Enhancing the Intelligence Community’s Exploitation and Dissemination Capacity as an Offset to Information Overload
Understanding AI’s Ability to Address Information Overload
An important first step towards understanding the potential utility and applicability of AI throughout the IC is providing a useful definition and explanation of AI and its related technologies. AI is a field of computer science dedicated to the creation of intelligent programs and machines—usually computer systems—that operate according to human-like principles. These machines are created to conduct operations at greater speeds and utilize more complicated thought processes than naturally available to humans. To develop common sense reasoning and problem-solving power in machines, researchers embed knowledge engineering programs within the machine to allow for objects, categories, properties and their complex data relationships to be understood in real-time. These concepts can extend to machine learning, a key sub-field of AI that is particularly important in the discussion of intelligence analysis and information dissemination.
It is important to highlight that the core benefits of AI and machine learning technologies relate to the speed at which they conduct analysis combined with the capacity to recognize complex patterns in enormous datasets. Machine-machine learning algorithms are capable of labeling, classifying and identifying objective patterns of information while other AI technologies—particularly dimension reduction, anomaly detection and artificial neural networks—are capable of explaining the impact or context such patterns may indicate.  When discussing notions of information overload or the over-collection of source material, the combination of AI and machine learning technologies allow the IC analyst or decision-maker to consistently focus on broad issues as opposed to dedicating the majority of their efforts towards information management. Not only can machine-machine pairing efforts and AI programs create dynamic and objective analysis of vast source material, but it can also perform these tasks at speeds well beyond the capacity of individual and teams of analysts.
Leveraging AI to Enhance IC Exploitation and Dissemination Practices
To take advantage of the vast sensory and information collection apparatus the U.S. has positioned throughout the world, the IC is challenged with analyzing all of the data flowing into each agency at any given moment. Further to this, accurately relaying certain raw data and finished reports to other agencies, military commanders, and policymakers add another layer of demand that traditional dissemination models cannot support. To contextualize the vastness of the data susceptible to U.S. collection capabilities, it is worth noting a study by the EMC Corporation in 2011. The study found that the amount of data stored on Earth doubles every two years, meaning the amount of data that will be created over the next 24 months will be greater than the all the data ever produced in history.
Dawn Meyerriecks, the Central Intelligence Agency’s (CIA) current deputy director for science and technology, stated in October of 2017 that the CIA at the time had 137 pilot projects directly related to AI. Meyerriecks noted that these projects ranged from, “automatically tagging objects in live full-motion video to better predicting future events based on big data and correlational evidence.” This highlights how AI technologies are not a new phenomenon in the IC, an understanding that is reinforced after considering the CIA’s Scientific and Technical Intelligence Committee’s report in 1993 titled, “Preparing U.S. Intelligence for the Information Age: Coping With the Information Overload.” Although the discourse surrounding the overload issue is not itself revolutionary, recent advancements in AI and machine learning technologies have transformed the IC’s capacity to leverage big data and vast information collection at a scale never fully recognized in early AI discussions.
A primary model of new transformative AI capability being researched and deployed is Activity-Based Intelligence (ABI), including related models such as Object Based Production (OBP) and Structured Observation Management (SOM). These approaches to GEOINT significantly reduce the data management role analysts at NGA, NRO or other agencies face on a given assignment, thereby improving the speed of the exploitation process. For example, ABI emphasizes the development and understanding of patterns of life, which enable analysts to differentiate abnormal from normal activities as well as identifying new trends in the scope of what is considered normal. This is a direct application of supervised machine learning algorithms processing and then categorizing data inputs based on previously analyzed information. If a persistent tactical ISR platform over a small city in Iran routinely tracked ten people entering a suspicious facility every day for a month, supervised machine programs would signal any instances of more or less than ten people entering that facility. However, if over the next month eight people routinely entered the facility, unsupervised machine programs would recognize a pattern change. These programs are all operating within the context of a specialized AI architecture, which would then be able to extract the new pattern and implement it as the updated guiding template for the supervised machine program classifying all new data from the ISR platform. This entire ABI process can occur with the analyst only providing an oversight and review role, demonstrating one capability of advanced AI that agencies can use to effectively exploit information and unburden the analyst from labor intensive data analysis.
Strategic Advantages of Increasing AI Capabilities in the IC
Understanding how AI can significantly enhance the exploitation speed of incoming data and information in addition to disseminating it more accurately can reveal an opportunity for completely leveraging and improving American collection capacity. From global ISR platforms extracting data from combat environments to developing analytical models to predict long-term global events, the more information capable of being collected and properly analyzed in a timely manner, the more the warfighter’s mission and the policymaker’s awareness will benefit.
A joint study in 2017 between the Belfer Center and the Intelligence Advanced Research Projects Activity (IARPA) recognized that the U.S. IC collects more raw information and data than their entire workforce could effectively analyze in their collective lifetimes. This challenge combined with the continued improvement in source collection—across platforms, sensors and domains—will result in an increasingly overwhelmed IC. During a statement before the House of Representative’s Subcommittee on Strategic Forces in May of 2017, NGA Director Robert Cardillo called for the continued growth of GEOINT collection programs being undertaken by NGA and NRO’s Geospatial Intelligence Systems Acquisition Directorate.
Similar trends are occurring with the IC’s SIGINT operations, where efforts led by the NSA are seeking to collect information from all systems and devices networked and connected to the Internet of Things (IoT). At a technology summit in 2016, former NSA Deputy Director Richard Ledgett was asked whether collecting information from billions of IoT devices would be a ‘security nightmare or a signal intelligence goldmine,’ he responded by saying, “Both.” This statement demonstrates that the growing scope and scale of collection concepts of operations (CONOPS) is currently viewed as an opportunity to better inform consumers of intelligence while at the same time is also viewed as a potential source of information overload. AI technologies are a feasible solution capable of alleviating the negative aspect of expanding collection CONOPS, meaning AI implementation across the IC could directly result in an information analysis advantages to the U.S. intelligence apparatus and its consumers.
Considering US adversaries are providing their warfighters and policymakers with war and peacetime information collected from highly advanced ISR platforms similar to American systems, it is fair to conclude that the U.S. has lost a component of its comparative advantage in global technical collection. Therefore, to ensure that American warfighters and policymakers are still equipped with superior intelligence to support their decision-making processes compared to adversarial counterparts, two general courses of action are possible: the first seeks to offset foreign technical collection capabilities with parallel U.S. improvements, while the second seeks to offset foreign capabilities with an enhanced analysis capacity.
Utilizing AI as the foundation of this analysis-based approach would induce a positive feedback loop where the expansion and the IC’s collection apparatus would in turn leverage operational AI technologies, which are reliant on large volumes of data and information inputs to be successful. Under this approach, the IC could continue drawing from a robust and diverse private industry to develop advanced collection capabilities—focused on greater speeds, greater detail, and more complex data sensors—without concern of overburdening analysts. Implementing AI systems and ensuring exploitation and dissemination remain effective and dynamic, rather than agencies simply aiming to sift through enough data and collect more information than adversaries, will make the U.S. more likely to regain a comparative advantage in global ISR mission utility.
AI augmented analysis throughout the IC can provide a long-term strategic advantage to the U.S. policymaker and warfighter, who could have access to large quantities of analyzed information foreign adversaries would technologically be unable to produce at a similar rate—relative to current capabilities. IC has heavily relied on collection superiority over foreign counterparts to support its importance within the national security apparatus; AI capabilities have now provided the IC with an opportunity to gain a new advantage that draws on enhanced exploitation and dissemination speeds to outcompete adversaries. This enhanced capacity can strengthen the IC’s significance as a tool of the policymaker and warfighter in a period of high-end strategic competition stemming from near-peer adversaries such as Russia and China.
Implications for the Human Analyst
Although AI provides a technological capacity that can outperform the human analyst in exploitation and dissemination roles, it must be understood that AI and its subset programs operate in closed systems. This means that AI and machine learning programs are suited for conditions where humans provide command, oversight and review roles, leaving the labor-intensive and complex information management duties to the more capable machine technologies. In such conditions, the human analyst must adapt to a new environment where their role is not replaced by AI programs, but elevated and refined.
As private industry and government research laboratories continue to invest in AI program advancements, it is possible that universal analysis capabilities may become conceptually and practically feasible. However, even if these technologies were to reach operational status, it would not necessarily coincide with the termination of the human analyst, largely due to the non-functional role humans support in the IC. NGA Director Robert Cardillo, during the Intelligence and National Security Summit 2017 in Washington, D.C., stated that “The bedrock of our profession is credibility, it is trust, and that means keeping human analysts involved.” At the same summit, Navy Admiral Michael Rogers, the director of the National Security Agency and head of U.S. Cyber Command, stated that “Increased use of AI does not mean agencies should abandon human intelligence analysis. I don’t see us abdicating everything to a machine. I see us asking ourselves where does [using] this technology make sense.” Consistent with Adm. Rogers’ comments, this piece has demonstrated that AI technologies can be highly effective at leveraging components of the IC’s exploitation and dissemination practice to enhance the capabilities of the analyst and improve finished products leaving the community. This would all occur under the guidance, oversight and final review of human analysts, allowing the IC to maintain its bedrock of legitimacy—trust and credibility in the human-facilitated intelligence process.
Sam Cohen is currently completing a Master of Science (M.S.) in Defense and Strategic Studies at Missouri State University's Washington D.C. Campus. He is currently working as a Security Risk Analyst with Horizon Intelligence - a geopolitical consulting and advisory firm.
Joe Zeman is currently completing a Master of Science (M.S.) in Defense and Strategic Studies at Missouri State University's Washington D.C. Campus. He is currently working as a Cyber Intelligence Analyst at LookingGlass Cyber Solutions and recently accepted a position in Lockheed Martin's Space and Missile Defense Department.
Ackerman, Robert. “Multiple Thrusts Define Geospatial Agency Big Data Efforts.” Signal Magazine: AFCEA’s International Journal. August 1, 2014. https://www.afcea.org/content/multiple-thrusts-define-geospatial-agency-big-data-efforts
Allen, Greg and Taniel Chan. “Artificial Intelligence and National Security.” Harvard Kennedy School: Belfer Center for Science and International Affairs on behalf of Intelligence Advanced Research Projects Activity. (July 2017): 1-67.
Atwood, Chandler P. “Activity-Based Intelligence: Revolutionizing Military Intelligence Analysis,” National Defense University Press: Joint Force Quarterly 77 no. 2 (2015): 24-33.
Arians, Anne. “Supercomputing: Maximizing Decision Advantage.” Pathfinder: Journal of the National Geospatial-Intelligence Agency 15 no. 2 (June 2017): 10-14.
Beaver, Matt S. “Dealing with Information Overload: CCIR as Ockham’s Razor in the Hands of the Joint Force Commander.” Naval War College: Joint Military Operations Department (May 2003): 1-20.
Brueggemann, Charles E. “Mitigating Information Overload: The Impact of ‘Context-Based Approach’ to the Design of Tools for Intelligence Analysis.” PhD. Thesis. Naval Post Graduate School. March 2008.
Cardillo, Robert. “Priorities and Posture of the National Security Space Enterprise.” House Armed Services Committee. Washington, D.C., May 19, 2017. Accessed through House Minute Archives at http://docs.house.gov/meetin gs/AS/AS29 /20170519/105974/HHRG-115-AS29-Wstate-CardilloR-20170519.pdf.
Choudhury, Saheli R. “China's Artificial Intelligence Technology Is Fast Catching Up To The US.” CNBC Technology News Division. September 1, 2017. https://www.cnbc.com/2017/09/01/goldman-says-china-has-talent-data-and-infrastructure-to-embrace-ai.html.
Cordova, Amado, Lindsay D. Millard, Lance Menthe, Robert A. Guffey and Carl Rhodes. “Motion Imagery Processing and Exploitation (MIPE).” RAND Corporation prepared for the United States Air Force no. 4 (2013): 1-36.
Crawford, David A. “Why AI Will Not Replace All Human Data Analysts.” Venture Beat. May 23, 2017. https://venturebeat.com/2017/05/23/why-ai-wont-replace-all-human-data-analysts/.
Easton, Ian “China’s Evolving Reconnaissance- Strike Capabilities: Implications for the U.S.-Japan Alliance.” Project 2049 Institute jointly with The Japan Institute of International Affairs. February 2014:1-31.
Erickson, Andrew S. “Chinese Air- and Space-Based ISR: Integrating Aerospace Combat Capabilities over the Near Seas.” In Peter Dutton, Andrew S. Erickson, and Ryan Martinson, eds., China’s Near Seas Combat Capabilities. Naval War College China Maritime Study 11. February 2014. 87-117.
Faggella, Daniel. “What is Machine Learning?” Technology Emergence. September 19, 2016. https://www.techemergence.com/what-is-machine-learning/.
Fitzgerald, Dennis D. “Risk Management and National Reconnaissance From the Cold War Up to the Global War on Terrorism.” Center For Space Nuclear Research, original (classified) published with National Reconnaissance-Journal of the Discipline and Practice no. 21 (2003): 1-20.
George, Roger Z. and James B. Bruce. Analyzing Intelligence: National Security Practitioners' Perspectives. Washington, D.C.: Georgetown University Press, 2014.
Goldstein, Phil. “Why Intelligence Agencies Are So Interested in AI.” FedTech Magazine. October 13, 2017. https://fedtechmagazine.com/article/2017/10/why-intelligence-agencies-are-so-interested-ai.
Gookins, Amanda J. “The Role of Intelligence in Policy Making.” School of Advanced International Studies Review 28 no. 1 (Washington, D.C.: Johns Hopkins University Press, 2008.): 65-73.
Granick, Jennifer S. American Spies: Modern Surveillance, Why You Should Care, and What to Do About It. Cambridge: Cambridge University Press, 2017.
Knopp, Bradley M., Sina Beaghley, Aaron Frank, Rebeca Orrie and Michael Watson. “Defining the Roles, Responsibilities, and Functions for Data Science Within the Defense Intelligence Agency.” RAND Corporation (2016): 1-96.
Lettre, Marcel. “The Military’s Third Offset Strategy and the Revolution in GEOINT.” USGIF GEOINT Symposium 2016. Orlando, FL. May 16, 2016. Accessed through Trajectory Magazine at http://trajectorymagazine.com/wcontent/uploads/USDI GEOINT.pdf.
Mass, Peter. “Inside NSA, Officials Privately Criticize ‘Collect it All’ Surveillance.” The Intercept. May 28, 2015. https://theintercept.com/2015/05/28/nsa-officials-privately-criticize-collect-it-all-surveillance/.
Murphy, Kevin P. Machine Learning: A Probabilistic Perspective. Cambridge: Massachusetts Institute of Technology Press, 2012.
National Geospatial-Intelligence Agency. Office of Corporate Communications. “Analysis Technology Vision for 2020.” NGA Public Affairs Division (August 2014): 1-12.
Nilsson, Nils J. The Quest for Artificial Intelligence: A History of Ideas and Achievements. Cambridge: Cambridge University Press, 2010.
O’Connell, Kevin M. “The Role of Science and Technology in Transforming American Intelligence.” In Allan E. Goodman and Bruce D. Berkowitz, Best Truth: Intelligence in the Information Age. Hoover Press, 2000. 139-174.
Office of the Inspector General. National Security Agency. Organizational Structure and Ongoing Operations with Metadata. ST-09-0002. Fort Meade, Maryland. 2009.
Patterson, Emily S. “Challenges in Intelligence Analysis Under Data Overload.” Ohio State University, Columbus, OH, 2007. Accessed through Cognitive Systems Engineering Laboratory at http://csel.eng.ohiostate.edu/productions/inte lligence/4_Institutes/2005/ Patterson 20 05_CPoDKickoff-Slides.pdf.
Pereira, Fernando. “Machine Learning in Natural Language Processing.” University of Pennsylvania, Philadelphia, PA, 2002. Accessed through Stanford University’s School in Logic, Language and Information at https://web.stanford.edu/group/nasslli/courses/pere ira/NASSLLI.pdf
Pomerleau, Mark. “Here is How Technology Can Help Unburden DIA Analysts.” Fifth Domain. August 4, 2016. https://www.fifthdomain.com/intel-geoint/isr/2017/08/04/heres-how-technology-can-help-unburden-dia-analysts/.
Pomerleau, Mark. “Inside Near-peer Capabilities: How Does Russia Stack Up Against the U.S.?” C4ISR NET. Mach 29, 2017. https://www.c4isrnet.com/c2-comms/2017/03/29/inside-near-peer-capabilities-how-does-russia-stack-up-against-the-u-s/.
Scagliarini, Luca. “What is Machine Learning? A definition.” Expert System Semantic Intelligence. April 2017. http://www.expertsystem.com/machine-learning-definition/.
Scientific and Technical Intelligence Committee. Central Intelligence Agency. Preparing U.S. Intelligence for the Information Age: Coping With the Information Overload. STIC-93-001. McLean, Virginia. 1993.
Shrener, Dmitri. “Artificial Intelligence and Video Mining: Audio Event Detection Using SVM.” The Dynamic Lab Zone. May 2017. http://dynadmic-lab.com/artificial-intelligence-and-video-mining-audio-event-detection-using-svm-2/.
Storm, Darlene. “NSA Interested in Exploiting Internet-connected Medical Devices, and Spying on IoT.” Computer World. June 13, 2016. https://www.computerworld.com/article/3082162/security/nsa-interested-in-exploiting-internet-connected-medical-devices-spying-on-iot.html.
Young, Alex. “Too Much Information: Ineffective Intelligence Collection.” Harvard International Review 3, no. 3 (August 2013): 1-8.
Zhang, Sarah. “China’s Artificial Intelligence Boom.” The Atlantic. February 16, 2017. https://www.theatlantic.com/technology/archive/2017/02/china-artificial-intelligence/516615/
 Kevin P. Murphy, Machine Learning: A Probabilistic Perspective (Cambridge: Massachusetts Institute of Technology Press, 2012), 1, 16-17.
 Nils J. Nilsson, The Quest for Artificial Intelligence: A History of Ideas and Achievements (Cambridge: Cambridge University Press, 2010), 439-440.
 Murphy, Machine Learning: A Probabilistic Perspective, 17-18.
 Nilsson, The Quest for Artificial Intelligence: A History of Ideas and Achievements, 92-93.
 Daniel Faggella, “What is Machine Learning?” Technology Emergence, September 19, 2016, https://www.techemergence.com/what-is-machine-learning/
 Beaver, “Dealing with Information Overload: CCIR as Ockham’s Razor in the Hands of the Joint Force Commander,” 2-3, 13.
 Greg Allen and Taniel Chan, “Artificial Intelligence and National Security,” Harvard Kennedy School: Belfer Center for Science and International Affairs on behalf of Intelligence Advanced Research Projects Activity (July 2017), 27.
 Phil Goldstein, “Why Intelligence Agencies Are So Interested in AI,” FedTech Magazine, October 13, 2017, https://fedtechmagazine.com/article/2017/10/why-intelligence-agencies-are-so-interested-ai
 Scientific and Technical Intelligence Committee, Central Intelligence Agency, Preparing U.S. Intelligence for the Information Age: Coping With the Information Overload, STIC-93-001 (McLean, Virginia, 1993), 1-2.
 Chandler P. Atwood, “Activity-Based Intelligence: Revolutionizing Military Intelligence Analysis,” National Defense University Press: Joint Force Quarterly 77 no. 2 (2015): 25-26.
 Ibid., 24, 29.
 Luca Scagliarini, “What is Machine Learning? A definition,” Expert System Semantic Intelligence, April 2017, http://www.expertsystem.com/machine-learning-definition/
 Greg Allen and Taniel Chan, “Artificial Intelligence and National Security,” 27.
 Robert Cardillo, “Priorities and Posture of the National Security Space Enterprise” (official remarks, House Armed Services Committee, Washington, D.C., May 19, 2017) Accessed through House Minute Archives at http://docs.house.gov/meetin gs/AS/AS29 /20170519/105974/HHRG-115-AS29-Wstate-CardilloR-20170519.pdf
 Darlene Storm, “NSA Interested in Exploiting Internet-connected Medical Devices, and Spying on IoT,” Computer World, June 13, 2016, https://www.computerworld.com/article/3082162/security/nsa-interested-in-exploiting-internet-connected-medical-devices-spying-on-iot.html
 Greg Allen and Taniel Chan, “Artificial Intelligence and National Security,” 32.
 Kevin M. O’Connell, “The Role of Science and Technology in Transforming American Intelligence,” in Allan E. Goodman and Bruce D. Berkowitz, Best Truth: Intelligence in the Information Age (Stanford: Hoover Press, 2000), 150.
 Phil Goldstein, “Why Intelligence Agencies Are So Interested in AI.”