TY - JOUR
AU - Crawford, Kate
AB - Kate Crawford What does it mean to historicize the materiality of artificial intelli- gence? What aspects should be included? These are harder questions than they may initially seem. The term “artificial intelligence” is slip - pery: it can be a technical method, an infrastructure, a set of social prac- tices, an industry, and a way of seeing. Behind every system, there are physical infrastructures designed by people, with social histories that shape their operations, and political economies driving their produc- tion and deployment. It is a wide terrain with many different kinds of primary material to study, and it stretches well beyond the archives of individual technical inventors and organizations. Material histories of artificial intelligence reveal how systems are “made”—materially and discursively. We can think of the labor histo- ries of the women who were the original computers (as documented by Jennifer Light and Mar Hicks), to the role of the underpaid workers who moderate content and categorize data (researched by Sarah Rob- erts, Mary Gray, and Siddharth Suri), to histories of how data practices shifted in industrial labs in the twentieth century (seen in the work of Xiaochang Li and Mara Mills), economic and political histories (such as 
TI - Archeologies of Datasets
JO - The American Historical Review
DO - 10.1093/ahr/rhad364
DA - 2023-09-26
UR - https://www.deepdyve.com/lp/oxford-university-press/archeologies-of-datasets-5bRcsDVUK7
SP - 1368
EP - 1371
VL - 128
IS - 3
DP - DeepDyve
ER -