Knowledge check

Quiz

Complete

Test your knowledge.

Evaluate your understanding of data privacy, governance frameworks, and data quality in machine learning by completing this knowledge check.

What does the principle of "privacy by design" mean in the context of machine learning systems?

A) Privacy features are added after deployment.B) Privacy is ensured by keeping all data encrypted at all times.C) Privacy is built into the system architecture from the outset.D) Privacy is the responsibility of third-party vendors, not internal teams.

Your team is building a machine learning model for a wellness app that processes user health histories. Which regulation must you comply with to ensure lawful handling of this health-related data in the United States?

A) GDPRB) CPRAC) HIPAAD) LGPD

Which of the following is a core principle of the AREA governance framework?

A) Human oversightB) Data minimisationC) ExplainabilityD) Fairness

Which framework focuses on sustainable, fair, and ethical AI data use?

A) AREAB) ISO/IEC 27001C) SAFE-DD) HIPAA

Why is data lineage important in ML data management?

A) It helps reduce training time.B) It ensures data is always anonymised.C) It provides transparency and traceability throughout the ML pipeline.D) It guarantees model fairness.

Which example best illustrates metadata tracking in an ML project?

A) Using synthetic data to augment the training set.B) Storing each dataset's source, schema, and access history.C) Encrypting training data during transmission.D) Applying feature scaling to standardise inputs.

A logistics company deploys a predictive maintenance model using sensor data from different regions. After rollout, engineers report frequent failures in one region where sensor data formats differ. What aspect of data quality is likely at fault?

A) CompletenessB) ConsistencyC) TimelinessD) Accuracy

What is the purpose of disaggregated model evaluation in the context of fairness?

A) To optimise runtime performance.B) To compare model accuracy across demographic subgroups.C) To balance training batches.D) To anonymise outputs.

Your ML model is underperforming for female founders in a small business funding dataset. Upon review, you find the training data has far fewer records for that group. What type of bias is most likely at play?

A) Labelling biasB) Sampling biasC) Historical biasD) Model drift

An education startup is designing an ML system to predict student performance. The team decides to store raw datasets indefinitely in case they're needed later. What principle is being violated?

A) Data minimisationB) AccuracyC) Metadata trackingD) Fairness

Action item: Knowledge check

Complete the quiz below to verify your mastery of this unit's concepts.

Quiz interaction placeholder...

In progress