Under Efficiency’s Shadow: The Algorithmic Imaginary of Warfare

A soldier looks at a screen. A name and a photo appear, flagged by an algorithm as a militant. She has only seconds to decide: is the person a lawful target? That, and little else, risks being the human judgment separating an AI recommendation from lethal force.

This is not about autonomous weapons, where machines identify and attack targets without human involvement. It is the quieter reality of AI decision-support systems that increasingly shape military decisions, including those about life and death. Their appeal rests on a deceptively simple conviction: algorithmic systems can solve demanding decision tasks better than humans. Faster. More accurately. More rationally. Seductive though it is, that promise risks displacing the human judgment the law of armed conflict depends upon.

Yet there is a familiar echo in this promise. Long before algorithms took centre stage, strategists envisioned perfecting decision-making through speed and precision. Today, AI represents just the latest attempt to turn that vision into reality. The real difference is not the ambition itself, but in how much human judgment we are prepared to surrender in the pursuit of it.

The Legacy of Speed

The modern ideal of accelerated decision-making is most closely associated with military strategist John Boyd and his Observe, Orient, Decide, and Act (OODA) loop. Boyd argued that victory belongs to the side that can cycle through this loop faster than its adversary can respond. Decision-making speed, in this account, is not just one virtue among others. It is the decisive one.

That logic continues to shape military thinking. It influences how militaries are trained, how command structures are designed, and how new technologies are evaluated: do they help decisions to be made more quickly and efficiently?

A (Re)newed Dream: The Promise of Machine Efficiency

AI has gained traction largely because it promises to fulfil this longstanding ambition better than humans can. For decades, rapid decision-making was understood as a distinctly human achievement, cultivated through training and intuition. Now, humans are increasingly viewed as the main obstacle: prone to fatigue, cognitive overload, inconsistency, and delay. AI seeks to eliminate these constraints, drastically raising the bar for targeting emerging threats from minutes to an astonishing 1,000 tactical decisions every hour (equating to one every 3.6 seconds). This represents a decision compression far surpassing the speed of human thought.

The shift is easy to overlook because the relentless pursuit of faster, more efficient decisions remains the ideal. What has changed is only the hero at the centre. The subtle but profound result is that humans are transitioning from actively applying informed judgment to passively endorsing algorithmic suggestions.

This logic is already visible in AI-enabled military systems, such as “Lavender” and “Gospel”, reportedly used by Israeli forces in Gaza, or the “Maven Smart System”, reportedly used by the US during Operation Epic Fury against Iran earlier this year. Although described as AI decision-support tools, they increasingly perform tasks traditionally reserved for trained military officers. Tasks that hold profound legal implications: identifying lawful targets and recommending whether to use force. Formally, humans remain involved in military decision-making. However, in practice, this involvement can wane, especially when the speed of decision-making leaves little room for thorough review of algorithmic suggestions. Instead of supporting human judgment, the push for efficiency threatens to turn human oversight into a mere “stamp of approval” of algorithmic recommendations, drastically altering the very nature of military decision-making.

Machine Efficiency and War’s Legal Deliberation

Once machine speed becomes the defining measure of military decision-making, deliberation about civilian protection risks appearing as inefficiency rather than a legal duty. This redistribution of authority matters because it challenges one of the foundational assumptions of the law of armed conflict. The law presupposes that decisions about the use of force remain the product of genuine human judgment. Obligations such as the distinction between civilians and combatants, proportionality, and the taking of feasible precautions in attack are not purely computational exercises but require contextual and qualitative evaluation. If meaningful deliberation migrates from the human to the algorithm, the legal architecture remains formally intact while its underlying premise is quietly hollowed out.

The tension becomes clearest where efforts to optimise military decision-making collide with the protective logic of the law of armed conflict. Each rests on a different understanding of what constitutes ideal military decisions. Optimisation prizes “maximum lethality”. The law recognises the realities of combat, but insists that military necessity does not eclipse the obligation to protect civilians. Commanders, for instance, must take all feasible precautions before launching an attack, by assessing risks and considering whether civilian harm can be reduced. Those obligations demand deliberation, and deliberation takes time.

Once machine speed becomes the defining measure of military decision-making, deliberation about civilian protection risks appearing as inefficiency rather than a legal duty. This shift can quietly reshape the standards by which military decisions are judged. Civilian protection need not be explicitly abandoned to be weakened. It may instead be eroded more subtly, as the growing faith in faster, AI-assisted decisions leaves ever less room for the human judgment that the law demands.

The Future of Warfare: A Human Choice

It is tempting to treat this trajectory as inevitable, as if legally significant decisions will naturally migrate from humans to AI systems to win wars. But AI systems are not neutral “tools” that emerged from nowhere. They are general-purpose technologies that do whatever they were designed to do. Every system embodies human choices about what to optimise for and what kind of war it is preparing for.

But AI systems are not neutral “tools” that emerged from nowhere. They are general-purpose technologies that do whatever they were designed to do.  Therefore, the current trajectory reveals less about what AI truly is than about the priorities and blind spots of the institutions behind its development. The pursuit toward ever faster targeting decisions is not an objective insight into what war must become. It is a normative vision of what war should become, dressed in the borrowed authority of the machine.

This recognition offers both hope and caution. If today’s AI systems mirror human choices, tomorrow’s can reflect different ones. Changing course begins by rejecting the idea that this path is unavoidable. Then doing the harder work of imagining these technologies through a different set of priorities, values, and ambitions.

Ultimately, AI will not determine the future of warfare. We, humans, will. The real challenge, in warfare, as everywhere else, lies not in deciding what AI can do, but what kind of humanity we want it to embody.