Jour fixe: Challenges for the development of fair language-based assessments of health, education, behavior, and beyond
Tuesday, 29. November 2022
15:15 - 16:30
Y326 / hybrid
Damian Blasi (PI – Harvard University, USA – Max Plank Institute for the Science of Human History, Germany) Joseph P. Dexter (Harvard University, USA) Amber Gayle Thalmayer (University of Zurich, Switzerland) Camila Scaff (University of Zurich, Switzerland) and Adolfo Martin Garcia (University of Santiago of Chile, Chile)
Linguistic behavior serves as a reliable, inexpensive, and increasingly automated resource to assess different aspects of individuals and societies. Speech helps detect incipient health issues; newspaper corpora are used to identify stereotypes and societal biases; and wordlists are the basis to determine verbal development. However, these and other relevant developments (which we label language-based assessments or LanBAs) have been concocted, tested, and deployed primarily on a handful of large and commercially central languages, with English dominating the scene. Since the 6,500 extant languages can and do vary substantially, transferring LanBAs from English to them is fraught with technical and linguistic challenges. The consequences of this bias, which we are only starting to understand, is that users of minority languages have at their disposal more expensive, less efficient, and potentially biased LanBAs. A novel source of worldwide inequity looms large across multiple social arenas.