Knowledge check

Quiz

Complete

Evaluate your understanding of this unit

Questions cover lineage, metadata, bias in collection and preprocessing, and governance themes from Unit 2.

What is the primary purpose of data lineage in AI/ML workflows?

A) To improve the speed of AI model training.B) To ensure data integrity, reproducibility and regulatory compliance.C) To visualise relationships between machine learning algorithms.D) To automate model selection in machine learning pipelines.

Which of the following best describes provenance metadata?

A) Metadata that describes data origins, modifications and movement.B) Metadata that defines relationships between datasets in a relational database.C) Metadata that controls user permissions and access rights to datasets.D) Metadata that describes the content and meaning of a dataset.

How can organisations improve the quality of metadata in AI projects?

A) By manually tagging datasets without standardisation.B) By using metadata repositories and automated tagging tools.C) By limiting metadata collection to reduce storage costs.D) By relying only on descriptive metadata for data discovery.

What is a major consequence of failing to track data lineage in AI systems?

A) Increased model transparency and fairness.B) Reduced ability to audit, debug and ensure regulatory compliance.C) Faster AI deployment with improved accuracy.D) Improved scalability of machine learning workflows.

What is an example of sampling bias in AI models?

A) An AI system that performs well on all demographic groups.B) A facial recognition system trained mostly on light-skinned individuals that struggles to identify darker-skinned individuals.C) An AI hiring system that evaluates resumes based on skills rather than demographic data.D) A predictive maintenance AI that correctly identifies failing machinery.

What is the primary goal of bias mitigation in AI?

A) To make AI models prioritise specific demographic groups.B) To reduce the risk of unfair, discriminatory or inaccurate AI predictions.C) To remove all subjective elements from AI decision-making.D) To speed up AI training by simplifying datasets.

How can AI developers address measurement bias in datasets?

A) By ensuring consistent data collection methods across all demographic groups.B) By training AI models only on historical data.C) By avoiding fairness-aware machine learning techniques.D) By manually removing demographic data from all datasets.

Why is transparency important in AI bias mitigation?

A) It allows users to understand how AI models make decisions.B) It makes AI models more efficient by reducing complexity.C) It eliminates the need for fairness audits.D) It ensures that AI models will never produce biased results.

What role does administrative metadata play in AI governance?

A) It describes how data is structured and formatted.B) It tracks ownership, access controls and regulatory compliance requirements.C) It provides insights into how often datasets are used in machine learning models.D) It helps define the relationships between datasets in a database.

What is the benefit of using bias detection tools like Fairlearn or AI Fairness 360?

A) They automate the removal of bias from AI models.B) They help evaluate fairness metrics and apply corrective measures.C) They improve AI model speed and efficiency.D) They eliminate the need for human oversight in AI ethics.

Action item: Knowledge check

Use the feedback blocks to self-check your answers.

In progress