AI Concerns, Ethical Principles, and Responsible AI CRAGE Practice Questions (Page 8)

Questions 36–40

1. 36expert · hard

   A university uses an AI system to screen student applications. The system was trained on historical admission data. An audit reveals that the system gives lower scores to applicants from certain high schools, which are predominantly attended by minority students. The university wants to increase diversity while maintaining academic standards. Which approach best achieves this goal?

   • ARetrain the model with a fairness constraint that ensures similar admission rates across high schools, while monitoring first-year GPA and retention.
   • BAdd a bonus to applicants from underrepresented high schools, but keep the model otherwise unchanged.
   • CUse the model only as a first filter and have human reviewers make final decisions.
   • DRemove high school from the model and rely on standardized test scores.
   Reveal Solution Discussion Select an answer first
2. 37expert · hard

   A social media company is using an AI to recommend content. The AI is optimized to maximize user engagement, which leads to the spread of misinformation and polarizing content. The company's leadership is concerned about the societal impact but also wants to maintain user growth. Which approach best balances these concerns?

   • AKeep the algorithm as is, but add warning labels to content that is flagged as misinformation.
   • BModify the recommendation algorithm to include a diversity metric that reduces the spread of misinformation, while monitoring engagement metrics.
   • CIncrease the engagement objective to make the algorithm more effective.
   • DDisable the recommendation algorithm entirely and show content in chronological order.
   Reveal Solution Discussion Select an answer first
4. 38application · medium

   An AI system used by a recruitment firm ranks job applicants. The firm discovers that the system gives lower scores to applicants who graduated from certain universities. The firm wants to ensure fairness without sacrificing the system's ability to identify qualified candidates. Which approach best achieves this?

   • AUse a different AI vendor that claims their system is fair.
   • BAdd a penalty term to the model that forces equal average scores across all universities.
   • CEvaluate the model's predictions against actual job performance to determine if the university-based disparity reflects genuine differences in qualification.
   • DRemove the university name from the model inputs and assume the problem is solved.
   Reveal Solution Discussion Select an answer first
5. 39application · medium

   A social media company uses an AI content moderation system that flags posts for review. The system is found to have a higher false-positive rate for posts written in certain dialects. The company's responsible AI team is asked to address this. Which action most directly addresses the technical concern of bias while respecting user privacy?

   • ACollect additional labeled examples of those dialects and retrain the moderation model with a focus on dialect-specific performance.
   • BPublish the false-positive rates by dialect on a public transparency report.
   • CIncrease the sensitivity threshold for all posts to reduce false positives globally.
   • DDisable automated moderation for those dialects and rely on human review only.
   Reveal Solution Discussion Select an answer first
6. 40expert · hard

   A rideshare company uses an AI system to set dynamic pricing. The system is accused of surge pricing during emergencies, which some call price gouging. The company's responsible AI team must respond. The team has limited time and must balance public perception, legal risk, and business revenue. Which action best addresses the ethical concern while maintaining business viability?

   • ADisable dynamic pricing entirely to avoid any appearance of price gouging.
   • BKeep the current system but add a message that prices reflect supply and demand.
   • CConduct a full audit of the pricing algorithm before making any changes.
   • DImplement a price ceiling during declared emergencies and publish the policy in advance.
   Reveal Solution Discussion Select an answer first