Humanwashing: The Empty Promise of Human Oversight in AI

A recent study published on arXiv claims that the term 'human in the loop' is misleadingly associated with safety in AI decision-making frameworks. The authors argue that although human oversight is frequently proposed as a solution to issues like bias, discrimination, misinformation, manipulation, accountability, and transparency, there is a lack of thorough analysis regarding its true implications. The careless application of the loop metaphor clouds both processes and results, leading to 'humanwashing'—similar to 'greenwashing,' where terminology is used more to foster a sense of security than to guarantee it. The paper raises concerns about whether this metaphor enhances comprehension of what is necessary and accomplished in particular decision-making scenarios, suggesting it is often misapplied in current systems.