Build Human-Machine Dream Teams

IBM made history in 1997 when its computer Deep Blue won a chess match against the reigning human world champion. Since that time, human-machine pairings have become the best chess-playing teams in the world. In freestyle chess, players may use any aid, including computers, to conduct moves in a limited time. Thomas H. Davenport and Julia Kirby in their book, Only Humans Need Apply: Winners and Losers in the Age of Smart Machines (Harper 2016), catalog how winning teams used multiple computer tools, applying moves from each as the development of the matches created situations where the relative strengths of each automated system offered advantages.

To form dream teams, the human in a human-machine partnership must step up to the next cognitive level. AI will discern patterns, but it will be poorly equipped to understand and articulate the adversary’s mind-set in dialog with the commander. At the highest level, the most difficult and sophisticated questions from a commander are about adversary actions in cases the adversary has not yet contemplated. The question “what would they do if . . . ?” cannot be collected. There is no document or computer file where our opposite number has considered the new and novel. The answer must be built on an extended examination of the target; a cognitive, deep penetration that yields what one commander called “expert intuition.” Developing this ability requires that analytic specialization on the adversary target be deliberately cultivated within the ranks.

To develop and effectively apply analytic tools, naval intelligence professionals must understand the intellectual elements of analysis. By tradition, naval intelligence analysis is taught “on the job.” Individuals with a “knack” for intelligence—able to recognize and articulate patterns in an OPINTEL environment—are rewarded, but formal skill development is not a priority. To take advantage of AI, the intelligence community should take a page from the Central Intelligence Agency, which benefits from a culture that recognizes analysis as a learned skill that must be cultivated deliberately over a career.

The Navy’s ability to develop naval intelligence analysts and the tools to support them is impeded by the lack of a program of record (POR) for analysis. PORs mean money—and in the dynamic of the Chief of Naval Operations (OpNav) staff—money means relevance and influence. Absent senior-level focus, analysis is an assumed capability, relied on but rarely discussed, and never measured or formally developed. Automated analytic systems, however, are tangible, fundable developments, and they have contractors and government developers who will advocate for them. This advocacy focuses on the material solution first, when the answer starts on the human side of the equation.

Currently, the Advanced Maritime OPINTEL Course (AMOC) is the only dedicated Navy course in OPINTEL analysis techniques. The “advanced” moniker, however, is more a tradition than fact. In the past, most students arrived at the AMOC having completed a tour focused on maritime analysis. Today, students arrive lacking the required knowledge, forcing the course to focus on the basics of the craft. The AMOC must upgrade its curriculum to produce advanced analysts, and that means the intelligence community must assign students who meet prerequisites prior to the course.

Naval intelligence should also embrace the Naval Information Warfighting Development Center to build experts in these tools and to teach analysis teams how to use what is available at an advanced level. Complementing the established training process, these naval intelligence weapons tactics instructors could bring best practices with existing advanced tools to units during their workups.

Problematic fleet experience with the Distributed Common Ground Station–Navy (DCGS-N) underscores that warfighting needs must be central to the creation of advanced analytic systems. Building the right tools requires sophisticated engagement with the development and acquisition process that naval intelligence has not been well positioned to provide. By only supporting the acquisition process, it misses opportunities.

As a long-term investment, naval intelligence should send several officers each year to the Naval Postgraduate School Information Sciences curriculum. These graduates can become the link between warfighting intelligence requirements and the AI development process. Each Maritime Intelligence Operations Center (MIOC) should have at least one graduate of the program. The MIOCs are the backbone of the Navy’s OPINTEL structure, focusing on the same adversary over time and generally enjoying the most robust information systems. An innovative officer or sailor with a background in advanced data analytics would be a powerful enabler of OPINTEL tool development and the integration of advanced analytics.

Placing empowered technical expertise with the fleet user emulates a model used successfully in the Joint Special Operations community. System experts and developers, sitting side-by-side with the analysts and operators, provide invasive support—identifying, adapting, and developing tools to answer immediate needs. As computing power becomes ubiquitous and sophisticated tools become standard software loads, better solutions do not need to be a POR. The Pacific Fleet MIOC, for example, developed an in-house tool to allow collaborative management of Pacific maritime contacts. The federation order-of-battle management tool (FOM) was produced by a talented lieutenant using a core of existing data management engines. Used by each node within the Pacific Fleet Intelligence Federation, the FOM automates many routine tasks—even producing standardized briefing slides. While it takes time, effort, and expertise, this kind of innovation is a game-changer.

Handling big data in an OPINTEL setting differs from the big data problems faced in industry. Our adversaries work actively to deny us information and to insert false or misleading information into our analysis.  The adversary defines normal and can chose to alter it dynamically or shift operations without warning to deny us data input. While AI tools can be made sensitive to sudden anomalies in the data, the human side of the team must be trained to consider deliberate adversary deception as a routine possibility.

It is essential that the Navy construct tools that enable decentralized operations when communications links are degraded or severed by enemy action. The vision of a Navy tactical cloud can be augmented by national and theater data with analytics conducted locally when communications are limited. Dependency on exquisite communications and centralized reach-back support ignores the reality that contested maritime superiority will be accompanied by intense electromagnetic and cyber conflict. The machine part of the team must degrade gracefully in the face of adversary non-kinetic fires, and the human part must be trained and prepared to manage that degradation.

Years ago, Admiral Bill Studeman, former director of the National Security Agency and deputy director of the Central Intelligence Agency, summarized the imperatives of the naval intelligence profession as “deep penetration of the adversary” and “deep penetration of the customer.” The job is to uncover, understand, and deliver combat-essential insights our commanders need to be victorious in battle. Today, advances in technology enable the pairing of human conceptual tools of analysis with AI-enabled machines to deliver unparalleled knowledge of the enemy and the environment. Chief of Naval Operations Admiral John Richardson has observed that in naval combat today, “the margins of victory are razor thin.” That margin will hinge on strengthening the naval intelligence profession and its machine partners.

Captain Rielage serves as director for Intelligence and Information Operations for U.S. Pacific Fleet. He is a frequent contributor to Proceedings .

This article appeared originally at the U.S. Naval Institute's Proceedings Magazine.