Drivers, warehouse pickers, call center agents, and even freelance writers are managed by systems that optimize for one variable above all others: throughput . The algorithm learns your fastest possible pace, then sets that as the baseline. Slow down even slightly, and you are flagged as “underperforming.” Take a legitimate break, and your rankings drop.
We tend to think of sabotage as dramatic—a wrench in the gears, a hammer to a circuit board. But in the age of platform capitalism, the machinery is no longer physical. It is code. The modern workplace is governed not by foremen with stopwatches, but by performance scores, real-time tracking, and predictive analytics.
Protect the core recommendation/classification algorithm from manipulation by detecting and quarantining "sabotage" inputs (adversarial examples or poisoned data). algorithmic sabotage work
From a corporate perspective, this is "fraud" or "theft of time." From a labor perspective, it is a digital form of —a classic protest tactic where employees follow every regulation to the letter to slow down production.
Preventing automation from unfairly evaluating worker performance. Algorithmic Accountability: Drivers, warehouse pickers, call center agents, and even
This example implements a for a machine learning classifier. It detects "Adversarial Examples"—inputs specifically crafted by an attacker to force the model to make a wrong prediction.
Delivery riders may collectively "ghost" low-tip or high-distance orders. By repeatedly rejecting a specific "bad" job, they force the algorithm to increase the base pay offered for that task to get it fulfilled. Profile "Swapping": We tend to think of sabotage as dramatic—a
Algorithmic sabotage is not about destroying value. It is about reclaiming a margin of humanity. That thirty-second pause between scanning and lifting? That is not theft. That is a breath. That is a blink. That is a worker saying: I am not a node in your network.