Professional Standards & Ethics Flashcards
7 cards from real GCP practice questions. Tap to flip, then mark Knew It or Still Learning — missed cards come back until you master them.
Read the first 7 Professional Standards & Ethics flashcards as text
What is 'algorithmic bias' in the context of machine learning systems?
Answer: Systematic unfair outcomes resulting from prejudiced training data or flawed model design
Algorithmic bias occurs when a model produces systematically unfair results due to biased training data, flawed assumptions, or discriminatory design choices.
Which practice most directly supports transparency in AI systems used for high-stakes decisions such as loan approvals?
Answer: Applying model explainability and interpretability techniques
Explainability techniques allow stakeholders and regulators to understand why a model made a particular decision, supporting accountability.
When deploying a facial recognition system for law enforcement use, what is the primary ethical concern that professionals must address?
Answer: Potential for demographic bias causing disproportionate impact on certain groups
Facial recognition systems have documented higher error rates for certain demographic groups, raising serious civil rights and due process concerns.
What does the principle of 'human oversight' require in automated AI decision-making pipelines?
Answer: That consequential automated decisions include mechanisms for human review and intervention
Human oversight requires that humans can review, override, and correct consequential AI decisions rather than treating them as infallible.
A company wants to use ML to rank employees for promotions. What ethical safeguard is most critical?
Answer: Auditing the model for discriminatory bias against protected classes before deployment
Employment decisions affect people's livelihoods, so auditing for bias against legally protected classes is essential before deployment.
What is 'data poisoning' as an attack against ML systems?
Answer: Deliberately corrupting training data to cause a model to learn incorrect or harmful behaviors
Data poisoning involves injecting malicious or misleading examples into training data to manipulate the model's learned behavior.
Under responsible AI principles, what should an organization do when an AI system begins producing harmful or discriminatory outputs in production?
Answer: Immediately investigate, pause the system if necessary, and implement corrective measures
Responsible AI practice requires halting harm promptly, investigating root causes, and implementing fixes rather than waiting for scheduled reviews.