Tuesday, June 10, 2014

Thinking in Parallel (Part Three - Testing The Mercyhurst Model Against The Real World)

Part 1 -- Introduction
Part 2 -- The Mercyhurst Model

For the last 11 years, I have been using the model described in Part 2 to structure my Strategic Intelligence class at Mercyhurst University.  This is a capstone class for seniors and 2nd year graduate students within the Intelligence Studies program at Mercyhurst.  This class is centered on a real world project for a real-world decisionmaker, often within the US National Security Community.  To date, I have overseen 133 of these types of projects.

The broad parameters of the projects have remain unchanged since 2003.  Students in the class are divided into teams and are assigned by the instructor to one of 4-5 projects available during that term.  Each project is sponsored by a national security, business, or law enforcement organization that has a strategic intelligence question.  To date, sponsors of these questions have included organizations such as the National Geospatial-intelligence Agency, the Defense Intelligence Agency, the National Intelligence Council, the National Security Agency, 66th Military Intelligence Group, and the Navy’s Criminal Investigative Service to name just a few.  To give readers a sense of the wide variety of questions intelligence studies students are expected to answer in this course, I have listed a few recent examples of them below:
1. What role will non-state actors (NSAs) play and what impact will NSAs have in Sub-Saharan Africa over the next five years?o What is the likely importance of NSAs vs. State Actors, Supra-State Actors and other relevant categories of actors in sub Saharan Africa?o What are the roles of these actors in key countries, such as Niger?o Are there geographic, cultural, economic or other patterns of activity along which the roles of these actors are either very different or strikingly similar?o What analytical processes and methodologies were applied to the questions above and which proved to be effective or ineffective? 
2. What are the most important and most likely impacts on, and threats to, US national interests (including but not limited to political, military, economic and social interests) resulting from infectious and chronic human disease originating outside the US over the next 10-15 years?  
3. What are the likely trends in Brazil’s oil/liquid fuel market and electric power sector in the next ten years?  Where will these trends likely manifest themselves?o What energy capacity and security issues are likely to be the most significant to Brazil’s economy in the next ten years?o How will Brazil likely address current and/or future energy security issues over the next ten years?o Where will Brazil address these energy shortfalls?
In each case, students had only 10 weeks to conduct the research, write the analysis and present the final product to the decisionmaker.  The students had no additional financial resources available to them and, other than the question itself, received no support directly from the decisionmaker.  Students rarely had any subject matter expertise in the area under question and were only allowed to use open sources.  Students were expected to integrate lessons learned from all previous intelligence studies classes and to manage all aspects of the project without significant supervision.  Finally, all the students, in addition to this project, were taking a full academic load at the same time.  

After all of the deliverables had been produced and disseminated, the decisionmakers sponsoring the projects were asked to provide objective feedback directly to the course instructor.  This feedback, in turn, was evaluated on a five point scale correlated with traditional grading practices and professional expectations.  In short, a 3 on this scale is roughly equivalent to a "B" and a “4” on this scale is roughly equal to “A” work in a university setting.  A “5”, on the other hand, is the kind of work that would be expected from a working (albeit junior) intelligence professional.  The chart below indicates how the annual averages have changed over time.

(Note:   While this chart may appear to reflect grade inflation more than any other suggested effect, it should be noted that “A” is essentially “average” among Mercyhurst University Intelligence Studies seniors and 2nd Year graduate students and has been for the entire time frame shown above.  The current dropout rate from the program is approximately 50% and much of that is due to a strict 3.0 minimum GPA in order to stay in the program.  As a result, seniors and second year graduate students (the only students allowed to take the class), typically have GPAs that average 3.6 or above.  For example, two years ago, 18 of the top 20 GPA’s in the entire University belonged to Intelligence Studies students.)
Anecdotally, it is possible to state the exact impact of these reports within national security agencies in only a few cases.  For example, the report that answered the question on global health mentioned earlier earned this praise from the National Intelligence Council: 
“Although the Mercyhurst "NIE" should not be construed as an official U.S. government publication, we consider this product an invaluable contribution to the NIC's global disease project: not only in terms of content, but also for the insights it provides into methodological approaches. The Mercyhurst experience was also an important lesson in how wikis can be successfully deployed to facilitate such a multifaceted and participatory research project.”
Likewise, in David Moore’s book Sensemaking:  A Structure for an Intelligence Revolution (published by the National Defense Intelligence College in 2011), the study on non-state actors in sub-Saharan actors produced in answer to the question mentioned above was judged more rigorous than a similar study conducted by the National Intelligence Council (in cooperation with the Eurasia Group).

Beyond the national security community, however, the impact of these reports on various businesses and other organizations is often easier to determine.  For example, senior managers at Composiflex, a mid-sized composites manufacturer, indicated, “We used this project as a seed for our new marketing plan in 2007 and now an industry that we had not even tapped before is 30% of our business.” 

Likewise, Joel Deuterman, the CEO of Velocity.net, an Internet Service Provider, stated, “The analysts discovered that our approach was actually a cutting-edge, developing standard in our industry…What really substantiated the data for us was to see many of our existing customers on the list. Then we knew we could rely on the validity of the ones they had found for us.”  

Even foreign organizations have seen the benefit of these products including Ben Rawlence, the Advisor for Foreign Affairs and Defense in the Whip’s Office of the Liberal Democrat Party in the UK, stating, “The research carried out by your students was first class, and has been of substantial use to Members of Parliament…  It was comprehensive, well sourced and intelligently put together.  I have had no hesitation recommending it to our MPs and Lords in the same way that I recommend briefings provided for us by professional research organisations…” 

While it is possible to imagine more rigorous testing of this model of the intelligence process, the long term success of the process in generating actionable intelligence for a wide variety of customers on a range of difficult problems in a very short time using limited resources is hard to ignore.  More importantly, not only has the process proven itself successful but this success has trended upwards as improvements have been made over the years in terms of structuring the course and teaching material consistent with this approach to the intelligence process.


Intelligence in the 21st century is best thought of as a series of sub-processes operating interactively and in parallel.  

This conclusion, by itself, has significant implications for the training and education of intelligence professionals.  In the first place, it suggests that it is no longer possible to specialize in one area to the exclusion of another.  Intelligence professionals will have to be trained to think more broadly, to be able to jump more fluidly from modeling to collection to analysis to production and back as the process of creating intelligence moves forward over time.  

Likewise, hardware and software support systems will need to be designed that facilitate this leaping back and forth between the various sub-processes.  Designing products that work sequentially in a parallel world will not only frustrate but will also slow down the process of generating intelligence – a result that is absolutely counter to the intelligence needs of modern decisionmakers.  

Finally, as dramatic as this type of change might appear to be, it is, perhaps, better thought of as merely aligning the training and education of intelligence professionals with what it is they already do.

Monday, June 9, 2014

Thinking In Parallel (Part 2 - The Mercyhurst Model)

Part 1 -- Introduction

While a number of tweaks and modifications to the cycle have been proposed over the years , very few professionals or academics have recommended wholesale abandonment of this vision of the intelligence process.  

This is odd.  

Other fields routinely modify and improve their processes in order to remain more competitive or productive.  The US Army, for example, has gone through several major revisions to its combat doctrine over the last 30 years, from the Active Defense Doctrine of the 1970’s to the AirLand Battle Doctrine of the 80’s and 90’s to Network Centric Operations in the early part of the 21st Century.  The model of the intelligence process, the Intelligence Cycle, however, has largely remained the same throughout this period despite the criticisms leveled against it.  The best answers, then, to the questions, “What is the intelligence process?” and “What should the Intelligence process be?” remain open theoretical questions, ripe for examination.

There are common themes, however, that emerge from this discussion of process.  These themes dictate, in my mind, that a complete understanding of the intelligence process must always include both an understanding of intelligence's role in relationship to both operations and the decisionmaker and an understanding of how intelligence products are created.  Likewise, I believe that the process of creating intelligence is best visualized as a parallel rather than as a sequential process.  I call this the "Mercyhurst Model" and believe it is a better way to do intelligence.  More importantly, I think I have the evidence to back that statement up.  

The first of the common themes referenced above is that the center of the process should be an interactive relationship between operations, the decisionmaker and the intelligence unit.  It is very clear that the intelligence process cannot be viewed in a vacuum.  If it is correct to talk about an “intelligence process” on one side of the coin, it is equally important for intelligence professionals to realize that there is a operational process, just as large if not larger and equally important if not more so, on the other side and a decisionmaking process that includes both.

The operational and intelligence processes overlap in significant ways, particularly with respect to the purpose and the goals of the individual or organization they support.  The intelligence professional is, however, focused externally and attempts to answer questions such as “What is the enemy up to?” and “What are the threats and opportunities in my environment?”  The decisionmaking side of the coin is more focused on questions such as “How will we organize ourselves to take advantage of the opportunity or to mitigate the threat?” and “How do we optimize the use of our own resources to accomplish our objectives?”  In many ways, the fundamental intelligence question is “What are they likely to do?” and the decisionmaker’s question is “What are we going to do?”  The image below suggests this relationship graphically.

The second theme is that it should be from this shared vision of the organization’s purpose and goals that intelligence requirements “emerge”.  With few exceptions, there does not seem to be much concern among the various authors who have written about the intelligence process about where requirements come from.  While most acknowledge that they generally come from the decisionmakers or operators who have questions or need estimates to help them make decisions, it also seems to be appropriate for intelligence professionals to raise issues or provide information that was not specifically requested when relevant to the goals and purpose of the organization.  In short, there seems to be room for both “I need this” coming from a decisionmaker and for “I thought you would want to know this” coming from the intelligence professional as long as it is relevant to the organization’s goals and purposes.

Theoretically, at least, the shared vision of the goals and purpose of the organization should drive decisionmaker feedback as well.  The theoretical possibility of feedback, however, is regularly compared with the common perception of reality, at least within the US national security community, that feedback is ad hoc at best.  There, the intelligence professionals preparing the intelligence are oftentimes so distant from the decisionmakers they are supporting that feedback is a rare occurrence and, if it comes at all, is typically only when there has been a flaw in the analysis or products.  As former Deputy Director Of National Intelligence for Analysis, Thomas Fingar (among others), has noted, “There are only two possibilities: policy success and intelligence failure” suggesting that “bad” intelligence is often a convenient whipping boy for poor decisions while “good” intelligence rarely gets credit for the eventual decisionmaker successes.

It is questionable whether this perception of reality applies throughout the intelligence discipline or even within the broader national security community.  Particularly on a tactical level, where the intelligence professional often shares the same foxhole, as it were, with the decisionmaker, it becomes obvious relatively quickly how accurate and how useful the intelligence provided is to the operators.   While most intelligence professionals subscribe to the poor feedback theory, most intelligence professionals also have a story or two about how they were able to give analysis to decisionmakers and how that analysis made a real difference, a difference willingly acknowledged by that decisionmaker.  The key to this kind of feedback seems less related to the issue or to intelligence writ large and more related to how closely tied are the intelligence and decisionmaking functions.  The more distance between the two, the less feedback, unsurprisingly, there is likely to be.

The third theme, is that from the requirement also emerges a mental model in the mind of the intelligence professional regarding the kinds of information that the he or she needs in order to address the requirement.  This model, whether implicit or explicit, emerges as the intelligence professional thinks about how best to answer the question and is constructed in the mind of the intelligence professional based on previous knowledge and the professional’s understanding of the question.  

This mental model typically contains at least two kinds of information; information already known and information that needs to be gathered.  Analysts rarely start with a completely blank slate.  In fact, Phillip Tetlock has demonstrated that a relatively high level of general knowledge about the world significantly improves forecasting accuracy across any domain of knowledge, even highly specialized ones (Counter-intuitively, he also offers good evidence to suggest that high degrees of specialized knowledge, even within the domain under investigation does not add significantly to forecasting accuracy). 

The mental model is more than just an outline, however.  It is where biases and mental shortcuts are most likely to impact the analysis.  It is where divergent thinking strategies are most likely to benefit and where their opposites, convergent thinking strategies such as grouping, prioritizing and filtering, need to be most carefully applied.  One of the true benefits of this model over the traditional Intelligence Cycle is that it explicitly includes humans in the loop - both what they do well and what they don't.

Almost as soon as the requirement gains enough form to be answerable, however, and even if it continues to be modified as a result of an exchange or series of exchanges between the decisionmakers and the intelligence professionals, four processes, operating in parallel, start to take hold: The modeling process we just discussed, collection (in a broad sense) of additional relevant information, analysis of that information with the requirement in mind and early ideas about production (i.e. how the final product will look, feel and be disseminated in order to facilitate communicating the results to the decisionmaker).

The notional graphic below visualizes the relationship between these four factors over the life of an intelligence product.  Such a product might have a short suspense (or due date) as in the case of a crisis or a lengthier timeline, as in the case of most strategic reports, but the fundamental relationship between the four functions will remain the same.  All four begin almost immediately but, through the course of the project, the amount of time spent focused on each function will change, with each function dominating the overall process at some point.  The key, however, is that these four major functions operate in parallel rather than in sequence, with each factor informing and influencing the other three at any given point in the process.

A good example of how these four functions interrelate is your own internal dialogue when someone asks you a question.  Understanding the question is clearly the first part followed almost immediately by a usually unconscious realization of what it would take to answer the question along with a basic understanding of the form that answer needs to take.  You might recall information from memory but you also realize that there are certain facts you might need to check out before you answer the question.  If the question is more than a simple fact–based question, you would probably have to do at least some type of analysis before framing the answer in a form that would most effectively communicate your thoughts to the person asking the question.  You would likely speak differently to a child than you would to an adult, for example, and, if the question pertained to a sport, you would likely answer the question differently when speaking with a rabid fan than to a foreigner who knew nothing about that particular sport.

This model of the process concludes then where it started, back with the relationship between the decisionmaker, the intelligence professional and the goals and purposes of the organization.  The question here is not requirements, however, but feedback.  The intelligence products the intelligence unit produced were, ultimately, either useful or not.  The feedback that results from the execution of the intelligence process will impact, in many ways, the types of requirements put to the intelligence unit in the future, the methods and processes the unit will use to address those requirements and the way in which the decisionmaker will view future products.

This model envisions the intelligence process as one where everything, to one degree or another, is happening at once.  It starts with the primacy of the relationship between the intelligence professional and the decisionmakers those professionals support.  It broadens and redefines, however, those few generally agreed upon functions of the intelligence cycle but sees them as operating in parallel with each taking precedence in more or less predictable ways throughout the process.  This model, however, explicitly adds the creation and refinement of the mental model of the requirement created by the intelligence unit as an essential part of the process.  This combined approach captures the best of the old and new ways of thinking about the process of intelligence.  Does it, however, test well against the reality of intelligence as it is performed on real-world intelligence problems?

Part Three - Testing The Mercyhurst Model Against The Real World