PROJECT 4: CREDIT REDUX
DSCI 451 Fall 2022
At the beginning of the semester, you created a model to predict whether a given borrower would repay their loan. For this project, you will revisit that work and revise your based based on what you have learned over the course of the semester.
The revision of your work from Project 1 can draw on two sources. First, it should draw on material we have covered in class since you initially submitted Project 1. You may apply whichever concepts and techniques from the subsequent three units that you find relevant. Second, it should draw on other students’ response to your work. In Project 2, your model was assessed for fairness by another student. In Project 3, your model received a values audit by another student. In Project 4, you will revise your model as informed by those assessments.
In addition, you will also document the performance of your model using Mitchell et al.’s (2019) Model Cards approach. You have seen examples of Model Cards, both in Mitchell et al.’s paper and in class. You will need to decide, based on on the guidance provided by Mitchell et al., what to include in your model card and how to report it.
DELIVERABLES
This project includes three deliverables:
• Original R Notebook: You should include the R Notebook that you originally submitted for Project 1. This will assist with the use of CourseSite’s blind grading.
• Revised R Notebook: This notebook should include the revised version of your credit prediction model. It should both describe and justify changes made from the originally submitted model, both in response to the other student’s assessments and in response to subsequent course material.
• Model Card: A card details about the model (see Mitchell et al. 2019).
There are no strict length limitations for this project. Submissions should be as long as they need
to be and not any longer.
REFERENCES
Margaret Mitchell, Simone Wu, Andrew Zaldivar, Parker Barnes, Lucy Vasserman, Ben Hutchinson, Elena Spitzer, Inioluwa Deborah Raji, and Timnit Gebru. 2019. Model Cards for Model Reporting. In Proceedings of the ACM Conference on Fairness, Accountability, and Transparency (FAccT), 220–229. https://doi.org/10.1145/3287560.3287596