SDS Governance, ethics and storytelling

A credit scoring model is accurate overall but declines applications from one group at a much higher rate for reasons unrelated to risk. The main issue is:

• 1 Speed
• 2 Storage
• 3 Accuracy
• 4 Bias and fairness

Explainability of a model is most important when:

• 1 No one uses the output
• 2 The model is for a game
• 3 The data is small
• 4 Decisions affect individuals and must be justified

Which chart is generally best for showing a trend over time?

• 1 3D bar chart
• 2 Line chart
• 3 Pie chart
• 4 Scatter plot of unrelated data

When explaining uncertainty to a non-technical board, it is best to:

• 1 Use only a single number
• 2 Hide the uncertainty
• 3 Give a range and describe the chance of outcomes in plain words
• 4 Use technical terms only

A model inventory records:

• 1 The passwords of staff
• 2 The share price
• 3 The models in use, their owners, purpose and validation status
• 4 The office layout

Under privacy principles for data science, personal data should be:

• 1 Kept forever
• 2 Used for any purpose
• 3 Shared with any party
• 4 Collected for a clear purpose and kept secure, using only what is necessary

A chart's vertical axis starts at 90 instead of 0 to make small differences look large. This is a concern about:

• 1 Misleading presentation
• 2 Encryption
• 3 Model drift
• 4 Data quality