On Warning

18 SEPT 95


Some observations on the hazardous duty of conveying early warning in political and other "soft" areas of intelligence.


Keith Clark

The capabilities of U.S. intelligence have improved markedly in the course of the last fifteen years or so, but in the same period expectations about what it ought to be able to accomplish have probably grown even faster. This is natural enough, and probably professionally salutary for those who ply the trade, since most people need demanding requirements to keep them up to the mark. In any case, the government spends a great deal of money to equip itself with good intelligence and is rightly impatient with anything less than the best. But the situation does carry irritations and hazards for the professional. It is comparable to that in modern medicine, wherein improvement in techniques and medications, by giving rise to anticipation of consistent success, makes occasional failure a doubly grievous matter.

And by some standards intelligence fails more than occasionally, since it is considered in many quarters to have fallen down on the job if there takes place anywhere in the world an important, or sometimes even mildly interesting, political event which it had not heralded in advance in a way to make the warning stick in the minds of its consumers. We are all familiar with the queries and the resulting search of the record to find out whether top officials had been warned of such and such a development prior to its occurrence, and if not why. The short answer is often that these officials had indeed been warned, sometimes repeatedly, but won't admit it. This is the one likely to jump to the tongue of the participant in the post mortem, whether intelligence collector, analyst, or estimator: he had reported a week or a month ago that coup plotting was afoot in Ruritania and the government's position was shaky, so nobody should have been surprised when it was thrown out last night.

Whether or not anyone should have been surprised, however, the fact is that they often enough were surprised, and so inclined to ask why. Except in a narrow and not very profitable way, the analyst or estimator cannot meet the question by pointing out that an estimate or a current intelligence daily "covered" yesterday's big event when it noted weeks or months ago the possibility of a coup in Ruritania. Too often that report has been forgotten in the intervening stream of intelligence issuances and other papers or their equivalent in briefing sessions. Unless the consumer has been informed recently, and with sufficient emphasis and impact to make it stick, he has not in an effective sense been warned.

The following observations on this subject are intended neither as a defense of the intelligence community's record nor as definitive analysis and solution of the difficulty. The problem of crisis anticipation and early warning will continue with us, I suspect, despite the recurrent efforts of this computer age to gear up machines for effective and reliable prophecy in these soft areas of intelligence; here art, old-fashioned expertise, and a judicious amount of imagination still count for more than science. But while these reflections can offer no new secret insights or intellectual breakthrough, it may nonetheless be useful in a professional journal to record some guidelines and techniques derived from experience in asking the questions, if not always giving the right answers.

[Top of page]


Varium et Mutabile

The obvious first consideration is that the world itself is a chancey and uncertain place, in which change, sudden or gradual, is more the rule than the exception. One need only compare the world today, or any one area of it, with what prevailed 10 years ago to get a measure of the flux we live in. Technology, altering the lives and the thinking of men everywhere, has been accelerating the pace of even the most massive historical trends, the kind that used to take decades to work themselves out. To take one conspicuous example: with some stretching of the historical imagination one can imagine a colonial revolt against imperialism getting under way a century ago and gaining wide support in various parts of Asia, Africa, and Latin America, but one can scarcely picture such a movement winning hands down, but for a few isolated spots, in little more than a decade. Yet this is what has happened in the last 15 years, and the accompanying turbulence has generated some of the principal problems for U.S. foreign policy and intelligence during most of our official careers.

In this world and this period of history, the intelligence analyst and his customers are going to be nearer the mark if they think of change as more or less constant, and the main question as being not whether but when and how it will manifest itself. Unless proven otherwise, it should be assumed that a given society is changing daily. We face a semantic pitfall in the possibility of inferring from the overworked term "stable" or "stability" that things are remaining static; this attribute is often ascribed to a kind of mere surface calm below which change and flux are going on all the time.

If accepting the fact that change is normal and widespread predisposes us favorably, it still does not begin to solve the problems arising from what we have to work with in forecasting a particular change. In most cases the raw material of the evidence is necessarily fragmentary and inconclusive, and as it is rounded out it normally becomes not the stuff of early warning but news of current events. A number of things contribute to the poor quality of evidence on future developments.

One is the sheer impossibility of keeping track of the moves of every individual, organization, or government that may be in a position to change things in some part of the world. This difficulty is compounded when the success of the move for change depends on the ability of the promoters to keep it secret. If the coup plan that gets leaked is the one most likely to be frustrated by its enemies, it follows that a lot of such impending moves that have been reported either do not come off or go quite differently than anticipated. No one in the early-warning business can afford to overlook such reports in his own calculations, but some-of them are going to prove ill founded by reason of the same lack of secrecy that led to our getting them.

There is also the intrinsic element of caprice in the affairs of men and nations. Some events cannot be predicted because the principals seize sudden opportunities to act or are reacting to sudden stimuli, unforeseen and quite often unforeseeable by those on the spot. f the participants themselves could not have predicted the turn of events, the most sensitive and pervasive of intelligence systems would not be likely to do better. It is probably a salutary sign of awareness of such limitations that the unanticipated fall of Khrushchev was not followed, at least to my knowledge, by stem admonitions to the intelligence community to reform its procedures and sharpen its sense of urgency.

[Top of page]


Shotgun and Pinpoint

For those charged with intelligence warning there is of course a simple and appealing solution to these dilemmas--to point the gun in all possible directions. Warning always of everything gives you a technical defense against the charge that you failed to provide warning; it is also likely to lose you most of your readers or listeners and beat the remainder into a state of permanent hysteria or hopeless apathy. It is doubtful that anyone could be got to read an estimate or current intelligence paper big and fat enough to cover all the dire possibilities, and it is certain that the inflationary effect of this course on the value of intelligence warning would be ruinous.

A cardinal principle of effective warning intelligence, then, has to be selectivity. Selectivity involves rejection, and rejection involves risk. If intelligence is to eschew the shotgun approach in the interests of being read and respected, it will have to pick from the voluminous mass of often fragmentary and sometimes contradictory data a limited number of items to pass along, and sometimes what it rejects will later prove to be important. The hope is that the error will be corrected in time by the receipt of information supplementing or shedding new light on the rejected item and so promoting it out of the rejection category. Or perhaps another, better or luckier human mind will encounter the same fragment of information and respond more sensitively and perceptively--hopefully well in advance of the event it foreshadows. In the best of circumstances, however, selection will occasionally eliminate something that subsequently proves to have been important stuff. It is the argument of this essay that an occasional miss of this type is preferable to the overprudent shotgun alternative.

[Top of page]


Criteria: Likelihood

Now even a highly selective warning system will have to deal in possibilities more often than in probabilities or near-certainties. Reasonable prudence requires that a government be prepared, at any given moment, to cope or at least live with a number of contingent possibilities only some of which will in fact materialize. If something could happen, it had better be borne in mind, whether it will "probably" happen or not. This being the case, some fairly substantial proportion of the warnings delivered will in the event prove exaggerated or will otherwise not be borne out by subsequent developments. (Sometimes the fact that a warned-of development fails to come off may be due to U.S. action triggered by the warning; here intelligence has done its job to perfection even as its prophesies fail to come true.)

Errors on the side of caution are less harmful than neglect of warning, but they are not harmless. A false alarm will normally be overlooked or forgiven much more easily than a failure to call the shot on something that does happen; but both are errors and both ought to be on the consciences of those in the warning business. Most of us recall with acute pain instances in which intelligence failed to forecast something that did occur. A review of the dangers and opportunities warned of that did not materialize may give less pain but is still sobering.

[Top of page]



The area between these two kinds of error thus represents one of the criteria in the process of selection--degree of likelihood. The standard is admittedly a fuzzy one. A second criterion offers somewhat solider ground, namely the importance of the matter being warned of. It is often, though not always, easier to judge how significantly some contingency would affect our interests than how likely it is to occur. Common sense and a reasonable familiarity with the scope of our government's interests and activities usually enable us to tell whether some foreseeable event would be of critical, great, moderate, little, or no importance to national or departmental interests. In any case the policy makers' judgment on this score can supplement our own.

The complexity and many responsibilities of a government like ours suggest that very few foreign developments would fail to be of concern to some department or program. As a criterion for warning selection, then, the question of importance probably refers less to whether than to whom to warn and how. Some predictions should have top billing in national intelligence publications or briefings, others more subdued treatment in departmental or specialized issuances. The criterion is thus most usefully relevant to selection for briefings and publication at the highest levels.

It is this writer's subjective and purely personal opinion that the application of more vigorous standards in this respect would have a salutary effect on the bulk and readability, and hence on the impact, of most intelligence publications, not excluding the national estimates. An urge for completeness and detailed perfection is a good thing, but sometimes an inordinate amount of time and energy is spent in perfecting presentations of detail which can make no earthly difference to policy decisions but confront already overburdened readers with more information than they want or need to know. This is not an argument for either carelessness or super-finality but a plea for the classic virtues of brevity and concentration on the essential as still useful in our line of work.

[Top of page]



The criteria of likelihood and importance for determining whether, how, and to whom to give early warning are supplemented by a third, that of imminence, which is most relevant to the choice of when to warn. This timing is often of critical importance, for policy makers are as human as the rest of us and busier than most. On the higher levels they are subjected to a mentally exhausting barrage of publications and briefings on a host of subjects, and in the daily round of attending to inescapably urgent things, some of the rest are going to be remembered and some are not.

Selection in the light of imminence is a matter of avoiding unacceptable extremes, warning too early or too late. Logically it might seem the earlier the better, giving as much time as possible to do something about it, but this logic leads to presenting a catalog of all kinds of important things that may or are likely to happen eventually. Though it is unquestionably desirable to look ahead, in appropriate context, with a general prediction of developments that seem ultimately probable, our problem here is a pointed particular warning at a time when something can and should be done about it.

Even the most prudent and forward-looking administration cannot give as serious attention to a problem foreseen five years ahead as to one shaping up next week. It is not just that something postponable is crowded off the stage by real and present dangers; there is often little that can or should be done about some foreseen events until they are closer at hand. There is always the chance that the contingency will not arise when expected or not at all.

It is true that in addition to delivering specific warning at the right time, intelligence has a responsibility to keep its consumers sufficiently aware of the remoter contingencies, of what Walt W. Rostow recently described as "the relevance of the less obvious."1 It has to do this without dulling their senses or straining their patience with frequent laundry lists of all imaginable horrors. I confess it is much easier to state this problem than to offer any but the most banal answers. One line of procedure, however, while more the result of evolution in the art of policy making than of intelligence innovation, does offer the intelligence officer some help. I refer to the increased emphasis in recent years on isolating and studying very long-range policy problems--issues of a sort which may not require U.S. counter action for several years to come. It may be debated whether the policy lines worked out in these exercises will in most cases be followed when the moment for action comes--certainly it will not be just a matter of lifting a ready-made "courses of action" formula out of the files--but the long lead-time concept is salutary for policy planning, and its acceptance makes the job of intelligence warning a few degrees easier and conceivably a bit more fruitful. In an uncertain world perhaps we can't ask for much more.

1 In a lecture on "The Planning of Foreign Policy," given at the School of Advanced International Studies of the Johns Hopkins University and published in The Dimensions of Diplomacy (E. A. Johnson, ed., Johns Hopkins Press, Baltimore, 1964).


[Top of page]

Posted: May 08, 2007 07:58 AM
Last Updated: Aug 04, 2011 03:49 PM