THE SCREEN WE LOOK THROUGH ALSO TARGETS US
Oct. 2026
We still call the screen a window. Yet before an image appears, most of what could have appeared has been eliminated. On YouTube, one model reduces millions of videos to a few hundred; another orders them according to how long each is expected to keep the person watching (Covington, Adams, and Sargin 2016). What reaches the eye is not an open world but a field narrowed and arranged to produce a response.
Looking through the screen and being targeted through it occur in the same rectangle. From one side, a person sees a face, a war, a product, a message. From the other, an adaptive apparatus has selected the moment and sequence, then measures the pause, replay, click, exit, or return. The same surface delivers a payload and captures a response. A failed image is not wasted: it teaches the next selection. This is a ballistic interface—not because every image wounds, but because exposure has a trajectory, an address, an objective, and a correction mechanism. The apparatus aims statistically; every miss can improve the next shot.
Vision is a powerful entrance because seeing already participates in preparing action. Visual processing identifies and locates objects, recruits memory and affect, and organizes movement (Goodale and Milner 1992). The interface places an executable gesture inside this mobilization: share, buy, enlarge, continue. It compresses the interval in which the impulse might be questioned or another intention might intervene.
The screen is not itself the weapon. It is the terminal surface of what I call a cognitive autonomous weapon: a distributed system that changes its interventions according to measured responses while an institution defines success (Poenaru, forthcoming). If anxiety produces checking, agitation prolongs viewing, or exhaustion favors effortless continuation, the metric can register victory while the organism absorbs the cost. At one in the morning, the rectangle records retention. The following morning, the body receives the shortened sleep. The screen has stopped glowing; the intervention continues biologically.
The two directions are radically unequal. The person sees what appears but cannot inspect why, what was suppressed, or how the response will alter the next selection. The apparatus records it and returns through the same rectangle with a revised intervention. Perceptual freedom cannot mean merely choosing among objects that survived an invisible elimination. It must include the power to govern the machinery arranging the visible. Otherwise the window remains a firing slit decorated as choice.
References
Covington, P., J. Adams, and E. Sargin. 2016. “Deep Neural Networks for YouTube Recommendations.” RecSys ’16, 191–198.
Goodale, M. A., and A. D. Milner. 1992. “Separate Visual Pathways for Perception and Action.” Trends in Neurosciences 15: 20–25.
Poenaru, L. Forthcoming. Cognitive Autonomous Weapons. Geneva: The Mental Forecast PRESS.
JOIN THE LAB
Commentaires