Unit 5: Responsible AI Principles - Practice Quiz

1 In Responsible AI, fairness primarily aims to ensure that an AI system does what?

2 Which of the following is an example of unfair AI behavior?

3 Transparency in Responsible AI mainly refers to which of the following?

4 Which practice best supports transparency in an AI system?

5 Accountability in Responsible AI means that:

6 Which of the following best reflects accountability in AI development?

7 What is the main purpose of prototyping in design thinking?

8 In the design thinking process, testing a prototype is mainly done to:

9 Prototyping is valuable because it allows teams to:

10 A low-fidelity prototype is best described as one that is:

11 A high-fidelity prototype is one that:

12 The term fidelity in prototyping refers to:

13 Which statement best describes an assumption?

14 A hypothesis in design thinking is best described as:

15 What is the key difference between an assumption and a hypothesis?

16 Why is testing with customers important in design thinking?

17 When testing with customers, the best approach is to:

18 When communicating innovation to stakeholders, the goal is to:

19 Which tool is commonly used to communicate an innovative idea to stakeholders?

20 Ethical design in AI focuses on building systems that:

21 A loan-approval AI model approves men at a rate of 80% and women at a rate of 55% for applicants with identical financial profiles. Which fairness concern is most directly demonstrated?

22 A hospital deploys an AI that recommends treatments but cannot explain why it made a recommendation to clinicians. Which Responsible AI principle is most compromised?

23 An autonomous delivery robot causes property damage. Establishing who is answerable for the harm reflects which Responsible AI principle?

24 During design thinking, a team builds a quick paper mockup to gather feedback before writing any code. What is the primary purpose of this prototype?

25 A team wants to test the overall navigation flow of an app early in the process without committing to visual details. Which prototype fidelity is most appropriate?

26 Which statement best distinguishes a high-fidelity prototype from a low-fidelity one?

27 A team states: "We believe busy professionals will pay for a meal-prep app because they lack cooking time." To convert this into a testable hypothesis, what must be added?

28 Which of the following is an assumption rather than a hypothesis?

29 While testing a prototype, a designer asks: "Don't you love this new feature?" Why is this problematic?

30 During usability testing, which approach gives the most reliable insight into how customers actually use a product?

31 When presenting an innovation to non-technical executives, which approach is most effective?

32 A team uses a storytelling narrative with a customer persona to pitch a new idea. What is the main advantage of this technique?

33 A facial recognition system performs poorly on darker skin tones because its training data was mostly light-skinned faces. This is an example of which issue?

34 Embedding privacy-by-design into an AI product means privacy is:

35 A hiring algorithm is trained on historical data from a company that rarely hired women for technical roles. Without intervention, what is the likely outcome?

36 Which practice most directly increases the transparency of an AI-driven recommendation system for end users?

37 An organization creates an internal review board and audit logs to trace decisions made by its AI systems. This primarily strengthens which principle?

38 A startup spends three months building a fully polished product before showing it to any user, then discovers customers don't want it. Which prototyping principle was violated?

39 Why is it risky to build a product based on untested assumptions?

40 A team adds a human-in-the-loop review before an AI system makes final decisions about medical benefits. This design choice mainly supports:

41 A hiring model achieves equal accuracy across gender groups but selects men at twice the rate of equally qualified women. Which fairness criterion is violated, and why can it not coexist with the one already satisfied?

42 A bank deploys a complex model but provides users with post-hoc local explanations (e.g., SHAP values) rather than an inherently interpretable model. What is the most defensible critique of relying solely on this approach for high-stakes decisions?

43 An AI system causes harm through a decision made autonomously in production. The vendor, the deploying company, and the data provider each blame the others. Which governance concept most directly addresses this situation?

44 A team must test whether users can navigate a new app flow, but the visual design is still undecided. Which prototype fidelity choice best isolates the variable of interest while minimizing wasted effort?

45 Which statement correctly distinguishes an assumption from a hypothesis in a design experiment?

46 During usability testing, a facilitator notices participants keep succeeding at tasks but rating satisfaction low. What is the most rigorous next step to avoid a misleading conclusion?

47 An engineer presents an innovation to executives using detailed model architecture diagrams, but the pitch fails. Applying stakeholder-communication principles, what is the primary flaw?

48 A recommender system optimizes engagement and inadvertently amplifies polarizing content. Which responsible-AI concept best explains why a purely metric-optimized design produced this outcome?

49 A team runs five iterative prototype tests but keeps observing the same three usability issues without them ever resolving. What does this pattern most strongly indicate?

50 A model is 'calibrated' within each group (predicted probabilities match observed frequencies) yet activists argue it is still unfair. Which fairness tension best explains this?

51 A regulator requires 'meaningful information about the logic involved' for automated decisions. A company argues its deep model is too complex to explain simply. What is the most principled resolution consistent with transparency obligations?

52 Which combination of practices most directly strengthens ex-post accountability (ability to investigate after harm occurs) for a deployed AI system?

53 A well-known risk of high-fidelity prototypes in early-stage testing is that they can distort feedback. Which mechanism best explains this distortion?

54 A team writes: 'We believe busy professionals will pay $20/month for meal planning because they lack time to cook.' To convert this into a proper testable hypothesis, what must be added?

55 A startup validates demand via a landing page with a 'Buy Now' button that leads to a 'coming soon' page, measuring click-through. What is the strongest ethical/methodological critique of this 'fake door' test?

56 When communicating a disruptive innovation whose ROI is uncertain, which approach best balances honesty with securing stakeholder buy-in?

57 A facial analysis product performs well overall but poorly for a minority subgroup underrepresented in training data. Which responsible-AI practice would have most directly surfaced this before deployment?

58 A designer wants to compare two onboarding flows but the difference in observed conversion between 12 test users is small. Which conclusion is most methodologically sound?

59 To reduce bias, a team removes the 'race' feature from a model but disparities persist. Which phenomenon best explains this failure?

60 A company wants to embed ethics into its ML workflow rather than treating it as a final compliance check. Which approach best reflects 'ethics by design'?