Imagine you have a binary classification model that determines whether an applicant is eligible for a loan. As a financial institution, it's essential to provide a reason for rejection to each applicant who doesn't qualify. Without access to the feature weights, how can you explain to each rejected applicant why their application was denied?