Small Wars Journal

Fitting Intelligence to the Fight: Lessons from Afghanistan

Sat, 07/20/2013 - 7:12pm

Fitting Intelligence to the Fight: Lessons from Afghanistan [i]

David J. Katz [ii]


The types of information needed by the military to conduct population-centric counterinsurgency operations in Afghanistan shares little in common with intelligence used for lethal targeting. Years into the Afghanistan campaign, recognition that success required effective population-centric interventions prompted a redirection of intelligence assets to focus more on acquiring detailed sociocultural information about target populations. Other reforms have sought to remedy information deficiencies by revamping the assumptions and concepts that frame the analysis of populations, their identities and attitudes and the so-called “hearts and minds” efforts seeking to change them. These adjustments during a campaign are welcome but still fall short of supplying information that can confer a strong confidence in the success of these non-kinetic activities. Appreciation of the inherent limits to the information and the level of confidence that it confers for such activities needs to be brought into the planning and execution of population-centric interventions of any ilk, whether stability operations, counter-insurgency or humanitarian assistance, regardless of their size and political significance.

Fitting Intelligence to the Fight

As we head for the exits in Afghanistan, reflecting on our performance there yields value if it institutionalizes practice and experience in doctrine so that it is not lost as operations end and people move on. In particular, we can benefit by studying the role of intelligence in conducting population-centric counterinsurgency. While Defense Secretary Panetta announced in January 2012, that the Army and Marines will no longer be sized to support large-scale, long-term stability operations, the military can count on getting involved in complex contingency operations which will have many similarities to the population-centric operations conducted during the campaigns in Iraq and Afghanistan. [iii]

Two broad questions frame this examination: Has the intelligence community been doing the job? Has it been the right job?

In January 2010, Major General Mike Flynn co-authored a paper published by the Center for a New American Security, a Washington think-tank, titled, Fixing Intel: A Blueprint for Making Intelligence Relevant in Afghanistan. In it Flynn, who was serving as the intelligence chief for ISAF Commander General McChrystal, delivered a blunt critique of intelligence effectiveness in Afghanistan, declaring that, “Eight years into the war in Afghanistan, the U.S. intelligence community is only marginally relevant to the overall strategy.” [iv]

The problem for Flynn was that our massive intelligence effort was directed at finding and destroying the enemy but, “Lethal targeting alone will not help U.S. and allied forces win in Afghanistan.” [v] Intelligence was hard at work but it was doing the wrong job, or more precisely, not enough of its capacity was devoted to getting information about “the environment in which we operate and the people we are trying to protect and persuade." [vi] Commanders and senior government leaders lacked vital information about the Afghan people. In a counterinsurgency, tactical-level information has far greater strategic significance than in conventional conflicts, argued Flynn. “U.S. intelligence officers and analysts can do little but shrug in response to high level decision-makers seeking the knowledge, analysis, and information they need to wage a successful counterinsurgency.” [vii]

Flynn made clear that what he proposed in this paper involves “changes that must occur” and “should be considered as a directive by the senior author.” [viii] He called for stepping up collection and production of information about the population through creating special teams of analysts “empowered to methodically identify everyone who collects valuable information, visit them in the field, build materially beneficial relationships with them, and bring back information to share with everyone who needs it.” [ix]

While Fixing Intel casts its concerns in terms of the conduct of a counterinsurgency campaign, the focus is mainly on what is needed for the conduct of stability operations which are those missions, tasks, and activities undertaken by the military, and in some instances along with civilian organizations, to maintain or reestablish for the population a safe, secure environment, deliver essential services, tend to critical infrastructure, and supply humanitarian relief.

The teams that Flynn proposed were to work in new units he called “Stability Operation Information Centers,” that would be located at the regional command headquarters throughout the country. These facilities would research and write “meaty, comprehensive descriptions of pivotal districts.” [x] The products would be kept at the lowest classification level possible so that they could not only be used for intelligence purposes but made available to “all elements with demand for information – including Afghan partners and non-government actors.” [xi]

Flynn sought to separate SOICs from the intelligence operations focusing on the enemy in order to ease outsider access and to promote at these centers a different approach to gathering, processing and disseminating information. He even suggested that in the south and east where the international presence was the greatest, SOICs be placed under the U.S. State Department’s senior civilian repre­sentatives who administer governance, development and stability efforts, outside of the military chain of command. [xii]

The SOICs were launched and commenced operations. During Flynn’s tenure and that of his successor, Brigadier General Stephen Fogarty, SOICs were brought under a new umbrella organization, called the “Civil-Military Integration Program” which was established in response to the problems identified in Fixing Intel. Also brought under the Civil-Military Integration Program were the Human Terrain System with its Human Terrain Teams deployed in the field, the Atmospherics Program – Afghanistan, which gathered open source information, and other units involved in gathering and processing sociocultural information.

In a review issued June 2012, 30 months after publication of Fixing Intel, the Department of Defense Inspector General asked whether SOICs were doing the job. The IG found that SOICs had managed to improve the ability to provide the type of information that Flynn had called for in his article. However, there were problems as well, the kind that are common for such ad hoc initiatives launched in the middle of a high-tempo combat operational environment. [xiii] The IG recommended doctrinal and organizational changes together with adjustments to the training for analysts assigned to collect and analyze population-centric information.

In Fixing Intel Flynn sought to drive home to the intelligence community the need to put more importance and resources on collecting information about the population. The IG report suggested that this effort was successful as far as it went but that significant institutional changes needed to be undertaken in order to maintain such capabilities into the future.

The intelligence community’s institutional inertia noted by the IG in the SOIC review was also cited by Ben Connable as one reason that its intelligence products are ill-suited to support counterinsurgency. In a 2012 Rand study, Military Intelligence Fusion for Complex Operations: A New Paradigm, Connable, like Flynn, argues that intelligence organizations are failing to provide commanders and policymakers an effective understanding of the complex counterinsurgency environment. [xiv] Their products give an intelligence picture that artificially deconstructs the environment and the people and groups within it and does so in ways that simplifies matters while distorting their interrelatedness. Connable is particularly bothered by approaches involving color-coding people and groups and treating those in the different categories as distinct: as enemy (red), friend (blue or green) or neutral (white). This scheme fails to accommodate the reality of persons having multiple identities, sentiments and affiliations, often simultaneously. Putting them into color-coded bins may satisfy a “nearly unquenchable need for clarity,” but introduces distortion and artificial simplicity. [xv] Connable also criticizes the reliance on the “System of Systems Analysis” approach that reifies and deconstructs a population into discrete and interactive sectors. This is again artificial and produces an ungainly, mechanistic and ultimately unworkable analytic approach.

Interrelated complexity in the target population needs to be recognized and conveyed accurately, writes Connable, without resort to either artificial simplicity or mechanistic models that link contrived sectors in contrived ways. Connable proposes a “new paradigm,” what he calls a “behavioral intelligence analysis” that eschews channelization through color-coding and focuses on factors critical to inducing behavioral change.

Both Flynn and Connable recognize the essential difference between approaches suited for intelligence with an enemy-centric focus and those supporting population-centric counterinsurgency. Connable sees a need for the counterinsurgent to have information that can be used to change attitudes and behavior of all persons in the environment and for this purpose he discounts the relevance of treating persons in the enemy (red) category different from those put into the other color-coded bins. Flynn, on the other hand, proposes a shift to an approach for counterinsurgency intelligence on the environment and population which involves modifying the five components of the kinetic targeting approach: find, fix, finish, exploit and analyze. “Now, put that model into a non-kinetic social engineering construct, and maybe it’s not find, fix, finish; maybe it’s find, feel, understand.” [xvi]

Unlike kinetic targeting of the enemy, population-centric counterinsurgency involves a different relationship between information and the level of confidence derived from the information. For kinetic targeting there is a positive relation between information and the level of confidence: the more relevant information collected and analyzed, the higher the confidence in the success of the planned action. For counterinsurgency social engineering efforts, the correspondence between information and confidence is not so straightforward. Instead, the level of confidence follows a bell-shaped curve: additional information initially increases confidence to a point and then may cause confidence to plummet.

In the early stages of a counterinsurgency campaign, when the environment is unfamiliar, information is limited as is confidence about the soundness and efficacy of possible interventions. Given this uncertainty about efforts and their results, the obvious inclination is to seek additional information. In our recent campaigns in countries about which we had little direct experience, institutionalized knowledge or detailed information before we launched operations, this requirement spurred a frantic drive to gather information from official sources, academia and also private-sector contractors.

The limited time to prepare and the lack of adequate information readily available resulted in hastily concocted products that were often superficial and subsequently shown to have little practical value for conducting counterinsurgency interventions. Regardless of the shortcomings of such material, when it is supplied to the counterinsurgents with assurances that experts had a hand in preparing it and it appears to supply the detailed understanding needed to accomplish their tasks, their confidence rises.

It may remain high throughout a unit’s deployment and beyond. However, if the counterinsurgents subsequently acquire more and better information about the target population or they gain a fuller appreciation through their direct experience then it is almost inevitable that their level of confidence about the effectiveness of their interventions will fall and fall dramatically.

Abundant evidence for this comes from reports and assessments that describe shortcomings and failures common to counterinsurgent interventions that have been undertaken in Afghanistan and how they have failed to achieve their intended objectives. [xvii] Without a single exception, to my knowledge, the researchers and specialists having extensive on-the-ground experience in Afghanistan that spans decades dismiss the effectiveness of stability operations and, in fact, cite instances where the interventions produce counter-productive results.

Is it possible to gather sufficient information for the counterinsurgent to understand how to intervene successfully? Should the counterinsurgent even aspire to attain a level of detailed knowledge comparable to the intelligence used in kinetic targeting where requirements demand a high degree of confidence? What is possible for the counterinsurgent with respect to obtaining information ensuring success? The ideal obviously is a high degree of confidence based on a thorough understanding of the target population. In this regard the gold standard, the absolute best knowledge and information about a society that we can envision is the population about which we have the fullest knowledge, understanding and experience: our own.

When engaging in social engineering in our society, we are wary of those ‘experts’ who present themselves as “understanding” our society, or even a narrow sector of it. And, if a person offers himself or is said to have such expertise, we regard it as foolhardy to rely on a single expert or even a handful of experts for guidance on any non-trivial matter. We recognize that experts on a particular topic will see things differently and take different positions. This is often the case even for experts steeped in same discipline or specialty. And in situations where there is agreement among the best and the brightest on the way forward, the cases where “slam-dunk” forecasts about likely outcomes in our society turn out to be anything but are legend.

Given the difficulties and uncertainty that characterize social and behavioral engineering in our own society and the obvious factors that make barriers to obtaining detailed knowledge about other populations, we must accept how far we are and will remain from “understanding” another society, even ones that are small-scale and strike us as simpler and less complex than our own.

Instead of aspiring to gain an understanding adequate to engage in social engineering with high confidence about the outcome, it is prudent to lower our expectations and set ourselves on a more realistic path where we seek to acquire a fuller appreciation than we have hitherto. We can then bring this limited yet marginally better knowledge to our planning together with an explicit recognition of the uncertainty and low confidence about the effectiveness of prospective interventions. SOICs, in concept, are a step in the right direction but they are still limited in their effectiveness and, even if they deliver as intended, that does not alter what they can accomplish, which falls far, far short of delivering products that allow for “understanding.”

Another factor that weighs heavily on the effectiveness of population-centric counterinsurgency interventions, yet receives too little attention, concerns the methodological difficulties involved in assessing and evaluating intervention effects. The social and behavioral sciences, no less than laboratory sciences, recognize the myriad ways that bias and distortion enter into measurement of effects even in tightly controlled laboratory settings. The opportunities for bias and compromise increase dramatically in natural settings which render efforts to monitor, measure and assess challenging if value is accorded to accurate, reliable, valid results.

Determining the effects of ‘warhead to the forehead’ kinetic targeting is many orders of magnitude simpler and less susceptible to error than assessment of counterinsurgency interventions with populations. Acknowledging the possibility of so-called second- and third-order effects from interventions and broadening the time horizon for conducting the assessment do nothing to overcome the intrinsic difficulty of meaningfully monitoring and assessing consequences. The intractable problems inherent to such tasks facing the counterinsurgent differ only in their added complexity from those bedeviling comparable efforts undertaken by civilians in peacetime.

As the difficulties and limits on measurement do not preclude such efforts in peaceful settings, they need not for the counterinsurgent. However, the quest for reliable metrics and assessment demands a lucid appreciation of what is involved, what can be accomplished and the need to present results that are forthright about the factors that must be weighed in evaluating their significance.

The intent here is not to build an iron-clad case against population-centric counterinsurgency activities. Instead, it is to point to complexities and uncertainties of such interventions, difficulties that cannot be overcome by recognizing, as Flynn does, the need to rebalance the intelligence effort to collect more information about the target population or adopting analytic processes, as Connable does, that apply more appropriate models of human behavior.  

Beyond the issues raised here, there are many other serious ones related to counterinsurgency interventions which deserve to at least be recognized as posing legitimate and profound concerns about the morality of the endeavor and the nature, scope and limits of gaining relevant knowledge both with respect to designing the intervention and determining their effect. These cannot be wished away but are left to academic faculties and think-tanks to tackle. Warfighters have enough to worry about without venturing into these murky ethical and epistemological matters, but they cannot dismiss the need to admit to and take into consideration the complexities, limits and difficulties inherent to their efforts.

MG Flynn chose a public and unconventional pulpit to press his intelligence community to reallocate its resources and commitment to matters that many intelligence professionals do not consider as falling within their ambit. [xviii] As he had done in Iraq with the creation of enemy-focused fusion cells, Flynn directed organizational reforms he saw as necessary to make intelligence relevant for the type of campaign being waged in Afghanistan. Connable called for reform that focused on shifting the ways that analysts think in order to improve their capability to generate material useful for altering behavior of target populations.

While both such reform initiatives are useful, population-centric intelligence needs to be considered in its broader context. What counterinsurgents have attempted to do in Afghanistan through stability operations is armed nation-building or perhaps more neutrally, armed social engineering. Whether we call it nation-building or social engineering, doing this in the midst of hostilities saddles it with added complexity and difficulty and, of course, greater danger. That the military is doing this; that we deem this to be virtuous; that we consider success a matter of urgency and vital to our nation’s well-being and security; all of these factors do not make what we are attempting any easier, or the prospects for success any greater. Nor does it mean that interventions that do not work in a peacetime setting will produce satisfactory results and that all of the inherent problems and complexities can be ignored without consequences. In the best of circumstances, with the fullest knowledge and information, these activities are hard, beset by uncertainty, and seldom unfold in the manner anticipated. The added difficulties piled on to such efforts during a military intervention push them right up to the edge of abyss of the impossible.

For intelligence to be relevant in a counterinsurgency or other complex operation, it needs to supply commanders and policymakers with products that are explicit and put in boldface the risks, uncertainties and difficulties involved. That information, along with much else including the realistic presentation setting out the entire range of possible outcomes, the time needed to achieve the objectives, and the intractable issues tied to information collection must be presented in ways so they are integrated fully into the design and planning for the intervention itself.

Until our leaders are supplied with the products that do this in campaigns where the effort is directed at populations rather than eliminating enemies, our intelligence establishment may be doing the job, but not the job it should be doing.


Connable, B 2012 Military intelligence fusion for complex operations: A new paradigm, Occasional Papers, Santa Monica: Rand Corporation. Available at [Last accessed 15 July 2013].

Deputy Inspector General for Intelligence and Special Program Assessments 2012 Review of stability operations information centers in Afghanistan, Alexandria, VA: Inspector General, Department of Defense, June 18, 2012. “Results in Brief” available at [Last accessed 21 June 2013].

Flynn, M 2011The situation in Afghanistan. In: Pool, R Sociocultural data to accomplish Department of Defense missions: Toward a unified social framework. Washington, DC: Planning Committee on Unifying Social Frameworks, National Research Council, pp. 9-16. Available at [Last accessed 12 June 2013].

Flynn, M, Pottinger, M and Bachelor, P D 2010 Fixing intel: A blueprint for making intelligence relevant in Afghanistan, Washington, DC: Center for a New American Security. Available at [Last accessed 12 June 2013].

Fishstein, P and Wilder, A 2012 Winning hearts and minds? Examining the relationship between aid and security in Afghanistan. Feinstein International Center, Tufts University. Available at [Last accessed 12 June 2013].

Regional Command West Stability Operations Information Center undated The stability operations information center (SOIC): Comprehensive understanding for comprehensive operations, Herat, Afghanistan: International Security Assistance Force. Available at [Last accessed 12 June 2013].

[i]  A version of this paper was presented at the EMC Chair Workshop, “Intelligence, National Security, and War,” March 25-27, 2013, Newport, RI. The author wishes to thank Karen Finkenbinder, Paul Overby and Roy Herrmann for offering suggestions and comments on drafts of this article.

[ii]  The author is not the same person as an author with the same name of the article, Reforming the village war: The Afghanistan conflict, Middle East Quarterly, Spring 2011, pp. 17-30.

[iii]  Defense Secretary Panetta, Defense Strategic Guidance Briefing from the Pentagon, January 5, 2012,

[iv]  Flynn et al., 7.

[v]  Ibid., 8.


[vi]  Ibid., 4.

[vii]  Ibid., 7.

[viii]  Ibid., 9-10.

[ix]  Ibid., 17.

[x]  Ibid., 19.

[xi]  Ibid.


[xii]  Ibid., 21.

[xiii]  Deputy Inspector General for Intelligence and Special Program Assessments.

[xiv]  Connable, 1.

[xv]  Ibid., 17.

[xvi]  Flynn, p. 14. Another population-centric targeting model similarly elevates “understand” to a primary importance. It is organized around five interconnected functions: understand, decide, detect, deliver, and assess. Regional Command West Stability Operations Information Center, p. 14.

[xvii]  See, for example, Fishstein and Wilder.


[xviii]  Flynn rebuts those who argue that such information is not the responsibility of intelligence shops:  “Some intelligence officers contend that ‘white’ topics are not intel’s job but the responsibility of civil affairs and stability staffers – the CJ9. However, CJ9 lacks the analysts, training, and resources to systematically gather, process, and disseminate relevant ‘white’ information.” Flynn et al., p.22.


About the Author(s)

David J. Katz is a career Foreign Service Officer currently serving as State Department Advisor to the U.S. Army Peacekeeping and Stability Operations Institute, Carlisle, Pennsylvania. His involvement with Afghanistan dates from the mid-1970s when he conducted field research for his PhD dissertation (Anthropology, UCLA 1982) in what is now Nuristan Province. The most recent of his six State Department assignments dealing directly with Afghanistan and Pakistan was serving as Deputy Director, Force Reintegration Cell at ISAF HQ, Kabul, from December 2009 to June 2011. 


Bill M.

Mon, 07/22/2013 - 12:04am


Impressive article that is also timely as we debate our future force structure (organization and capabilities). We have recognized the need for gaining better understanding of sociocultural aspects, but as you point out we'll always have severe limitations in our ability to do so due to a range of epistemological matters. While the military can't afford to focus on philosophy and epistemology, its leaders must understand that our information in this area will frequently be incorrect regardless of good our collection systems and analytical processes may be. By at least recognizing this, we can limit the risk of getting it wrong and not recognizing we got it wrong and stay on a path to nowhere. If we realize our understanding is inherently flawed, then we will more likely to look for indicators we're on the wrong path, instead of using metrics to artificially validate our assumptions. This allows us to adjust course as needed, which ties into something I read in a GEN (Ret) McCrystal article about not be able to steer a ship that isn't moving.

We don't know what type of wars we will fight in the future, so it is risky to divert too much capability from our intelligence structure to sociocultural intelligence, especially since as you have pointed out it is a fuzzy field to begin with. I didn't see the human terrain teams produce anything of real value downrange, and our overall intell products briefs), especially the graphic ones that are color coded or use numbers to assess villages were not helpful in gaining anything close to ground truth. I felt we didn't really move one past the disaster of McNamara's role of DOD when he demanded metrics of no value that eventually led to failed strategy. That begs the questions: what sociocultural information do we need? What can we realistically expect to get? How do will we use that information?

I think Connable's proposal to focus on behavioral analysis is more reasonable than the wide range of measures we currently assess to no end. I don't know if this can realistically be done, but Steve Tatham and Andrew MacKay wrote a darn good book on the topic titled Behavioral Conflict.…

I think a greater appreciation of strategy would help us focus on what type of intelligence we need. If we're simply out to kill members of Al-Qaeda, and don't care about the impact of our operations and have no larger vision for the country or region, then our current intelligence system isn't that bad. Probably give it a C+, they do enough sociocultural analysis to vector us in on targets. However if we have larger aims such as a relatively stable region, then we need to determine if simply killing our adversaries will achieve that strategic end. If not, then what do we think we should do to achieve that end? Then what intelligence do we need to inform decision makers? We have seen that killing our adversaries isn't working, nor is our economic aid working, and nor is building up Afghanistan's security forces. The behavior of those we're trying to influence for the most part hasn't changed, so it would seem there is adequate evidence there are gaps in our understanding that prevent us from effectively shaping the behavior we desire, or perhaps more realistically the gaps prevent us from realizing our objectives are unrealistic? Either way, there is a demand for better sociocultural intelligence, and even a slightly better understanding could be decisive. Since a slightly better understanding is all we can realistically expect to achieve, how do we do organize and train to achieve it? Who should do it? Is it a DOD function, USAID, Dept of State? All the above?


Sun, 07/21/2013 - 10:46am

Do you feel this second intel shop/capability should be housed in DoD or elsewhere, like State? What would be your org and op models?