Respect for Autonomy in the Machine Learning Pipeline.

The machine learning (ML) community recognizes the potential impact of ML systems on human rights, especially regarding privacy and discrimination. To address these concerns, the community has conducted various studies on fairness, accountability, and transparency in developing and deploying ML systems. Despite these efforts, the importance of autonomy, a fundamental principle underlying many human rights, has often been overlooked. This oversight is concerning as it could jeopardize individuals’ decision-making and exercise effective control over their lives, resulting in a violation of their rights. This article examines the principle of autonomy and its significance in the ML pipeline from a transdisciplinary perspective. The authors argue that autonomy remains a theoretical concept and does not translate well into the real use of ML, leading to contradictory outcomes. The absence of an effective approach to integrating autonomy leads to the persistence of disparities.