The right to human decision: analyzing law, ethics, and implementation

As automated decision-making advances, many governments have adopted AI ethics guidelines that endorse a "right to human decision" mandating that important decisions made by AI algorithms (i.e. housing loan eligibility, judicial rulings) are subject to review by human decision-makers with final decision-making authority. The African Union, Canada, China, European Union, United Arab Emirates, United States, Uruguay, and political entities with jurisdiction over most of the world's population have all endorsed a right to human decision, even though their proposals have received relatively little public attention. This paper examines the major proposed forms of the right to human decision: a right to contest decisions made by AI to a human reviewer, the right to opt out from AI decision-making in favor of a human decision-maker, and the right to exclude machines entirely from particular kinds of decisions. I examine key arguments for and against a right to human decision. I argue that several practical challenges must be addressed before this proposed right is implemented. For instance, the distinction between a “human” and “AI” decision is not always clear and the advent of hybrid human-AI decision makers will only blur that line.