ghost work
In plain wordsThe hidden human labor behind AI: people, often poorly paid and far away, who label data, check outputs and handle cases the software cannot, while the product appears fully automatic.
Task-based, often on-demand human labor hidden behind services presented as automated: labelingdata labeling, n.: The work of attaching correct answers to examples — often done by large paid workforces, sometimes by experts — so that AI systems can learn from them or be tested against them. and annotatingannotation, n.: Adding information to raw data by hand or with software — marking objects in photos, tagging names in sentences, transcribing speech or ranking chatbot answers — so machines can learn from it. training data, moderating content, transcribing, rating chatbot answers for reinforcement learning from human feedbackreinforcement learning from human feedback, n.: A way of training AI using people's judgments: humans compare pairs of answers, a scoring model learns their preferences, and the AI is then trained to produce answers that score well., writing demonstrations, and stepping in when software fails. It is typically sourced through crowdsourcingcrowdsourcing, n.: Getting many people over the internet to each do a small piece of a big job, such as labeling pictures for a few cents apiece, with anyone free to join in. platforms or outsourcing firms, paid per task, and performed by contractors without the protections of employment, frequently in lower-income countries. Researchers document low and unpredictable pay, algorithmic managementalgorithmic management, n.: When software, rather than a human manager, assigns work, sets pay, tracks performance and even disciplines or fires workers, as in ride-hailing, delivery and warehouse jobs., exposure to disturbing material and the erasure of workers' contributions from public accounts of AI. As frontier training shifted toward expert feedback, part of the work moved to better-paid specialists, but the pattern of invisibility persists.
Behind the fully automated receipt scanner was ghost work: contractors correcting every third scan by hand.
Origin
Coined by anthropologist Mary L. Gray and computer scientist Siddharth Suri in Ghost Work: How to Stop Silicon Valley from Building a New Global Underclass (2019), based on studies of Amazon Mechanical Turk and similar platforms. A 2023 TIME investigation into Kenyan workers paid less than $2 an hour, via the outsourcing firm Sama, to label toxic content for OpenAI made the issue widely known.
Usage note
Related terms include microwork, click work, data work and heteromation (Hamid Ekbia and Bonnie Nardi, 2014): the shifting of essential tasks onto unpaid or low-paid human labor inside apparently automated systems.
Related
- data labelingdata labeling, n.: The work of attaching correct answers to examples — often done by large paid workforces, sometimes by experts — so that AI systems can learn from them or be tested against them.
- crowdsourcingcrowdsourcing, n.: Getting many people over the internet to each do a small piece of a big job, such as labeling pictures for a few cents apiece, with anyone free to join in.
- annotationannotation, n.: Adding information to raw data by hand or with software — marking objects in photos, tagging names in sentences, transcribing speech or ranking chatbot answers — so machines can learn from it.
- content moderationcontent moderation, n.: Checking what people post, or what an AI produces, and removing, limiting or labeling whatever breaks the rules or the law — done by people, by AI, or by both together.
- algorithmic managementalgorithmic management, n.: When software, rather than a human manager, assigns work, sets pay, tracks performance and even disciplines or fires workers, as in ride-hailing, delivery and warehouse jobs.
- AI colonialismAI colonialism, n.: The criticism that the AI industry repeats patterns of colonial empires, taking data, labor, land and resources from poorer places and peoples while profits and control stay with a few rich companies and countries.
- reinforcement learning from human feedbackreinforcement learning from human feedback, n.: A way of training AI using people's judgments: humans compare pairs of answers, a scoring model learns their preferences, and the AI is then trained to produce answers that score well.
Sources
- Gray, M. L. & Suri, S. (2019). Ghost Work: How to Stop Silicon Valley from Building a New Global Underclass. Houghton Mifflin Harcourt.
- Perrigo, B. (2023). Exclusive: OpenAI Used Kenyan Workers on Less Than $2 Per Hour to Make ChatGPT Less Toxic. TIME, 18 January 2023.
Mentioned in
AI colonialismAI colonialism, n.: The criticism that the AI industry repeats patterns of colonial empires, taking data, labor, land and resources from poorer places and peoples while profits and control stay with a few rich companies and countries., AI washingAI washing, n.: Falsely claiming that a product or company uses AI, or exaggerating how much it does, to look more advanced and attract customers or investors., algorithmic managementalgorithmic management, n.: When software, rather than a human manager, assigns work, sets pay, tracks performance and even disciplines or fires workers, as in ride-hailing, delivery and warehouse jobs., crowdsourcingcrowdsourcing, n.: Getting many people over the internet to each do a small piece of a big job, such as labeling pictures for a few cents apiece, with anyone free to join in., data labelingdata labeling, n.: The work of attaching correct answers to examples — often done by large paid workforces, sometimes by experts — so that AI systems can learn from them or be tested against them., Scale AIScale AI, prop. n.: A company that pays large numbers of people to label examples, write model answers and grade chatbot replies, producing the data AI developers train and test on. Meta bought nearly half of it in 2025., Amazon Mechanical TurkAmazon Mechanical Turk, prop. n.: A website Amazon opened in 2005 that acts as a hiring hall for tiny online jobs, such as labeling images for AI, which registered workers pick up and finish for a few cents each., Distributed AI Research InstituteDistributed AI Research Institute, prop. n.: Timnit Gebru set up this institute in 2021, a year after her departure from Google, to research the damage AI can do, especially to marginalized communities, without relying on big-tech money.