rather than a claim they made for themselves. Schelling described himself as a social scientist who used game theory on occasion. ‘I don’t see that game theory is any more involved than Latin grammar or geophysics’, he observed of its presumed role in strategic theory, ‘but its quaint name makes mysterious and patronizing references to it an effective ploy’.31
The importance of Game Theory was that it provided a means of reducing strategic problems to a manageable form in which the dilemmas and the paradoxes of the age could be bared and solutions explored. It exemplified a certain type of thinking, presenting it in its purest form. It was certainly not a sufficiently powerful instrument to shape opinion itself. Nor could it be taken as some sort of scientific proof for particular policies, though it was on occasion presented as such. As a form of logical reasoning it depended on premises which were subjectively determined. It could offer valuable insights into behaviour. What it could not do was predict how actors in the real world would respond to situations approximating to those in the theoretical models, or even to say how they ought to respond, given that the relevant decision-makers could well attribute different values to the alternative outcomes of a particular ‘game’.
the intercontinental strategic nuclear confrontation was about as simple as any confrontation since maybe the Peloponnesian War. It was essentially bilateral. It didn’t involve knowing in detail anything about the adversary; it involved a very limited variety of weapons—essentially, nuclear explosives and long-range delivery systems or, in Europe, intermediate-range.33
The second important assumption was that of rational behaviour, based on calculations aimed at maximizing values. No particular set of values was associated with ‘rational’ behaviour in this sense. All that was required was that whatever the values, they were held in an explicit, transitive and internally consistent manner. The values attached to alternative outcomes of games were known as payoffs. The aim of a game was to maximize payoffs while being cognizant of the adversary’s similar intention. The range of choices permitted to each player was decided either by rules which could be mutually agreed, or by restraints imposed by the situation. Because each player affected the moves of the other, the resources available to shape alternative outcomes influenced the other’s moves through their potential. The actual application of resources came when the moves had been decided and were being played out.
Games could vary according to the number of participants (two-person, three-person … n-person games) and the extent of the conflict of interest. A zero-sum game was one in which the potential gains were fixed and so one player gained exactly what the other lost. Non-zero sum or variable sum games were those in which one’s gain did not necessarily correspond to the other’s loss. Both could win or both could lose, or there could be various mixes of gains and losses. For the most part, the application of Game Theory to strategic studies concentrated on two-person non-zero sum games. The two super-powers provided the players and the possibility of mutual nuclear destruction ensured that there was a shared interest in avoiding certain outcomes. If the payoffs for each player depended on decisions taken by the other player, and if there were certain outcomes which could suit the interests of both players, then what resulted was not all-out conflict but a carefully constructed bargain, which included elements of co-operation.
Some of the uses to which strategists put game theory were severely criticized.34 A frequent complaint was of unreal assumptions. The concern was not so much with the assumption of interdependence of players or the possibility of co-operation. Anatol Rapaport’s argument was its use was suited to the analysis of conflict and so allowed far too little scope for the analysis of forms of co-operation. The basic problem was with the assumption of rationality. The prescriptions of the theory only seemed to work if the players would, in practice, be rational, calculating and ‘utility-maximizing’. No regard was taken of a whole range of psychological and sociological factors—such as mental quirks, lack of awareness, domestic political pressures, value-conflicts or sheer errors of judgment. ‘Purely instrumental choice unhindered either by emotional or ideological blocks on the one hand, or by ignorance on the other, exists only in the abstract world of game theory.’35 Furthermore, the opportunities for action and thus the range of choice were determined by the structure of the game, which was the creation of the theorist. As the structure of any game was unlikely to be reproduced exactly in real life, one could not predict the course of rational decision. Finally the situations described in Game Theory were static. There was a single judgment about values, the calculation of moves, and the receipt of payoffs. Yet in international affairs policy-makers had to deal with a sequence of events in which their assumptions and circumstances were liable to rapid change. Their sense of values could also be affected as the full meaning of certain outcomes became apparent.
The tenuous relationship between the contrived world of the theory and the real world limited the practical validity of any conclusions. Yet, it was argued, the opposite was often believed. The excessive use of quantification and the scientific pretensions created an aura of objectivity that was quite undeserved. For example, the pay-offs connected with the alternative outcomes had to be compared along some scale. One course of action that might be rational if one assigned a high value to the prevention of a Soviet take-over of Western Europe would seem less sensible if one was indifferent to the fate of Western Europe. This created a danger, through the manipulation of payoff values, of putting exceptional and awful strategic moves likely to result in mass death and destruction into the realm of the possible and the rational.
A great deal of argument about military strategy … postulates the ‘rational action’ of a kind of ‘strategic man’, a man who on further acquaintance reveals himself as a university professor of unusual intellectual subtlety. In my view this kind of formal theorizing is of great value in the discussion on strategic matters when it represents not a prediction of what will happen in the real world but a deliberate and conscious abstraction from it, which must later be related back again to the world.37
The models could not predict or explain actual behaviour, though it would suggest criteria for judging behaviour. The most important point, however, was that there was no valid, empirically founded theory available. Herman Kahn’s response to an officer disputing his credentials to speak about thermonuclear war was: ‘Colonel, how many thermonuclear wars have you fought? Our research shows that you need to fight a dozen or so to begin to get a feel for it.’38 There were no analogous situations to draw upon. Human imagination or intuition was inadequate to cope. The abstractions of Game Theory and similar devices were useful as much because of the lack of suitable alternatives than anything else.
Despite the attention that game theory acquired it is important to keep its contribution in perspective. In one of the first surveys of the development of the field, written in 1963, RAND is accorded no more than a respectful place for its ability to explore the connections between weapons technology, military operations and broad strategic issues and for getting physical and social scientists to work together. Its staff, it was noted, had become ‘increasingly in demand as guest lecturers and visiting professors in academic circles’. The most striking feature of this survey, however, was the sheer variety of contributors identified.39 In addition to the established institutions such as the Council on Foreign Relations and university centres, contributions were noted from the atomic scientists concerned about the consequences of their invention, retired military men, leading figures from the Truman Administration who had left office in 1953, and Congressional staff who had been obliged to get up to speed with the complexity of defence policy. Of the civilian strategists those from the more classical tradition, such as Brodie and Kissinger, were marked out for their influence, although Kahn and Schelling got honourable mentions.