Ethics & Bias in AI �Activities

05/05/2026

Q1

• Given the table with ground truth of an allocation problem (Y_true), and a prediction made by a Machine learning (ML) model (Y_pred). Please note: 0 means an allocation was not made, 1 means it was made. One protected feature was listed for each instance in terms of {p, x}. Consider: p is unprivileged.

Please compute Statistical Parity Difference of the ML model.

Your answer:

Q2 ~ same dataset

• Given the table with ground truth of an allocation problem (Y_true), and a prediction made by a Machine learning (ML) model (Y_pred). Please note: 0 means an allocation was not made, 1 means it was made. One protected feature was listed for each instance in terms of {p, x}. Consider: p is unprivileged.

Please compute Disparate Impact of the ML model.

Your answer:

Q3 ~ same dataset

• Given the table with ground truth of an allocation problem (Y_true), and a prediction made by a Machine learning (ML) model (Y_pred). Please note: 0 means an allocation was not made, 1 means it was made. One protected feature was listed for each instance in terms of {p, x}. Consider: p is unprivileged.

Please compute Equal Opportunity Difference of the ML model.

Your answer:

Q4 ~ same dataset

• Given the table with ground truth of an allocation problem (Y_true), and a prediction made by a Machine learning (ML) model (Y_pred). Please note: 0 means an allocation was not made, 1 means it was made. One protected feature was listed for each instance in terms of {p, x}. Consider: p is unprivileged.

Please comment based on your understanding from Questions 1, 2 and 3:

on i) the original dataset, ii) the predictor model.

Your answer: