Responsible Data Science Quiz - Statistics.com: Data Science, Analytics & Statistics Courses

1. You are working on a loan approval model for a large bank and one of the available features for the model is race. Should your model use race as a feature in making loan approval recommendations?

a. Yes -you will need to have this data for reporting to government regulators

b. Yes -legal restrictions on the use of race allow its explicit use if there are also other predictors

c. No -it is not legal

d. No -it is legal, but might cause reputational harm to the bank

2. In a responsible data science framework, the phrase “performance-interpretability tradeoff” refers to:

a. The fact that to get maximum accuracy in a model, you must often turn to black-box models that lack explainability

b. The fact that model performance for the data scientist is defined differently from the model performance as regulators define it

c. What statisticians call the “bias-variance tradeoff”

3. Unintended group bias in an AI model can result when

a. A linear instead of data-centric (machine learning) model is used

b. A data-centric (machine learning) model is used instead of a linear model

c. A variable that is a proxy for a protected group is included

d. One or more feature is highly correlated with another feature

First name

Last name

Email(Required)

Responsible Data Science 3 Questions