A good video game usually introduces a new weapon not by stopping to explain everything it does, but by giving the player the weapon, enough information to make the possibilities legible, and then pointing him toward an opportunity to use it, with the rest of the new mechanic learned through experimentation:
Taken as a metaphor, the familiar fantasy of a teenager unexpectedly finding himself inside a mechsuit works a similar way, and the useful military fiction is what lies underneath: a sophisticated weapon arriving with an interface that allows a new operator to begin using it effectively before either he or the institution around him has completely worked out what it can do. America should try to make that part of the fantasy real.
[…]
Ease of use alone is not enough to make a weapon playable. Far from accelerating time-to-learn, systems made superficially simple by concealing so much of their behavior as to leave operators guessing what’s happening beneath the surface become useless as soon as they depart from the conditions the interface was designed around. A good game, on the other hand, hides an enormous volume of calculations, yet remains very deliberate about revealing the essential and necessary information. A playable military system, which faces the same design problem with higher consequences, should do much the same thing: provide curated legibility by hiding complexity which does not matter to employment while highlighting the essential information, making it visible enough that an operator can build an accurate working model of the machine and its state.
Ukrainian experience with automatic terminal guidance on FPV drones demonstrates the distinction. Experienced drone operators have at times chosen manual control for as long as they had a video link, in order to make better corrections on final attack approach. Other times they’ve preferred automatic guidance using pixel lock for the reason that it proved more than sufficient against targets with regular, recognizable outlines like enemy anti-air or jamming equipment.11 Therefore, the value of automatic or manual guidance is situation-, target-, and mission-dependent. To generalize this example, more than just providing an easily selectable automatic option, an effective, time-to-learn optimized interface will still convey sufficient information for the operator to understand each mode’s function, the limits of their advantages, and adapt to changing circumstances.
The division of work between operator and software changes with the target and the condition of the link. An operator will soon likely cease to “fly” an unmanned aircraft in any conventional sense, and instead tell it where to go, what to observe, and how it should work with other machines. The necessary elements of an aerial weapon’s flight, like navigation, obstacle avoidance, or even parts of terminal guidance can increasingly be submerged, freeing up the operator’s attention for the most important tasks, namely target identification, maintaining situational awareness, and coordinating teaming between the weapon and other systems. Such a configuration would aim to offload work from the operator downward into the machine, ultimately enabling the operator to give more attention to the more consequential work, while preserving interface legibility sufficient for the operator to appreciate when the machine’s behavior or the tactical situation requires intervention.