Private traits and attributes are predictable from digital records of human behavior

Published in PNAS, this empirical study demonstrates that personal traits and sensitive attributes can be accurately predicted from digital behavioral records, raising important privacy and AI safety concerns regarding data exploitation and algorithmic discrimination.