The conversation delves into the complexities of accessing protected attributes in highly regulated industries like banking. Parinaz emphasizes the need for data scientists to test models against biases, suggesting the use of APIs as a potential solution to navigate privacy concerns. The discussion also highlights the existence of proxy variables that can serve as indirect indicators of sensitive attributes, raising important questions about model testing and ethical data practices.
