Bias detection and fairness metrics in MLOps - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to calculate demographic parity difference using the AIF360 library.

MLOps

from aif360.metrics import BinaryLabelDatasetMetric
metric = BinaryLabelDatasetMetric(dataset, privileged_groups=[1], unprivileged_groups=unprivileged_groups)

Drag options to blanks, or click blank then click option'

A[{'sex': 1}]

B[{'race': 0}]

C[{'race': 1}]

D[{'sex': 0}]

Attempts:

3 left

💡 Hint

Common Mistakes

Using unprivileged group as privileged

Confusing attribute names

Using incorrect list format

The privileged group is usually represented by 'sex': 1 (e.g., male). This is used to calculate fairness metrics like demographic parity difference.

2fill in blank

medium

Complete the code to compute equal opportunity difference metric.

MLOps

equal_opportunity_diff = metric.[1]()

Drag options to blanks, or click blank then click option'

Aequal_opportunity_difference

Bstatistical_parity_difference

Cdisparate_impact

Daverage_odds_difference

Attempts:

3 left

💡 Hint

Common Mistakes

Using statistical parity difference instead

Confusing average odds with equal opportunity

The method equal_opportunity_difference() computes the difference in true positive rates between privileged and unprivileged groups.

3fill in blank

hard

Fix the error in the code to calculate disparate impact.

MLOps

disparate_impact = metric.[1]()

Drag options to blanks, or click blank then click option'

AdisparateImpact

Bdisparate_impact

Cdisparate_impact_score

DdisparateImpactScore

Attempts:

3 left

💡 Hint

Common Mistakes

Using camelCase method names

Adding extra suffixes to method names

The correct method name is disparate_impact() with underscores and lowercase letters.

4fill in blank

hard

Fill both blanks to create a fairness metric object for unprivileged and privileged groups.

MLOps

metric = BinaryLabelDatasetMetric(dataset, privileged_groups=[1], unprivileged_groups=[2])

Drag options to blanks, or click blank then click option'

A[{'gender': 1}]

B[{'gender': 0}]

C[{'age': 1}]

D[{'age': 0}]

Attempts:

3 left

💡 Hint

Common Mistakes

Swapping privileged and unprivileged groups

Using wrong attribute names

The privileged group is usually 'gender': 1 and unprivileged is 'gender': 0 to measure fairness across gender.

5fill in blank

hard

Fill all three blanks to create a dictionary comprehension that filters dataset features by threshold.

MLOps

filtered_features = {feature: value for feature, value in dataset.features.items() if value [1] threshold and feature [2] 'age' and value [3] 0}

Drag options to blanks, or click blank then click option'

A>

B!=

C>=

D<

Attempts:

3 left

💡 Hint

Common Mistakes

Using wrong comparison operators

Including 'age' feature by mistake

The comprehension filters features with values greater than threshold, excludes 'age' feature, and values greater or equal to zero.