Question

A bank is developing a machine learning model to help approve or deny personal loan applications. During testing, the data science team discovers the model approves applicants from certain demographic groups at significantly higher rates than others, even when financial qualifications are similar. According to Google's responsible AI principles, what should the team do before deploying this model?

Accepted Answer

Investigate and mitigate the bias to ensure the model treats applicants fairly and avoids unjust impacts

Answer

Deploy the model as-is because it produces accurate predictions on the historical training data

Answer

Increase the size of the training dataset without examining how the model treats different groups

Answer

Remove all human oversight so the automated system can make faster, more consistent decisions