A Joint Warfighting Concept for Systems Warfare

Note: This paper is an early, condensed version of a forthcoming longer, more detailed report to be published early in 2021.

The Bottom Line

• Future combat between peer and near-peer adversaries will be characterized, dominated, and decided by the collision of opposing systems of systems assembled to prosecute campaigns in wartime.
• An inherent component of these types of “systems confrontations” will be concerted system destruction attacks aimed at disrupting, disabling, and destroying the opposing system of systems.
• The U.S. Joint Force needs a new overarching Joint Warfighting Concept for systems warfare to cope with this emerging reality.
• This new concept will explain how new human-machine collaborative battle networks waging AI-enabled algorithmic operations will give the Joint Force a decided advantage in any future systems confrontation and the ability to defeat system destruction attacks.
• Demonstrating the ability of Joint Force battle networks to dominate opposing operational systems under realistic conditions, including system destruction attacks, will substantially achieve deterrence by denial.

Introduction

To maintain the Joint Forces’ competitive advantage, the 2018 National Defense Strategy (NDS) declared that it must pursue “urgent change at significant scale.”¹ Embracing the NDS, former Secretary of Defense Mark Esper remarked that “it remains [the department’s] guidepost and drives our decision-making.”² Among the implementation priorities Esper laid out to military secretaries and other top commanders in a January 27, 2020, memo was the need to develop a new Joint Warfighting Concept (JWC) “to align personnel, training and doctrine to win on any battlefield.”³ The next NDS should endorse and complete this important work.

Since the 1980s, the guiding JWC for combined arms warfare has been AirLand Battle.⁴ This concept is now well past its “sell-by date.” Future combat between peer and near-peer adversaries will be characterized, dominated, and decided by the collision of opposing systems of systems assembled to prosecute campaigns in wartime. Combined arms warfare remains a valid concept, but against a peer adversary with an advanced operational system, it has been rendered subordinate. Thus, the need for a new JWC.

The purpose of this new concept is to help guide Joint Force doctrinal and programmatic development by describing a vision for human-machine collaborative battle networks waging high-intensity algorithmic operations against an opposing system of systems. This concept focuses on employing human-machine collaborative battle networks in the 2040 timeframe to guide force development beyond the current future-year defense plan. Because the technologies for algorithmic operations are advancing so rapidly and at an ever-accelerating rate, this paper presents only the broad precepts that should shape the concept, rather than specifics.

The precepts and ideas outlined in this draft concept represent an initial step in a comprehensive, long-term, Joint development process that 1) trains the Joint Force to trust and exploit AI-enabled autonomous functionalities, allowing it to excel in human-machine collaboration and human-machine combat teaming; 2) redesigns and trains the Joint Force to be able to quickly assemble new human-machine collaborative battle networks, even while under attack; and 3) redesigns and trains the Joint Force to fight and win a systems confrontation and prevail against system destruction attacks.

The Key Operational Challenge

In the evolving era of systems warfare, global advances in artificial intelligence (AI), big data analytics, advanced computing, quantum science, AI-enabled cloud applications, 5G, and robotic and autonomous systems will provide both the Joint Force and its potential adversaries with the opportunity to enhance their systems of systems in different ways, making future confrontations between them increasingly sophisticated, intense, and lethal.⁵

How will the Joint Force maintain its conventional overmatch and achieve deterrence by denial in a future operating environment characterized by systems confrontation and system destruction attacks under conditions of rough military-technical parity in guided munitions warfare?

The Central Idea: Human-Machine Collaborative Battle Networks and Algorithmic Operations

The central idea of a new JWC for systems warfare is that Joint Force efforts should aim to achieve deterrence by denial through fielding new battle networks that operate better and faster than adversary operational systems, and ones that cannot be destroyed like the battle networks used today. Toward this end, the Joint Force will develop capabilities and platforms optimized for systems warfare, specifically human-machine collaborative battle networks, that leverage AI-enabled autonomy at scale to wage algorithmic operations at near machine speeds. This will lead to consistently better decisions faster than any adversary, giving the Joint Force both a decisive advantage in its observation-orientation-decision-action (OODA) cycle and the ability to prevail against system destruction attacks.

A good JWC must explain how the armed services are expected to exploit technology and organize themselves as a coherent warfighting force, and then how to employ that force to fight and win against a future military near peer. It is becoming increasingly clear that incremental adjustments to current concepts will be insufficient given the severity of the operational challenges confronting the Joint Force. Instead, entirely new warfighting concepts will be required to contest peer adversaries and provide the Joint Force with battlefield advantage in the next chapter of systems warfare.

This concept envisions future human-machine collaborative battle networks that capitalize on advances in artificial intelligence to exploit new applications of AI-enabled autonomy throughout its component sensor; command, control, communications, and intelligence (C3I); effects; and sustainment and regeneration grids. AI-enabled autonomy will lead to improved types of human-machine collaboration across all three levels of war (e.g., tactical, operational, and strategic) that compress the time between observation, orientation, and decision. It will also enable new forms of human-machine combat teaming that can apply effects on the battlefield more rapidly and discriminately at both the tactical and operational levels. Human-machine collaborative battle networks aim to synchronize operations in space, air, sea, undersea, ground, underground, cyberspace, and the electromagnetic spectrum in order to compel adversaries to seemingly confront the entire Joint Force—rather than individual components—operating at a much higher tempo than hitherto possible.

Human-machine collaborative battle networks waging algorithmic operations will render subordinate legacy operational systems that fail to exploit advances in artificial intelligence and AI-enabled autonomy, just as systems warfare rendered combined arms warfare subordinate at the operational level of war.

Precepts

The following precepts enable this central idea.

Precept 1: Going forward, the Joint Force must first be organized, trained, and equipped to fight and win against a near-peer battle network adept at systems warfare and system destruction operations.

The Joint Force must maintain its competencies in combined arms warfare, which will remain relevant and valuable over the timeframe of this concept. However, as a matter of urgency, it must develop new concepts, doctrine, and competencies in both systems warfare and system destruction warfare and test them in wargames and force-on-force exercises against opposing forces well trained in systems confrontation and system destruction operations.

Demonstrating the ability of our battle networks to dominate opposing battle networks under realistic field conditions and withstand network destruction attacks would substantially achieve deterrence by denial. A new JWC should be the first step on this transformational journey.

Precept 2: Joint Force human-machine collaborative battle networks must be able to consistently and reliably attack effectively first with both guided munitions and “invisible system strike” capabilities.

In guided munitions warfare, the first principle is to attack effectively first—to bring the enemy under fire before they can reciprocate.⁶ This is as true in systems and algorithmic warfare as it is in combined arms warfare. However, the object of these fires is not about the annihilation of the enemy force, but of disrupting and destroying the inner workings of the opposing system of systems. The specific targets chosen are those that, if destroyed, will allow the Joint Force to gradually gain an information and decision advantage in a systems confrontation. This will require a sophisticated understanding of adversary systems of systems, akin to the level of insight we had of our Cold War adversary, the Soviet Union. That means our knowledge of the adversary must go beyond insight into threat platforms and forces to include higher-level understanding of how the adversary intends to assemble and fight these elements as an operational system.

The ability to out-range an enemy has become far more difficult with the development of invisible system strike capabilities such as cyber, counter-AI, and electronic warfare. Cyber and cyber-enabled counter-AI weapons essentially have unlimited range. System destruction warfare uses these types of capabilities extensively. This helps explain why the Chinese have added a fifth information contestation system in their overarching system of systems and established a Strategic Support Force to wield these capabilities in systems destruction operations.

Precept 3: Joint human-machine collaborative battle networks must be prepared to weather the first salvo.

While the first principle of guided munitions warfare is to attack effectively first, Joint human-machine collaborative battle networks must account for the fact that an opposing operational system will likely have the advantage of firing first. The Joint Force can ill afford to assume a future president will initiate a preemptive attack on a nuclear-armed great power. It must therefore anticipate that the opponent will initiate hostilities, and the first salvos will consist of dense, guided munitions salvos and intensive, invisible system strike attacks. The Joint Force battle network must therefore be able to take these first punches, shake them off, and immediately shift to the counter-offensive.

Under these conditions, the battle network must therefore have early-entry “inside forces” that can operate within the effects envelope of the enemy’s operational system with relative impunity, either by operating from sanctuaries (e.g., undersea) or being protected by stealth, camouflage, cover, and concealment. Other network forces, units, elements, platforms, and systems inside the enemy effect envelope must be able to disaggregate to complicate enemy targeting and fight initially from a distributed posture, often without the benefit of mutual support. This will require a battle network with multiple redundant communication and data pathways. Even then, the Joint Force battle network must be designed and trained to operate without communications to higher headquarters during periods of network outages.

Precept 4: John Boyd’s conception of warfare—and particularly his OODA Loop Theory—is especially applicable to systems warfare and systems confrontation.

Philosophically, systems confrontation is entirely consistent with Boyd’s mind for war.⁷ One way to gain an information and decision advantage in a systems confrontation is to design battle networks and train the Joint Force to achieve a higher relative system operating tempo (SOT) in its observation-orientation-decision-action cycle.

Each step in the OODA cycle is a self-contained process. At the operational level of war, these processes encompass all the tasks and actions necessary to scan and scout the entire battlespace, including enemy, friendly, allied, and noncombatant forces and then make sense of what is happening and help the commander develop a common recognized operating picture, formulate intent, choose among courses of actions, develop plans, distill them into orders, and transmit the orders to the appropriate forces, elements, and systems of the Joint Force. These then apply the appropriate effects in accordance with the commander’s intent and plans, which restarts the cycle.

The processes are executed in the component grids of the battle network: observation by the sensor grid; sense-making (i.e., orientation) and decisioning in the C3I grid; and action in the effects grid, with all processes supported by the sustainment and regeneration grid. In this concept, the approach to gaining and maintaining an advantage in the OODA cycle at the system level is to start the smart teaming of humans and computers throughout the battle network, with each playing to its comparative strength. After playing humans alone, computers alone, and human-machine collaborative teams in chess, Grandmaster Gary Kasparov concluded that weak humans plus machines plus better processes were superior to both strong computers alone and strong humans plus machines plus inferior processes.⁸

This concept is based on the premise that the operational system with the strongest humans and machines and better (algorithmic) processes will perform better and operate at a much higher SOT than its adversary’s.

Precept 5: Adopting processes that exploit AI-enabled autonomy across the battle network is the path to achieving a higher relative system operating tempo than U.S. competitors.

Better processes will evolve through judicious decisions of how and when to team humans and machines and the appropriate division of labor and effort between them, especially as it relates to autonomy—the level of independence granted to a subordinate unit or system to execute a given task.⁹

In a system of systems, autonomy comes in two different types:

Autonomy at rest occurs virtually, in software, in things like intelligence support, planning, expert advisory, and predictive maintenance systems. Whatever its form, autonomy at rest is designed to help humans make decisions faster, better, and with greater precision and persistence. It is therefore often referred to as human-machine collaboration—in which human insight, creativity, and strategic guidance is combined with the tactical acuity of a computer.¹⁰ When humans playing with computers were able to beat both humans alone and computers alone in chess, Kasparov noted, “We [people] could concentrate on strategic planning instead of spending so much time on calculations. Human creativity was even more paramount under these conditions.”¹¹

Autonomy in motion occurs in the physical world as it empowers the far more capable robotics and autonomous systems that will be developed and fielded over the next 20 years.¹² Autonomy in motion will manifest itself in new forms of human-machine combat teaming and machine-machine combat teaming. The former will see humans and crewed platforms operating cooperatively with uninhabited platforms to accomplish complex battlefield tasks. The latter will see increasingly sophisticated collaborative weapon attacks and cooperative combat swarming of low-cost effectors.

The operational system that is better in human-machine combat teaming and machine-machine combat teaming will enjoy significant advantages over opposing systems. These include: the ability to learn; better situational awareness; higher performance; improved efficiency and effectiveness; greater flexibility; increased tempo by operating close to machine speeds; and easier distributed and dispersed operations.¹³

The measured exploitation of autonomy throughout future Joint battle networks will be the surest path toward achieving better battle network performance, a higher relative SOT, and a lasting information and decision-making advantage in any systems confrontation.

Precept 6: Autonomous functionalities throughout the Joint Force battle network will be foundational to prevailing against system destruction attacks.

System destruction operations are an inherent part of systems confrontation. The Joint Force must therefore “armor” its battle networks against system destruction attacks. In this regard, the autonomous functionalities so central to providing an information and decision advantage will also provide a measure of hardening and self-healing against attacks on joint battle networks.

The idea of building a self-healing system is inherent in the Joint Force’s embrace of command autonomy. Command autonomy at the Joint Force system-of-systems level is not a technology; it is a command-and-control approach that seeks to push power and decision-making to the warfighting edge.¹⁴ When coupled with an understood common recognized operational picture and commander’s intent, a battle network built on the idea and practice of command autonomy provides the means for every individual element in a system of systems to synchronize their operations in a bottom-up manner, even without communications to a command authority. A system-wide ability to do this will allow the Joint battle network to “fight through” adversary attacks on friendly data and communications links.

Under these conditions, when and where to grant delegated autonomy becomes an important battle network choice. Delegated autonomy occurs when a higher-level commander authorizes selected system elements—either human or machine, or a combination thereof—to exercise self-government and self-directed behavior. This involves granting them the freedom to develop, select, and execute a course of action required to achieve a high-level mission objective, even when communications between lower and higher command echelons is interrupted.

Over the course of the next 20 years, delegated machine autonomy will become more common. It is the level of independence, self-governance, and self-directed behavior that human commanders and operators grant an intelligent machine or system to execute a given task. It is the condition or quality of being self-governing to achieve an assigned task based on the machine’s own situational awareness (integrated sensing, perceiving, analyzing capabilities), as well as the means to generate and select among different courses of action.¹⁵

Pursuing widespread autonomy in future human-machine collaborative battle networks will achieve both an information and decision advantage and a stronger ability to withstand system destruction attacks. Indeed, they are two sides of the same coin. For example, as the Defense Science Board explains, when two sides are vying for control of the electromagnetic spectrum:

The opportunity presented by automating sensor, communications, and jamming coordination within the environment is to protect the ability to achieve information dominance while imposing high “cost” and disruption on adversaries.¹⁶

Embracing and adopting processes that exploit AI-enabled autonomy is a natural step for a Joint Force that values and celebrates initiative exercised by lower-level commanders and operators. Indeed, encouraging initiative through an emphasis on autonomy may provide a decided advantage when confronting an operational system designed by an authoritarian adversary.

Precept 7: At some tipping point, as more and more AI-enabled autonomous applications are injected into all four Joint Force battle network grids, a new type of human-machine collaborative battle network will emerge.

The nation that first masters human-machine collaboration and teaming across the levels of war will have a multi-decade advantage in battle. A collaborative human-machine system of systems will be able to demonstrate:

• More rapid sensing of the battlespace;
• More rapid sense-making of high volume, heterogeneous data;
• More rapid understanding of the operational environment and enemy plans;
• More rapid problem solving to deal with fog, friction, and chance—the permanent features of war;
• More rapid development of a common Joint recognized operational picture, shared more quickly throughout the force;
• More rapid development of commander’s intent, relevant courses of action and plans;
• More rapid, more relevant decisions, promulgated faster to manned, unmanned, and human-machine combat teams and combat organizations, able to apply more discriminate effects in a synchronized fashion across the breadth and depth of the battlespace.

Human-machine collaborative battle networks thus provide the operational architecture for the American way of algorithmic operations.

Precept 8: Fully embracing the tenets of algorithmic operations will require a rethinking of Joint Force development.

The emerging era of systems confrontation and system destruction operations will require new conceptions of algorithmic operations and battle network command and control; new ways to develop Joint Force capabilities; and new organizational constructs. The cumulative end result of all these measures is high tempo operations that are less focused on annihilating the opposing force and more about causing an adversary to collapse psychologically and their entire operational system to fail.

The psychological aspect of algorithmic operations should not be underestimated. As explained by Boyd, an enemy’s conception of a system confrontation may be overwhelmed in two key ways:

Generate a rapidly changing environment, quick clear observations, fast tempo, fast transient, quick kill; or you can turn it around the other way, to inhibit an adversary’s capacity to adapt to such environment [sic]… suppress or distort his observation by suppressing or distorting your signatures. … Always try to remain somewhat inconspicuous, at least more inconspicuous than he is ... unstructure [sic] your adversary’s system into a hodge-podge of confusion and disorder [thus] causing him to… under- or overreact to your activity, which appears uncertain, ambiguous, and chaotic to him.¹⁷

Suppressing and distorting battle network signatures calls to mind stealth and camouflage, cover, concealment, and deception. While these will all remain important in algorithmic operations, AI will transform the hider-finder game, rendering hiding harder and finding easier. Moreover, “finding” here includes the prompt ability to destroy whatever is found. Thus, suppressing and distorting signatures must be aided by a new “all-encompassing ‘force design’ targeted at blocking China’s strategy of hitting critical U.S. information nodes [e.g., systems destruction warfare].”¹⁸

Such a force design is the subject of the Defense Advanced Research Projects Agency’s (DARPA) Mosaic Warfare program. The JWC should build upon it.

As described by Timothy Grayson, director of DARPA’s Strategic Technologies Office (STO), Mosaic Warfare:

[I]s a “system of systems” approach to warfighting designed around compatible “tiles” of capabilities (think functions such as sensors and shooters), rather than uniquely shaped “puzzle pieces” (think platforms) that must be fitted into a specific slot in a battle plan in order for it to work.

Instead of exquisite platforms like the F-35 fighter, exquisite functional technology nodes (such as an advanced infrared sensor) that can be mixed and matched via AI-enhanced networks. [The] battle plan doesn’t exist until a field commander builds it, based on whichever capabilities are available in real time — like a kid building a LEGO spaceship not from a kit and a blueprint, but from a drawer of jumbled pieces that nonetheless all fit together.

Key to the concept … is flexibility. What we are trying to do … is to enable the ability to create what I’ll call a warfighting architecture, which would be a force package … [and] to be able to defer the decision as to what that warfighting architecture is going to be to as close as possible to time of need.¹⁹

Such an infinitely malleable architecture like this would help suppress or distort an adversary’s picture of friendly battle network operations, particularly in how the Joint Force human-machine collaborative battle network applies effects. Accordingly, the Defense Department’s defense program and future force development should be modified to better support systems warfare and thinking and the associated development of Joint Force capabilities. For example, the Office of the Secretary of Defense could stand up cross functional teams (CFTs) for each of the Joint Force battle network grids, each with a responsible leader. These teams would be tasked to describe grid processes and how they could be incrementally but steadily improved. In doing so, they would recommend the highest payoff exploitation of AI-enabled functionalities and autonomy in each grid. For example,

• The CFT for the sensor grid should encourage the insertion of machine-learning algorithms into as many Joint Force sensors as practical. This would allow the sensors themselves to look for specific objects and targets and download their findings directly to effectors without the need to send the data to a central processing node.
  
• Carl von Clausewitz described the genius for war as “coup d’oeil,” the ability to take in the entirety of the battlefield at a glance, determine the decisive point, and resolve to commit one’s forces against it.²⁰ Today, however, the battlefield encompasses an entire theater of operations, surveilled by thousands of multi-phenomenology sensors providing far too much data and information for any human commander to process and comprehend. The solution is to have machines process and learn from the data, to discover patterns unseen to the human eye and mind, and to make inferences and predictions from them. Properly designed, human-centric, human-machine collaboration will certainly help extend human cognition by providing “eyeglasses for the mind.” In other words, it will provide the means for a capable commander to take in the entirety of the data and information gleaned by the sensor grid, determine adversary plans and the decisive points in an adversary’s system of systems, and commit their forces against them. Accordingly, the CFT for the C3I grid should encourage, develop, and field a variety of AI-enabled analyst and commander decision-support tools. Human-machine collaboration will be central to the processes of the C3I grid.
  
• The CFT for the effect grid should champion all types of human-machine and machine-machine combat teaming in every operating domain and pursue those that promise a decided combat advantage.
  
• The CFT for the sustainment and regeneration grid would have the dual responsibility of developing capabilities to sustain and rapidly regenerate battle network forces under persistent attack, as well as sustaining and regenerating the network damaged by system destruction operations.
  
• An option would be to limit the remit of the sustainment and regeneration grid CFT to the capabilities necessary to sustain the force while under intense attack (e.g., predictive readiness; predictive maintenance; predictive resupply; unmanned delivery of supplies, food, and ammunition) and stand up a fifth CFT for an information contestation grid, as the Chinese have done. This CFT would focus on the joint force capabilities needed to wage system destruction operations, as well as those needed to regenerate damage to the joint force system of systems.
  

The work of all CFTs would be overseen by a systems architect for human-machine collaborative battle networks, whose job would be to establish common battle network communications and data standards and oversee the development of complementary network capabilities needed for systems warfare. Importantly, the system architect would develop a digital twin of notional collaborative battle networks that could be used to help develop battle network capabilities and new tactics, techniques, and procedures. The system architect, in close coordination with DARPA, would also use the digital twin to demonstrate new human-machine collaborative battle networks operating under realistic operating conditions and their ability to withstand system destruction warfare by 2025. These demonstrations would be followed by real-world exercises against capable opposing forces. Demonstrating the ability of our battle networks “to take a licking and keep on ticking” would substantially achieve deterrence by denial.

About the Author

Robert O. Work is the Distinguished Senior Fellow for Defense and National Security at the Center for a New American Security. He was the 32nd Deputy Secretary of Defense.

Learn More

From July to December 2020, CNAS will release new papers every week on the tough issues the next NDS should tackle. The goal of this project is to provide intellectual capital to the drafters of the 2022 NDS, focusing specifically on unfinished business from the past several defense strategies and areas where change is necessary but difficult.