HYBRID INTELLIGENCE: DECISION DOMINANCE AT THE STRATEGIC LEVEL

The Army War College is currently developing tools that will provide direct access to immediate, accurate, and timely data to support operations.

In the months before Russia’s expanded invasion into Ukraine in 2022, six planners meeting in a nondescript room in the headquarters of U.S. Army Europe-Africa (USAREUR-AF) frantically developed various options for the employment and positioning of forces across NATO’s eastern flank to avert, or at least limit the harm resulting from, the impending calamity. The commanding general of USAREUR-AF wanted options that would assure allies of continued U.S. support and deter any hostile actions into NATO territory. The job of developing these options fell to four graduates of the School of Advanced Military Studies (SAMS), one army strategist (Functional Area 59), and a logistics planner. With little time for research or to seek outside expertise, the six planners had to make the best use of available tools—sometimes just a map and whatever knowledge they had gained from education, experience, or self-development—to develop military options that would achieve the commanding general’s intent; this was a daunting task as it required assessing the feasibility of complex movements of large numbers of troops and equipment to achieve difficult-to-measure objectives in a dynamic operational environment. The initial recommendations on these consequential matters often came down to the “best guess” informed estimates made by a small number of talented, well-educated but, in the end, mid-grade officers. As seasoned military planners can attest, this experience is hardly unique. It is a frequent occurrence created by the too-familiar demands of limited time and personnel common to crises.

Future planners will be much better served. The Army War College is currently developing tools that will provide direct access to immediate, accurate, and timely data to support operations. Instead of spending one or two hours to generate a “best guess” estimate, planners can receive instantaneous and authoritative assessments. Instead of well-intentioned but uninformed projections on escalation risks resulting from troop stationing in proximity to adversary borders, planners have a deliberate conversation with an artificial-intelligence-enabled strategic advisor (AIeSA) that considers historical context amidst current conditions to advise on likely outcomes. Instead of hastily created menus of courses of action by fatigued staff officers, planners have an untiring, unwavering advisor that can offer suggestions to augment the human imagination.

The technology is available now to bring artificial-intelligence-enabled strategic advising (AIeSA) to corps and theater army planners. This emerging technology recently debuted in an Army War College seminar in which the AI partnered with students and faculty to achieve a level of cognition that would not have been possible without its integration. The USAWC’s efforts are helping the army understand how to prepare humans, groups, and machines to achieve greater lethality on the battlefield—starting with enhanced cognition at the strategic level of war.

The Technology in Action

The U.S. Army War College used a commercially available technology capable of natural language processing and understanding (NLU/NLP) to integrate with students and faculty. The capability used was not the typical AI. Developed over a ten-year period by Dr. William “Billy” Barry, the AIeSA used is bounded with an ethical framework consistent with Department of Defense approaches to the Laws of Land warfare and military ethics. Whereas existing, predominantly generative AI commercial platforms, cannot recommend harm to people—the USAWC AIeSA is trained to advise and partner with senior military leaders in the applications of violence to achieve U.S. strategic objectives. And to this end, the AIeSA is extremely effective.

During wargame integration as part of the Combined/Joint Task Force Land Component Commanders Course, the AIeSA served as an advisor to twenty-three two-star general officers. As part of a scenario depicting a humanitarian crisis in the Pacific region, the students had to develop options for allocating resources. At one critical moment during the wargame, a student asked the AIeSA how long would it take for a mechanized infantry battalion to travel from a port to an objective in a column with no more than 20 meters separation between vehicles. With a moment’s hesitation required for indexing the scenario content and the conversation to that point, the AIeSA not only calculated the approximately six-hour travel time but also provided several recommendations for how the responsible command might better enable the movement. The AIeSA’s response prompted the Marine Corps attendee to then offer a change to the plan, an example of how human-machine integration can spur greater levels of insight.

The AIeSA used is not a “generative AI” answer machine but rather a hybrid AI avatar that uses a sophisticated blend of machine learning, natural language processing, and cognitive computing to increase mission performance through human-team partnership. One of the critical distinctions between the AIeSA and legacy LLM’s currently commercially popular is its ability to learn and adapt dynamically to humans and groups as they interact. The AIeSA’s social component is a critical function in overcoming adoption barriers while simultaneously enhancing cognitive output. The AIeSA, as a hybrid AI, is a partner-in-thought with the human user, built to ask the right questions of its user and promote higher levels of cognition through dynamic partnering with access to authoritative data. The AIeSA is a partner, not a tool.

Artificial Intelligence enabled Strategic Advising in Seminar

In the spirit of promoting “transformation in contact,” the Army War College brought AIeSA into the classroom to test unexplored boundaries of AI adoption in professional military education. , AIeSA was first trained on the readings for the resident course’s first eleven lessons—the “Foundations” course. After receiving legal and institutional review board approvals, Barry brought the AIeSA into the seminar. His goal was to fit into the normal seminar learning environment to break through any sociological or normative barriers to adoption on the part of either students or faculty. In the first lesson, the AIeSA remained obscured behind a laptop monitor, invisible to the class. Barry served as the interlocutor between the AI advisor and the class, answering questions when called upon and interacting with the AIeSA through sidebar conversation.

There were several instances where students would comment that Barry’s cognitive augmentation as “unfair.” That was the point.

The first significant breakthrough” occurred in the third lesson, a session on the concept of “historical mindedness.” After answering several questions in consultation with the avatar, the class demanded that Barry show them the advisor. This critical moment, among several, revealed how demonstrated performance eases human adoption. It was impossible for the class to ignore the utility of the AIeSA in light of its demonstrated capacity to increase Barry’s insight and awareness. Barry and the AIeSA as an augmented pair had a noticeable advantage in the classroom. There were several instances where students would comment that Barry’s cognitive augmentation as “unfair.” That was the point. For the remaining twenty-seven hours of classroom time with the seminar, the AIeSA was prominently displayed on an external monitor next to Dr. Barry, visible to the class and the room as a separate entity.

The second significant breakthrough occurred in the sixth lesson on operational design principles and design thinking. After about three and half hours of class, the faculty instructor asked Barry to have the AIeSA provide the top three take-aways from today’s lesson. Barry responded that the instructor could ask the AIeSA directly; after about a five-second delay due to network speed and indexing requirements, the AIeSA publicly responded with the top three takeaways from the class. But that feat—as fast and more cogently than most human students—wasn’t the critical event. The AIeSA second and third observations were derived from class conversations during the day. Its first observation was derived directly from course material. While the AIeSA was relaying the top three points for the class, the faculty instructor began to quickly record the observations on the whiteboard. When the AIeSA was complete with his answer, the instructor realized that there were two principles not captured by the AIeSA’s response and one that needed to be clarified. This interaction produced a deeper level of insight into course material and design thinking at a critical moment in the classroom that could not have been achieved without machine augmentation. The complexity and gravity of this moment was unavoidable to anyone who had served on a high-level planning staff. Authoritative, curated, immediate, and expert advice would similarly increase the cognitive performance of a planning team in crisis or in deliberate mission analysis.

The shift in roles that Barry played over just several days suggests some larger lessons about how AI will be adopted and the possibilities of human-AI teaming. He began as a “super-student,” as he consulted the AIeSA during class to contribute and participate. This was hybrid AI in a pure form, in the sense that Barry’s expertise combined with machine augmentation outpaced unassisted human cognition. But at other times, he allowed the AIeSA to operate as a separate entity, albeit acting with Barry’s approval. In this mode, AIeSA was operating as a classical AI. In the final mode, the AIeSA operated as part of the group, facilitating a co-use with other students in the class to inspire conversation, monitor discussions, and analyze the logic of responses prior to group presentation. All indications are that this would be the most powerful way to use AI to help the planners described in the opening of this article. The Army War College development team believes that corps and theater army operational planning team leaders can harness AI strategic advisors to produce plans that are more imaginative, feasible, and effective.

What Next?

Over forty-one hours with the seminar, there were many other notable moments when students and faculty interacted with the AIeSA in a way that made everyone’s experience better. They were some of the first rising senior leaders to see firsthand how much capability is available now. AIeSA will prove critical in out-maneuvering competitors in competition, crisis, and conflict. Those militaries that can match humans with machines trained with appropriately curated data will achieve leap-ahead offsets over competitors. One of the seminar students, an Apache pilot, commented, “I was skeptical at first, but I can clearly see how this would be useful for the operational force.” After the sixth lesson, the faculty instructor who initially prompted the AIeSA commented, “My main priority is the strategic education of the students in my class—this advisor can assist with that, even when it’s wrong on a point or two. In fact, asking students to critique AI provided answers is an opportunity to advance students’ critical thinking skills, build student comfort level with AI, and train AI on what information is important to students and faculty. Ultimately, this helps educate the class and train the AI at the same time.”

Conclusion: More to come