Autre (Rapports)
Making Audits Meaningful: Overseeing the Use of AI in Content Moderation
Bloch-Weba, Hannah; FERNANDEZ, Angelica; Morar, David
2020
 

Documents


Texte intégral
[EoD] Policy Paper Auditing AI (1).pdf
Postprint Éditeur (254.75 kB)
Télécharger

Tous les documents dans ORBilu sont protégés par une licence d'utilisation.

Envoyer vers



Détails



Mots-clés :
Audits; AI; Content Moderation
Résumé :
[en] While platforms use increasingly sophisticated technology to make content-related decisions that affect public discourse, firms are tight-lipped about exactly how the technologies of content moderation function. The laconic nature of industry disclosure relating to their use of algorithmic content moderation is thoroughly unacceptable, considering that regulators need to understand the platform ecosystem in order to design evidence-based regulations and monitor risks associated with the use of AI in content moderation. This white paper sets out to explain how and why audits, a specific type of transparency measure, should be mandated by law within the four clear principles of independence, access, publicity, and resources. We go on to unpack the types of transparency, and then contextualize audits in this framework while also describing risks and benefits. The white paper concludes with the explanation of the four principles, as they are derived from the previous sections.
Disciplines :
Droit européen & international
Auteur, co-auteur :
Bloch-Weba, Hannah;  Texas A&M University > School of Law
FERNANDEZ, Angelica ;  University of Luxembourg > Faculty of Law, Economics and Finance (FDEF) > Department of Law (DL)
Morar, David;  George Washington University > Elliott School of International Affairs
Langue du document :
Anglais
Titre :
Making Audits Meaningful: Overseeing the Use of AI in Content Moderation
Date de publication/diffusion :
03 décembre 2020
Maison d'édition :
HIIG, Berlin, Allemagne
N° de rapport :
Research Sprint on AI and Content Moderation
Focus Area :
Law / European Law
Disponible sur ORBilu :
depuis le 20 octobre 2022

Statistiques


Nombre de vues
166 (dont 2 Unilu)
Nombre de téléchargements
57 (dont 2 Unilu)

Bibliographie


Publications similaires



Contacter ORBilu