Implementing bias testing and human-in-the-loop oversight for resume screening

[James.T]

I'm part of a product development team at a software company, and we're integrating a new AI-powered feature for resume screening. We're deeply concerned about potential bias in the training data and the opaque nature of the model's decision-making. We've established a basic review board, but we're struggling to create a practical, ongoing AI ethics framework that goes beyond a one-time audit. How are other teams implementing transparent documentation, bias testing protocols, and human-in-the-loop oversight for features that have real-world consequences for people's careers?