Raji ID, Kumar IE, Horowitz A, Selbst A. The Fallacy of AI Functionality. FAccT '22: Proceedings of the 2022 ACM Conference on Fairness, Accountability, and Transparency. 2022; Pages 959-972. https://doi.org/10.1145/3531146.3533158
Summary: This paper highlights the often overlooked issue of AI functionality in deployed systems, which frequently fail to work as intended. The authors argue that current discussions about ethical AI deployment focus too much on value alignment while neglecting the foundational question of whether AI systems are functional and beneficial. The paper presents a taxonomy of AI functionality failures based on case studies and stresses the importance of addressing these failures before considering ethical implications. The authors call for greater attention to functionality in AI policy as a crucial step toward mitigating algorithmic harm.
Commentaires