As machines become smarter, humans become more important.
Artificial intelligence and autonomous systems are changing the way modern military operations are planned and conducted. The war in Ukraine has demonstrated how rapidly unmanned aerial systems, maritime drones, advanced sensors and AI-supported decision-making tools have become integrated into military operations. At the same time, NATO countries are investing heavily in new capabilities based on artificial intelligence and autonomy.
These technologies offer the potential for faster decision-making, improved situational awareness and more precise use of military capabilities. But they also raise a fundamental question: As machines become increasingly intelligent, how do we ensure that humans retain meaningful control?
“In complex systems, control is not something we assume, it is something we must actively design.”
— Stig Ole Johnsen, NTNU
This question is at the heart of the Norwegian research project MIDAS – Humans in Future Ocean Space Operations, and of the book Safety by Design: Human-Centered Approaches to AI, Automation, and Remote Operations. Edited by Frøy Birte Bjørneseth, Stig Ole Johnsen, Ole Andreas Alsos, Vidar Hepsø and Gunhild Birgitte Sætren, the book brings together research and practical experience on human-centred approaches to AI, automation and remote operations across safety-critical sectors. [1]
The article draws on research, findings and practical experience from the Norwegian MIDAS project and Safety by Design: Human-Centered Approaches to AI, Automation, and Remote Operations, with particular emphasis on the implications for defence. Key themes include meaningful human control, human factors, system design and the interaction between people, technology and organisation.
Although MIDAS is rooted in ocean operations, the book draws on experience from several safety-critical sectors. Many of the challenges are therefore equally relevant to defence.
One of the most important lessons from research into autonomous and remotely operated systems is that autonomy should not be understood as a way of removing humans from the system.
Stig Ole Johnsen's research specifically addresses how Meaningful Human Control can be engineered into digitalisation, automation, AI and remote operations. His work emphasises that human control needs to be deliberately incorporated into the design of complex systems rather than simply assumed to exist because a human remains formally responsible.
“Autonomous and remotely operated maritime systems will not become safe simply by removing people from the bridge or the control loop. They become safer when we understand the work operators actually have to perform, design the technology around that work, and ensure that people can build situation awareness and intervene when conditions change.”
— Stig Ole Johnsen, NTNU
This is closely connected to the concept of Meaningful Human Control. Simply having a human formally “in the loop” does not necessarily mean that the human has effective control. The operator must have sufficient understanding of what the system is doing, access to relevant and timely information, the authority to make decisions and the ability to intervene when necessary. NATO's work on Human Systems Integration for Meaningful Human Control over AI-based systems addresses these issues at the system level. [2]
This has important implications for defence procurement. Human Factors Engineering, task analysis, user involvement, prototyping, simulation and continuous learning cannot be treated as activities that are added after the technology has been developed. They need to be incorporated from the beginning.
The collision involving the frigate KNM Helge Ingstad in 2018 provides a powerful Norwegian example of why technology, people and organisation must be understood as one system.
Research by Ole Andreas Alsos, Stig Ole Johnsen and Frøy Birte Bjørneseth addresses human, technical and organisational factors in accident analysis and safety-critical systems. Their work is part of the broader research presented in Safety by Design.
For Bjørneseth, the lesson from accident investigations is clear:
“When we investigate serious incidents, we often see that attention is focused on the person who made a mistake. But humans always operate within a system. If the system is poorly designed, the likelihood of errors increases.”
— Frøy Birte Bjørneseth, NTNU
Bjørneseth's research focuses on Human Factors, human-machine interfaces, ergonomics, user experience, operational management and safety in autonomous and maritime systems.
The broader lesson is consistent with the central argument of Safety by Design: in safety-critical systems, human performance cannot be separated from the technical and organisational context in which people operate. [1]
For autonomous military systems, this interaction becomes even more important. A technically sophisticated system can still fail to deliver its intended operational effect if information is difficult to interpret, interfaces are poorly designed, responsibilities are unclear or operators cannot intervene effectively.
The war in Ukraine provides an extraordinary real-world laboratory for the development and deployment of autonomous and semi-autonomous technologies. Drones, sensors, satellite data and AI-supported analysis are being integrated into military operations at unprecedented speed.
But the experience from Ukraine also demonstrates that technology alone does not determine military effectiveness.
Humans remain responsible for understanding the operational context, assessing uncertainty and making decisions involving tactical, legal and ethical considerations. The physical distance between an operator and the battlefield may increase, while the cognitive demands on the operator can become greater.
The same principle applies to autonomous maritime operations: the more responsibility is delegated to a system, the more important it becomes to understand what the system can and cannot be expected to do
There is a tendency to view AI adoption primarily as a technology challenge. Research suggests that this can be misleading.
The MIT NANDA report The GenAI Divide: State of AI in Business 2025 examined the gap between experimentation with generative AI and measurable organisational value, highlighting challenges related to implementation, workflow integration and organisational adaptation. [3]
For defence, the consequences of poor integration can be particularly serious. The issue is not simply whether an AI system works technically, but whether it works in the operational environment in which it is intended to be used.
This is where the Human Factors perspective becomes particularly important. Bjørneseth's research covers Human Factors, human-machine interaction, usability, ergonomics, operational environments and safety in autonomous systems. [1]
AI adoption should therefore begin with the mission and the tasks that people need to perform. What decisions have to be made? What information is required? Where is uncertainty greatest? What happens when communication fails? What happens when information is incomplete, misleading or deliberately manipulated?
These questions are as important as the technical performance of the AI model itself.
A Norwegian example can be found in the development of the next-generation Fire Distribution Centre for the Norwegian Advanced Surface-to-Air Missile System, NASAMS.
The system has been developed around the tasks operators need to perform under extreme time pressure. User involvement, observation, task analysis, prototyping and testing have been important elements of the development process.
The result illustrates a fundamental principle of human-centred design: advanced technology does not necessarily have to mean greater complexity for the user.
This approach is consistent with the principles of human-centred design described in ISO 9241-210:2019, which provides requirements and recommendations for human-centred design principles and activities throughout the life cycle of interactive systems. [4]
Norway has several advantages in this field. The country combines a strong defence industry with substantial research expertise in autonomy, Human Factors and human-machine interaction, as well as decades of experience with safety-critical maritime operations.
There are also relatively short distances between users, industry, research institutions and government. This creates an opportunity to develop and test new approaches in close cooperation with the people who will ultimately use the technology.
User involvement should extend from the definition of the problem through prototyping, exercises, acquisition and implementation – and continue as an integral part of operational learning. This is consistent with both the human-centred design principles of ISO 9241-210 and NATO's approach to Meaningful Human Control across the system lifecycle. [2][4]
The war in Ukraine reinforces the value of this approach. Technologies are developed, deployed, tested and modified at extraordinary speed. The ability to establish short learning loops between operational experience, users, researchers and industry can therefore become a strategic advantage.
The central message from MIDAS and Safety by Design is not that autonomy should be slowed down. Quite the opposite. Autonomous and AI-based systems will become increasingly important in both civilian and military operations. [1]
The question is how these systems are designed and integrated.
Testing should not be limited to whether a system performs correctly under ideal conditions. Systems must also be tested under time pressure, degraded communications, incorrect or uncertain information, cyber disruption and other hostile or degraded conditions. NATO's work on Meaningful Human Control addresses human-system integration across the lifecycle, including design, testing, training and operational use. [2]
ISO 9241-210:2019 provides an established framework for human-centred design, emphasising an understanding of users, tasks and environments throughout the development process. [4]
For defence, this approach is particularly important because failure can have consequences far beyond inconvenience or lost productivity. It can affect operational effectiveness, situational awareness, safety and accountability.
The most advanced system is therefore not necessarily the one with the highest degree of autonomy.
Instead, the critical question is whether the system enables people to understand what is happening, make appropriate decisions and intervene when necessary.
This is the essence of Meaningful Human Control – and it is why the human role does not disappear as autonomous systems become more capable. Johnsen's research specifically addresses how Meaningful Human Control can be engineered into digitalisation, automation, AI and remote operations.
The challenge for the defence sector is therefore to combine technological development with human competence, organisational development and continuous user involvement.
When machines become smarter, the human role does not disappear. It changes – and in many cases becomes more demanding.
Safety by Design: Human-Centered Approaches to AI, Automation, and Remote Operations is edited by Frøy Birte Bjørneseth, Stig Ole Johnsen, Ole Andreas Alsos, Vidar Hepsø and Gunhild Birgitte Sætren. The book was published by CRC Press in 2026 as an Open Access publication.
The book brings together research and practical experience on human-centred approaches to AI, automation and remote operations across several safety-critical sectors. Its central themes include Human Factors, human-centred design, automation, AI, remote operations and Meaningful Human Control.