Artificial intelligence (AI) isn’t as automatic as it claims to be. This technology runs on powerful computers that process calculations based on extensive databases. However, these databases require manual oversight, a role often filled by outsourced workers who are typically underpaid. The book Ghost Work (2019), authored by anthropologist Mary Gray and colleague Siddarth Suri, exposes this harsh reality of data labor.
Milagros Miceli, an Argentine sociologist holding a doctorate in Computer Science, has researched this critical aspect for over two years. At the German Internet Institute, she sought to examine the social consequences of algorithms from a practical angle, leading her to discover the hidden world of data annotators. These workers perform essential tasks such as tagging images so that machines can learn to recognize objects. Miceli states, ‘I wondered if anyone was talking to the people behind these algorithms,’ leading her to focus on data workers.
In a recent interview with EL PAÍS, Miceli explained the role of a data worker. ‘A data worker is a person who produces data to train AI systems,’ she noted. “They might record their voice, upload selfies, or label and classify data. This work is not only ongoing but also necessitates the maintenance of datasets and ensuring their accuracy.”
The misconception that data workers are unskilled is prevalent. However, Miceli countered: ‘These workers typically possess at least secondary or higher education. In fact, I’ve met individuals with PhDs engaged in this type of work.’ Many are located in high-unemployment areas yet maintain educational qualifications.
Miceli provided insight into the demanding nature of data work, stating, ‘Labeling and segmenting satellite images are very common tasks. It requires significant attention to detail and can be physically taxing.’ Many data workers face precarious situations, working in environments that heavily favor companies and often leaving them without necessary protections.
She elaborated, ‘They are in a situation of total precariousness, paid by task completion rather than time spent, leading to exploitation. If something goes wrong, clients may withhold pay or claim the task wasn’t completed as requested.’ In many cases, confidentiality agreements prevent workers from discussing their job roles, leaving them voiceless even when they suffer psychological effects from the content they handle.
Miceli has also raised concerns about the future of this labor force as AI continues evolving. ‘The number of data workers will likely grow,’ she predicts. “Even with advancements in AI-generated data, human oversight remains essential for accuracy.”
As she emphasizes, ‘AI would not function without the countless individuals performing manual labor behind the scenes. There’s a narrative being pushed that lies about automation; AI is fueled by human effort.”
Miceli’s work sheds light on the often-overlooked contributions of data workers to the AI landscape, calling for greater awareness and structural support for these essential members of the tech workforce.
- 0 Comments
- Data Workers
- Labor Rights