
Senior Principal Researcher at Microsoft Research
A scholar of AI’s social and material foundations, examining the labor, data, resources, and institutions behind automated systems.
Kate Crawford is a researcher and author who studies the social and political implications of artificial intelligence. She earned her doctorate at the University of Sydney and developed a research career spanning academic institutions and industry. Her work asks how technical systems are shaped by the resources, classifications, and institutional decisions that make them possible.
Crawford co-founded the AI Now Institute with Meredith Whittaker in 2017. The institute studies the social effects of AI, including accountability, power, and the implications of deploying automated systems in important public and private settings. Her work helped establish these issues as an organized area of research rather than a secondary discussion around model performance.
In Atlas of AI, published in 2021, Crawford examined AI through its material and social foundations, including extraction, labor, data collection, and classification. She has also collaborated with artist and researcher Trevor Paglen on projects including Excavating AI, which investigates the assumptions embedded in image datasets and the categories used to label people.
Crawford serves as a senior principal researcher at Microsoft Research and holds academic affiliations including USC Annenberg. Her contribution is analytical rather than the invention of a particular model architecture. She connects the operation of AI systems to the economic and political arrangements surrounding them, offering frameworks for understanding what is at stake when automated technologies are developed and deployed.
Co-founded AI Now in 2017 to study the social implications of AI and bring interdisciplinary research into public discussions about its deployment.
Published Atlas of AI in 2021, examining the labor, resources, data practices, and political choices involved in building AI systems.
Collaborated on Excavating AI, a study of image training datasets and the assumptions embedded in their categories and labels.