Suppose we have a 2-class classification algorithm. We also have a dataset on loan applicants similar to the one we considered in class but now defined on the following 4 classes:

The applicant should definitely be considered for the loan.

The applicant may be considered for the loan.

The applicant may default the loan.

The applicant most likely will default the loan.

Describe a procedure for inferring a classifier from this 4-class dataset where the above 2-class classifier could be used as a module (i.e., a black box). Explain how your approach would work.