CPMAI_V7 practice question 19 of 35

Your team is working on a new loan decision model that takes a number of factors and data points into consideration and then automatically approves…

Full set: 100 questions
Your team is working on a new loan decision model that takes a number of factors and data points into consideration and then automatically approves or denies a loan. After a month in operation someone does a review and notices that the system is denying a large number of loans from a certain demographic when all other factors from people in other regions (such as age, salary, and credit score) are the same. What is most likely happening here?
Answer options
A: Biased data sets leading to algorithmic discrimination Correct answer
B: Generative AI models hallucinating data results
C: Nothing is wrong, algorithmic decisions will never be 100%
D: Data privacy issues leading to data sharing concerns
Correct Answer:
A. Biased data sets leading to algorithmic discrimination
Explanation

The disproportionate denial rate for a specific demographic, despite equivalent qualifying attributes, indicates algorithmic discrimination produced by biased training data. If historical data reflect past prejudices or sampling imbalance, the learned model internalises that bias, producing unfair outcomes—an issue well documented in credit-scoring and lending AI systems.

Why Incorrect

B. Hallucination refers to fabricated text/image content in generative models, not structured loan decisions.
C. Accuracy limits do not systematically disadvantage one protected class; bias is the core problem.
D. Privacy breaches concern data exposure, not inequitable approvals.

References

1. Barocas, S., Hardt, M., & Narayanan, A. “Fairness and Machine Learning”, ch. 2 “Bias in, Bias out”, pp. 23-27. (https://fairmlbook.org)

2. U.S. Federal Reserve, “Machine Learning in Credit Underwriting” (2021) §3.2 Data Bias, pp. 11-12.

3. Stanford CS221 Ethics Notes, “Algorithmic Fairness in Lending”, slides 5-7.

Scroll to Top