Module Check: The Risk Landscape
You have worked through the full landscape of AI risk: why the field deserves serious study, how to classify risks into a principled taxonomy, what makes misuse risks distinct from accident risks and both distinct from structural risks, how to distinguish near-term documented harms from longer-horizon concerns, how to reason under genuine uncertainty about high-stakes outcomes, and how to weigh AI's substantial benefits against its risks in a way that is honest about tradeoffs. This module check is not a memory exercise. It is a test of whether these ideas have become tools you can think with — tools you can apply to new scenarios, use to evaluate arguments, and build into analytical conclusions of your own.
Flashcards — click each card to reveal the answer
Module Check Quizzes
An AI system trained to approve mortgage applications learns to associate certain zip codes with high default risk. The zip codes correlate with race due to historical residential segregation. The system was never told to use race as a factor. Which failure mechanism is most precisely responsible, and which risk category applies?
A researcher argues: 'We should invest heavily in AI safety research now, even though advanced AI systems capable of causing catastrophic harm do not yet exist, because preparing governance and technical safeguards requires significant lead time.' Which risk-reasoning principle most directly supports this argument?
A policy proposal would require AI companies to publish the demographic performance breakdown of their systems before deployment in any public-sector context. Which risk category is this policy primarily designed to address, and which failure mechanism does it target?
A company that builds widely used AI infrastructure argues against regulation by pointing out that its AI systems have never intentionally caused harm. A regulator responds that intentionality is not the relevant standard. Which concept from this module most directly supports the regulator's position?
A student reviewing an AI risk assessment of an autonomous vehicle writes: 'The main risk is that someone might hack the vehicle and use it as a weapon.' A second student adds: 'But the vehicle's collision avoidance system might also fail in heavy rain due to lidar performance degradation, causing accidents with no malicious actor involved.' How do these two risks differ in the taxonomy, and why does the distinction matter?
An analyst argues that AI-driven automation is not a structural risk because historically, new technology has always created more jobs than it destroyed. A critic says this historical argument does not resolve the structural risk concern. What is the strongest version of the critic's position?
Final Synthesis: Letter to a Policymaker
- You have spent this module building a rigorous understanding of the AI risk landscape. Now synthesize it.
- Write a formal letter (750-1000 words) to a hypothetical national policymaker who is about to vote on a broad AI governance framework. The policymaker has asked you — as a student who has studied AI risk seriously — for your honest assessment of the most important things they should understand before voting.
- Your letter must address all five of the following:
- 1. The taxonomy: Explain the three-category risk taxonomy in terms accessible to a non-technical policymaker. Use one concrete real-world example for each category. Explain why the distinctions matter for policy design.
- 2. Near-term and long-term: Distinguish risks that are happening now from those that require extrapolation. Tell the policymaker which near-term risks you believe are most urgently underaddressed, and why. Tell them which long-term risks you believe justify precautionary investment now, and why.
- 3. Reasoning under uncertainty: Explain to the policymaker how to think about risks whose probability cannot be reliably estimated. Name one cognitive trap they should guard against in evaluating expert testimony about AI risk. Give them a practical rule for how much uncertainty is acceptable before acting precautionarily.
- 4. Benefits and honest tradeoffs: Make the case that the goal of AI governance is not to eliminate AI risk but to improve the tradeoff. Give one example of a policy that would reduce a specific risk without eliminating the benefit, and one example of an overly broad restriction that would eliminate both. Explain what makes the difference.
- 5. Your recommendation: Tell the policymaker the single most important thing an AI governance framework should include, in your judgment. Justify this recommendation using the specific concepts and evidence from this module. Acknowledge one legitimate counterargument and explain why you nonetheless hold your position.
- Tone: rigorous, honest, and constructive. You are not an alarmist and you are not an AI booster — you are a careful analyst who wants good outcomes. Write accordingly.
- After completing your letter, exchange with a classmate. Your classmate should evaluate: (a) Is the taxonomy explanation accurate and accessible? (b) Is the near-term/long-term distinction clearly drawn? (c) Is the benefit-risk tradeoff genuinely balanced? (d) Is the recommendation specific and well-justified? Provide written feedback on each dimension.