Artificial intelligence as heteromation: the human infrastructure behind the machine

This article interrogates the widespread narrative of Artificial Intelligence (AI) as autonomous, intelligent, and self-sufficient, and instead centers on the largely invisible human labor that sustains these systems. Drawing on the frameworks of heteromation and human infrastructure, we analyze how AI systems are deeply reliant on distributed networks of ghost workers, crowdworkers, and microtaskers, often working in precarious conditions, to perform essential and low-paid tasks such as content moderation, data annotation, and fact-checking. Far from being fully automated, these systems operate as sociotechnical assemblages where algorithmic processes are scaffolded by fragmented, hidden, and undervalued human work. This article explores how platforms like Amazon Mechanical Turk and Appen organize and obscure this labor through algorithmic management and cultural erasure, creating an illusion of machinic intelligence while extracting value from unpaid or underpaid workers. By examining examples ranging from self-driving car training to misinformation moderation, we argue that understanding AI requires acknowledging the human infrastructures that animate it. This perspective challenges dominant techno-deterministic ideologies and reframes ethical debates around AI to focus on labor, visibility, and exploitation. Ultimately, this article calls for critical attention to the political economy of AI and insists that any ethical framework for AI must begin with the recognition and fair treatment of the human labor that makes it function.