This blog post discusses some important issues surrounding strategic management in the AI age and Multi-Domain Operations (MDO). The intention is to spark discussion that – hopefully – can lead to new insights and further studies. The examples are primarily taken from the maritime domain but should in principle be general.
Network Centric Warfare
In 1997, Captain Ross Mitchell, USN, published the article “Naval Fire Support: Ring of Fire” in the US Naval Institute. Proceedings. Here he describes how naval combat should be coordinated in networks. But the concept of Network Centric Warfare was only introduced the following year in Admiral Arthur K Cebrowski and John Garstka's article "Network-Centric Warfare: Its Origin and Future", in Proceedings, January 1998.
From a Swedish perspective, these articles led to the Network-Based Defense (NBF) project. We draw a veil of mercy over this project, which was too ambitious for the time. Some enthusiasts also drew too far-reaching conclusions from the outlined technical possibilities. However, networked defense is a given today and forms the basis for command in multi-domain operations (MDO).
The idea of coordinating the combat of combat forces is of course much older. In the era of sailing ships, various signaling systems were developed for this purpose. One of the first was Bigot de Morogues Tactique navale or Traité des Évolutions et des Signals [Naval Tactics or Treatise on Evolutions and Signals], from 1763.
When the author of the article was a young naval officer, command was sequential: signal messages with reconnaissance reports were entered into the plot; when the picture became clear, the unit commander made decisions and sent orders with signal messages to the subordinate ships. From the 70s, decisions could be sent digitally (MADAK), but it was still a matter of sequential command.
Today, information gathering, command and communication (C2) occur in parallel. In the “cloud”, command and all information are collected (in theory, anyway) and can be accessed by all units. And the “transparent battlefield” creates a lot of information; not least when we transition to MDO because the “battlefield” includes all domains: land, sea, air, space, cyberspace, the information domain, the cognitive domain and the electromagnetic domain. [1] With command information in the “cloud,” units can synchronize themselves and thus become more agile. That was how Network Centric Warfare was intended from the beginning. But does it work? How do you create a gathering of forces? The idea is interesting in any case.
However, a lot of information is not the same as a lot of knowledge and high understanding. This is for many reasons. Information, especially more aggregated information, can be contradictory. Small differences in situational or time perception can distort information data. The amount of information is overwhelming and requires technical systems to be able to be used. In addition, the information domain itself is a battlefield.
AI and Big Data are of course important tools in this context. But these systems are built around algorithms that the decision maker is likely not familiar with and which may not be built at all for the systems and needs he/she has. Furthermore, it is difficult to train AI regarding situations that have never occurred.
Several domains interact in a complex way. This is particularly clear in the information domain; information is a broad concept. One basis for classification is the source; for example, HUMINT (human sources) and SIGINT (information from signals systems) as well as other types of -INT. Another is level: strategic, operational, tactical and combat engineering. A third concerns the degree of processing: data, information, knowledge and insight. To which is added reliability assessment.
The flow of information can be intercepted, disrupted or distorted. Cyber warfare can affect in all three ways. Cyber warfare can use the electromagnetic domain for its purposes while the latter is its own battlefield; especially at the combat technical level. The cognitive domain focuses on the ability to understand and interpret information and to make decisions; it can be immediate (understanding of a certain situation) or long-term (attitudes and values).
History shows that the big problem is often not the information itself, but that the decision-maker does not accept the information or reinterprets it for ideological or other reasons; he/she refuses to accept the truth. Two examples. One reason why the United States was surprised by the Japanese attack on Pearl Harbor was racist; it was not believed that the Japanese would have the ability to plan and carry out such an operation. There was also a technical aspect: it would not be possible to drop torpedoes from aircraft in the relatively shallow waters of the harbor; it was possible. The Israeli intelligence failure before October 7, 2023 was similarly an underestimation of the enemy's capabilities.
War is a complex system
War can be seen as two (or more) complex adaptive systems that strive to limit each other's freedom of action. When one side no longer has any freedom of action, the system stops and that side has lost.
In a military strategic sense, a “complex adaptive system” refers to an environment where multiple actors, weapon systems and domains (land, sea, air, etc.) interact in non-linear ways. Complex because of all the different subsystems interacting in all possible ways, adaptive because one must constantly adapt its actions to the actions of the other and vice versa. Since the interaction between all the different parts is complicated, the outcome is difficult to predict.
In a complex military strategic system, small differences in decisions, timing or information can have disproportionately large consequences: a tipping points. The Battle of Midway in 1942 shows how the combination of a seemingly reasonable order to rearm, inadequate reconnaissance and doctrine meant that the Japanese aircraft carriers were in a position of maximum vulnerability just as the American dive bombers arrived. Japan was never able to replace the losses in aircraft and pilots. The result of the Battle of Midway was therefore that the United States was able to go on the offensive and gradually push back the Japanese combat forces.
In a complex system, even very small differences in the starting position can lead to large differences in the end result. [2] An example is the “miracle at the Marne” in 1914. If reconnaissance aircraft had not discovered the gap in the German line in time, the French countermeasures could have been delayed or misdirected. Then Germany’s chance of defeating France in the first weeks of the war would have been significantly greater.
There are many other characteristics of complex adaptive systems that have military strategic analogies. But that will have to wait for another blog.
In summary, war is an extremely complex activity where a multitude of systems must interact while the adversary does his utmost to prevent this very interaction. The result is that the outcome is never a given – as the war in Ukraine is an example of.
Clausewitz discusses these phenomena by pointing out how the theoretical absolute war changes to the real war when phenomena such as probability, hazard, friction and the commander's "creative mind" are introduced into the reasoning.
Decision
Before the outbreak of war, both sides – for example NATO and Russia – have a set of plans. When war breaks out, it generally turns out that none of the plans work as intended. Although the functionality of war changes, its character remains the same: a dialectic of wills in which wills use force to resolve their conflict (Beaufre).
The military commander must understand the information and what it means at his/her level: strategic, operational or tactical. In the old sequential command system, it was not particularly strange that the battle was divided into different companies at the same time as there were “continuous activities” such as mine clearance. But today, the battle will be more or less continuous – of course with great variations. But some major naval battles will probably not be fought except in special cases such as over-shipping companies. The question is then when should the commander step in and make a decision? At a combat technical level, this is hardly a problem. Admiral Sir John Fisher’s classic 3H still applies here: Hit First, Hit Hard, Keep on Hitting.
In the many wars since the end of the Cold War, the West (NATO, various coalitions) has demonstrated great operational capability. This capability is being further strengthened now that the MDO is about to be realized. An example is Maven Smart System (MSS) which is now seen as a core component of the US effort to Combined Joint All-Domain Command and Control (CJADC2). This consists of interconnected capabilities from sensor to top-level command, which will also connect the command systems of US allies for interoperability. During the beginning of the ongoing war between the US, Israel and Iran, MSS is said to have enabled attacks on over 1,000 targets in 24 hours.
The war against Iran, however, is yet another example of a long line of strategically unsuccessful wars. There was no need for any technical systems that CJADC2 to understand that Iran would close the Strait of Hormuz. Stupidity at the highest level of leadership – heads of state – requires functioning political structures and this is beyond the ambition of this blog to discuss. But there is a military-strategic leadership level underneath to think about.
Decisions are often either too late or too early; rarely on time. In both cases, this is problematic because on the one hand, technology drives decisions in a split second, but on the other hand, the inertia of the systems requires reflection.
The most common thing in the history of war is probably that decisions are made too late; the leadership is surprised. But today's technological systems mean that too early is becoming increasingly common. The general aim is to create "agile leadership organizations", "dynamic decision-making systems" that act on the basis of "the transparent battlefield". It is about creating information superiority - Dominant Battlefield Awareness where enormous amounts of information are translated into actual understanding, despite contradictory data, uncertain quality and cognitive attacks. But of course it is not that simple. Friction, chance and creativity have not played their part. No technical system can compensate for misjudgments of the kind Hitler made in 1944 when he insisted that the Allied invasion should go towards Pas-de-Calais and that the attack on Normandy was a diversionary maneuver; an example of the Magruder principle: make the opponent believe what he already believes.
Decision-making at the strategic level cannot and should not be instantaneous. It is of course good to be able to make quick decisions, but the combat forces are not that agile. There is a risk of jerkiness here – new decisions come before the old ones have had time to be implemented. The characteristics of the different domains and combat forces must be taken into account here.
Land and naval forces move relatively slowly but have long endurance, while air forces are fast with long range but short endurance. Building up adapted logistics takes time. The war in the Middle East, for example, has been affected by a shortage of qualified ammunition that can take several years to remedy. Launching new satellites takes some time, depending on what is available, among other things. A cyber attack requires careful planning; the American-Israeli attack on the IRA with Stuxnet required four to five years of preparation. Cognitive warfare that aims to change attitudes and thought patterns obviously takes a long time. Warfare in the electromagnetic domain is by definition very fast (3*108 m/s).
Military strategic decision-making is largely about the allocation of tasks and resources. Before the decision is made, the commander has freedom of action with a certain amount of resources. Once the decision has been made, he/she has tied up a certain amount of resources to solve a certain task. This reduces the freedom of action. However, if the task is solved, this generally means that the enemy's freedom of action decreases and my own increases. When his freedom of action is zero, I have won. However, it is important to remember that a change of task and/or reallocation of resources almost always involves a cost. This is why it is so important to carefully analyze the political goal against the strategy to achieve it.
Boyd's OODA loop[3] is often used to discuss decision cycles. The winner is the one who goes through the loop the fastest. But the question is whether this is right at a strategic level. Here it is rather analyzing the situation to find winning concepts that is important; that is the second O, Orient, which is important. The first O (Note) and A (Act) are more technical in nature.
Om Orient Boyd writes in summary: In order to succeed, we must act, which in turn means that we must decide how we should act. This requires that we first free ourselves from rigid perceptions and preconceived ideas. To make decisions, we create mental patterns of the environment. This process in turn involves an analysis of the environment, dividing it into pieces, and a synthesis, putting the pieces together but now in a different way. We go from the general to the specific and from the specific to the general; from deduction to induction and back again. The goal of this loop, deduction and induction or analysis and synthesis, is to create new mental images of the environment and in this way find new possibilities for action.
In strategic decision-making, the D is (Decides) is very much a political issue. For example, it could be about the direction of force concentration. In war, we will need strong naval forces both on the west coast and in the Baltic Sea. But these are too few to be spread out. No AI system in the world can decide what is right, but it becomes a question of political risk-taking.
Political-strategic decision-making is not a matter of a technical process or of bureaucracy. It is a matter of intellect, morality and character. It is a matter of experience and the ability to see things in a big way, the ability to think new things. Which requires knowledge of history, geography and people.
AI as a necessary but problematic component
AI is and is increasingly becoming an important support in the work of developing a winning decision. It is an aid; AI does not win wars. Victory requires weapons and people.
AI can handle narrow problems where there is a lot of information, but the less reliable the information, the greater the risk of unpredictable or unclear results.
Military AI is particularly useful in three types of applications:[4] planning, targeting and cyber operations. In terms of planning, AI can support the decision-making process by quickly creating an understanding of the different domains of the area of operation. In terms of targeting, it is AI's ability to quickly analyze very large amounts of data that is interesting. In cyber warfare, AI is an enabler: identifying target groups, creating content, distributing it and amplifying any reactions it arouses.
War is a battle of wills; it is not possible to construct an algorithm that can take into account all the factors involved. While automation involves following fixed rules and performing a certain action without prior analysis of the context, AI has learned to analyze, predict, and suggest.
At the operational and tactical level, AI can often propose decisions independently, especially when it comes to structured and well-known data. At the strategic level, where decisions are more complex and data is often unstructured, AI works best as a support for human decision-makers and not as the sole decision-maker. Strategic management in the AI age should be based on a kind of consultation between different models – historical, geographical, operational-technical, cognitive and algorithmic. AI can then structure the information and test different models and theories. The AI system thus constitutes a voice in this consultation, but cannot replace the human ability to weigh and synthesize the results.
But in such complex systems, even modern computer systems have to simplify, approximate and take “pretty good” solutions. And it is not certain that AI will announce that “I can’t”. Instead, perhaps an approximation or a solution will come that aims to make the user happy.
It is also possible that AI can “recognize” the current situation and suggest a course of action. But that is not how you win a war because if our AI can, so can the opponent’s AI. To win, you have to surprise – not do the probable.
AI cannot transform uncertainty into certainty, but it can structure the problems. How the questions are phrased is very important: “shit in, shit out.”
AI can also be manipulated. To answer the question of whether Ukraine can win the war, the machine will go through a large amount of articles from a variety of sources. When Russia uses automated systems to plant millions of articles saying that Ukraine will lose, our machine will say the same.
The manager's ability to see context, experience and a "commander's perspective" are therefore still indispensable.
This is where the problem with algorithms comes in. As pointed out above, today's managers are unlikely to know the algorithms that his/her system uses. It is likely that they originate from American industry and thus may not fit into our military strategic culture at all. Therefore, we should develop algorithms for decision support ourselves. It could become an important part of our strategic competence development and a means of strategic control. This is also a question of responsibility. "But AI advised me to..." is not an acceptable excuse when things have gone wrong.
For example, after Pearl Harbor on December 7, 1941, Admiral Nimitz could have chosen a defensive strategy with the main task of defending Hawaii and the sea links with Australia while increasing US military capabilities. Instead, he chose to carry out a number of offensive operations to take the initiative, worry the superior Japanese and thereby also raise morale on his own side. The overall objective can be said to be to create chaos out of order on the Japanese side and order out of chaos on his own. The result was the more or less inconclusive Battle of the Coral Sea (May 4–8, 1942) and the turning point at Midway (June 4–7, 1942). Had AI advised him on this winning strategy given the information that was available?
Conclusion
Some final questions among many possible ones:
- What does the outlined development mean for how we should design our future management systems – as decision support or decision substitutes?
- Is strategic competence at risk of eroding if AI systems become too dominant in planning and evaluation? That is, if creative thinking is replaced by machine solutions?
- How can we train managers to think in complex systems rather than just utilizing “transparent” situational pictures?
Finally. The command and control systems – C4ISTAR – are themselves a battlefield. It is therefore necessary to be able to conduct warfare – and combat – even with only limited access to technical systems. War is and remains a human phenomenon.
***
The author welcomes comments: lars@wedinstrateg.fr
The author is a retired commander, member of KÖMS and KKRVA, and associate member of the Academie de marine (Paris). Lars Wedin is also an honorary doctor at the Swedish National Defense University.
Footnotes
[1] I know that the Armed Forces use a different classification but I am not bound by the Ministry of Defence.
[2] Sensitive dependence on initial conditions in technical terms
[3] See Wedin, Lars: “Colonel John Boyd US Air Force – a postmodern strategist”, KKrVAHT No. 4, 2023.
[4] Op. cit., King, see footnote 632, p. 58