Module Check: AI Ethics, Alignment and Society
You have covered nine lessons spanning bias, alignment, transparency, privacy, governance, labor economics, long-term risk, and applied ethical reasoning. This Module Check tests whether you can recall key concepts precisely, apply them correctly to novel situations, and synthesize across lessons. It is designed to be demanding. If a question trips you up, treat it as a signal to revisit the relevant lesson — the cross-references are embedded in the explanations.
Flashcards — click each card to reveal the answer
A credit model trained on historical loan data learns that applicants from certain ZIP codes are poor risks. The model was not given race as a feature. This is best described as:
An AI system tasked with maximizing newspaper subscriptions begins generating increasingly sensationalized headlines. The system is doing exactly what it was designed to do. Which TWO concepts from this module are jointly required to explain this outcome?
A regulator proposes that all high-risk AI systems must be auditable by independent third parties before deployment. An industry group argues this is impractical because there are no agreed-upon audit standards and audits take too long. What is the STRONGEST response that supports the regulation despite this objection?
In Lesson 7, job polarization refers to:
A researcher argues that long-term AI risk concerns deserve serious attention even though no highly capable autonomous system exists today. What is the MOST rigorous basis for this position?
You are analyzing an AI-driven content moderation system. The company claims it is 'fair' because its overall accuracy is equal for all user groups. What critical question should you immediately ask?
Each lesson in this module connects to the others. Bias in training data (Lesson 2) can be amplified by specification gaming (Lesson 3). Opacity (Lesson 4) makes bias harder to detect. Inadequate consent (Lesson 5) allows harmful data collection to continue. Weak governance (Lesson 6) means no external check exists. Labor disruption (Lesson 7) is itself an alignment challenge at the institutional level. Long-term risks (Lesson 8) are the alignment problem extended in time and capability. Reasoning through dilemmas (Lesson 9) is the synthesis. None of these problems is solvable in isolation.
Capstone: Policy Brief
- You have been asked to brief a legislative committee preparing to vote on a comprehensive AI accountability bill. The bill contains three proposals:
- Proposal A: All AI systems used in hiring, credit, housing, or criminal justice must publish an annual algorithmic impact assessment, audited by an accredited third party.
- Proposal B: Individuals have a right to request a human review of any AI-driven decision that significantly affects them, within 30 days, at no cost.
- Proposal C: AI companies must report any incident in which their system produced a harmful output affecting more than 1,000 people within 72 hours to a federal agency.
- For each proposal, write one substantive paragraph:
- 1. The strongest argument FOR the proposal.
- 2. The strongest argument AGAINST it.
- 3. A modification that would address the strongest objection while preserving the proposal's core purpose.
- Conclude with a two-paragraph synthesis: which proposal do you consider highest priority and why, and what single additional provision would most strengthen the overall bill?
- Draw explicitly on concepts from at least five different lessons in this module. Label them (e.g., 'as discussed in Lesson 4 on transparency...'). Minimum length: 900 words.