Back to Subject

Unit 3: Introduction to Responsible AI - Practice Quiz

CSG202 — Generative Ai Fundamentals 60 Questions
0 Correct 0 Wrong 60 Left
0/60

1 Why did Google establish a set of AI principles?

Understand why Google has put AI principles in place Easy
A. To guide the responsible development and use of AI
B. To reduce the cost of cloud storage
C. To eliminate the need for human employees
D. To increase the speed of model training only

2 AI principles help an organization primarily address which type of concern?

Understand why Google has put AI principles in place Easy
A. Ethical and societal concerns
B. Server cooling concerns
C. Office furniture selection
D. Employee vacation scheduling

3 Which of the following best reflects a goal of Google's AI principles?

Understand why Google has put AI principles in place Easy
A. AI should avoid all testing
B. AI should replace all decisions
C. AI should be kept secret
D. AI should be socially beneficial

4 Google's AI principles include a commitment to avoid creating or reinforcing what?

Understand why Google has put AI principles in place Easy
A. Larger datasets
B. Unfair bias
C. Faster networks
D. Cheaper hardware

5 Having published AI principles helps build what with users and the public?

Understand why Google has put AI principles in place Easy
A. Debt
B. Confusion
C. Delay
D. Trust

6 What is a responsible AI practice within an organization?

Identify the need for a responsible AI practice within an organization Easy
A. A structured approach to developing and using AI ethically
B. A way to increase advertising revenue only
C. A method to speed up code compilation
D. A tool for managing office supplies

7 Why do organizations need a responsible AI practice?

Identify the need for a responsible AI practice within an organization Easy
A. To eliminate documentation
B. To avoid ever training new models
C. To manage risks and ensure AI is used ethically
D. To remove all human oversight

8 A responsible AI practice can help an organization reduce which of the following?

Identify the need for a responsible AI practice within an organization Easy
A. Potential harm from AI systems
B. The need for any strategy
C. The number of available features
D. The amount of teamwork required

9 Without responsible AI practices, an organization is more likely to face which outcome?

Identify the need for a responsible AI practice within an organization Easy
A. Automatic regulatory approval
B. Unintended negative consequences
C. Guaranteed higher profits
D. Perfect model accuracy

10 Responsible AI practices are best described as which of the following for an organization?

Identify the need for a responsible AI practice within an organization Easy
A. An optional afterthought
B. A one-time setup
C. An ongoing commitment
D. A single software download

11 At which stage of an AI project can decisions affect responsible AI outcomes?

Recognize that decisions made at all stages of a project have an impact on responsible AI Easy
A. Only after user complaints
B. Only during marketing
C. At every stage of the project
D. Only at the final deployment

12 Choosing which data to use for training an AI model is a decision that affects what?

Recognize that decisions made at all stages of a project have an impact on responsible AI Easy
A. Only the project budget
B. Responsible AI outcomes
C. Nothing important
D. Only the color of the interface

13 Which statement about responsible AI and project decisions is correct?

Recognize that decisions made at all stages of a project have an impact on responsible AI Easy
A. Small early decisions can have large downstream effects
B. Responsible AI is decided only once at launch
C. Only the CEO's decisions matter
D. Decisions never affect fairness

14 Who in an AI project should consider responsible AI?

Recognize that decisions made at all stages of a project have an impact on responsible AI Easy
A. Only external auditors
B. Only the legal team
C. Only the data scientists
D. Everyone involved across all stages

15 During which activity should responsible AI considerations be applied?

Recognize that decisions made at all stages of a project have an impact on responsible AI Easy
A. Only when errors occur
B. Only the design phase
C. Design, development, and deployment
D. Only during budgeting

16 Can organizations tailor their AI approach to their own values?

Recognize that organizations can design AI to fit their own business needs and values Easy
A. No, all AI must be identical everywhere
B. Only governments can define AI values
C. Yes, AI can be designed to fit an organization's needs and values
D. AI cannot reflect any values

17 When creating AI principles, an organization should reflect its own what?

Recognize that organizations can design AI to fit their own business needs and values Easy
A. Unrelated industries' rules
B. Random guidelines
C. Competitors' logos
D. Mission and values

18 Which of the following is true about designing responsible AI in different organizations?

Recognize that organizations can design AI to fit their own business needs and values Easy
A. Different organizations may adopt different responsible AI approaches that reflect their unique goals, culture, and the specific customers and communities they aim to serve
B. Values play no role in AI design
C. All organizations must copy one fixed template
D. Only technical teams decide values

19 A company's AI principles should ideally align with its overall what?

Recognize that organizations can design AI to fit their own business needs and values Easy
A. Software license count
B. Parking policy
C. Business strategy
D. Office layout

20 Why is it useful for an organization to define its own AI values?

Recognize that organizations can design AI to fit their own business needs and values Easy
A. To hide AI from all users
B. To avoid using AI entirely
C. To guide consistent, responsible decisions
D. To copy another company exactly

21 A company adopts Google's AI principles as a model. Which of the following best explains the primary purpose behind establishing such principles?

Understand why Google has put AI principles in place Medium
A. To provide a shared ethical framework that guides how AI is developed and applied
B. To reduce the cloud computing costs associated with training models
C. To guarantee that AI models achieve the highest possible accuracy scores
D. To ensure AI products are released to market faster than competitors

22 Google's AI principles include a list of applications it will not pursue. What does the inclusion of these prohibitions primarily demonstrate?

Understand why Google has put AI principles in place Medium
A. That regulatory bodies mandate specific banned use cases
B. That responsible AI requires setting explicit boundaries on acceptable uses
C. That prohibited uses are always the most technically difficult ones
D. That AI technology is generally too dangerous to deploy commercially

23 An engineer argues that because an AI model is technically feasible, it should be built. How would Google's AI principles respond to this reasoning?

Understand why Google has put AI principles in place Medium
A. Technical feasibility alone is insufficient; ethical and societal impact must also be weighed
B. Feasibility is the only meaningful criterion for approving a project
C. Projects should be approved if they pass automated fairness tests
D. Any feasible model should proceed unless it violates a specific law

24 Which scenario best reflects Google's principle that AI should "be socially beneficial"?

Understand why Google has put AI principles in place Medium
A. Ensuring the model maximizes user engagement metrics
B. Weighing the likely overall benefits against foreseeable risks before deploying a model
C. Releasing the model publicly as soon as it functions
D. Restricting the model to internal company use only

25 An organization deploys AI widely but has no formal responsible AI practice. Which outcome is most likely as a direct result?

Identify the need for a responsible AI practice within an organization Medium
A. Inconsistent, ad-hoc decisions that increase the risk of harm and reputational damage
B. Reduced need for human oversight in AI decisions
C. Faster and more reliable model performance across all teams
D. Automatic compliance with all future AI regulations

26 Why is it insufficient for an organization to treat responsible AI as the sole responsibility of a single ethics officer?

Identify the need for a responsible AI practice within an organization Medium
A. Responsible AI must be embedded across roles and stages, not isolated in one person
B. A single officer would be too expensive to employ full time
C. One person cannot legally sign off on AI deployments
D. Ethics officers lack the authority to review technical code

27 A startup wants to build trust with its customers around its AI products. Which action best supports this goal through a responsible AI practice?

Identify the need for a responsible AI practice within an organization Medium
A. Deploying models quickly to demonstrate innovation
B. Establishing clear governance, transparency, and accountability processes for AI
C. Keeping all model details confidential to protect trade secrets
D. Advertising that its AI is more advanced than competitors' products

28 Which statement best captures why a responsible AI practice is a continuous effort rather than a one-time checklist?

Identify the need for a responsible AI practice within an organization Medium
A. AI systems and their impacts evolve, requiring ongoing monitoring and reassessment
B. Continuous review is required solely to satisfy investor reporting
C. Models degrade only once and can then be permanently certified
D. Regulations require a new checklist to be completed every fiscal year

29 An organization notices that different teams interpret "fairness" in conflicting ways. What does a responsible AI practice provide to address this?

Identify the need for a responsible AI practice within an organization Medium
A. A legal exemption from fairness-related complaints
B. A directive to avoid using the term fairness entirely
C. A single automated tool that removes all bias from data
D. Shared definitions, standards, and processes to align decision-making

30 During data collection, a team omits certain demographic groups from the dataset. At which later stage will this decision most likely cause responsible AI concerns?

Recognize that decisions made at all stages of a project have an impact on responsible AI Medium
A. Model performance and fairness, producing biased outcomes for underrepresented groups
B. Only during the initial data storage stage
C. It will have no impact if the model achieves high overall accuracy
D. Only in the user interface design stage

31 Which example best illustrates that responsible AI decisions occur even at the problem-framing stage?

Recognize that decisions made at all stages of a project have an impact on responsible AI Medium
A. Naming the project appropriately prevents downstream bias
B. Choosing which outcome to optimize can embed unintended values or harms
C. Framing has no effect since bias only enters through data
D. Selecting a cloud provider determines the model's fairness

32 A team performs excellent bias testing during model training but ignores how the model is monitored after deployment. Why is this problematic for responsible AI?

Recognize that decisions made at all stages of a project have an impact on responsible AI Medium
A. Post-deployment monitoring is only relevant for cost tracking
B. Real-world data can shift over time, causing new harms that go undetected
C. Monitoring is unnecessary once the model passes validation
D. Training-stage testing permanently guarantees fair outcomes

33 When labeling training data, annotators apply subjective judgments inconsistently. What responsible AI risk does this decision introduce?

Recognize that decisions made at all stages of a project have an impact on responsible AI Medium
A. Increased transparency of the model's reasoning
B. A permanent reduction in model training speed
C. A guaranteed improvement in fairness metrics
D. Label bias that the model learns and reproduces in its predictions

34 A product manager decides to skip documenting model limitations to save time. Which responsible AI stage does this decision undermine?

Recognize that decisions made at all stages of a project have an impact on responsible AI Medium
A. Model architecture design, which requires no user-facing information
B. Hardware selection, which documentation does not affect
C. Data preprocessing, where limitations are irrelevant
D. Deployment and communication, where users need to understand appropriate use

35 Which choice best demonstrates that even evaluation metric selection is a responsible AI decision?

Recognize that decisions made at all stages of a project have an impact on responsible AI Medium
A. Metric choice affects only the training time, not fairness
B. Relying only on overall accuracy can hide poor performance for minority subgroups
C. Any single metric fully captures a model's societal impact
D. Evaluation metrics are chosen after deployment and have no ethical weight

36 Two companies adopt responsible AI but define their guidelines differently. What does this variation primarily reflect?

Recognize that organizations can design AI to fit their own business needs and values Medium
A. Differences indicate that responsible AI has no consistent meaning
B. Responsible AI guidelines are legally required to be identical
C. One company must be misinterpreting responsible AI standards
D. Organizations can tailor AI principles to their own values, context, and business needs

37 A healthcare firm and a retail firm each build their own responsible AI guidelines. Which approach is most appropriate?

Recognize that organizations can design AI to fit their own business needs and values Medium
A. Each defines principles reflecting its own risks, domain, and organizational values
B. Both should copy Google's principles verbatim without modification
C. Both should avoid principles until governments issue mandates
D. Each should adopt whichever framework is cheapest to implement

38 Why might an organization's responsible AI values legitimately differ from those of another organization?

Recognize that organizations can design AI to fit their own business needs and values Medium
A. Values must differ to comply with international standards
B. Only the largest companies are permitted to define their own values
C. Responsible AI values are randomly assigned to each company
D. Different missions, stakeholders, and industries shape which values are prioritized

39 An organization wants its responsible AI principles to be genuinely effective rather than symbolic. Which practice best achieves this?

Recognize that organizations can design AI to fit their own business needs and values Medium
A. Assigning the principles only to the marketing department
B. Reviewing the principles once at project completion
C. Publishing the principles on the website and taking no further action
D. Integrating the principles into everyday decisions, processes, and accountability

40 A company aligns its AI design choices with its stated corporate value of "customer privacy first." Which decision is most consistent with this value?

Recognize that organizations can design AI to fit their own business needs and values Medium
A. Minimizing data collection and applying strong data protection safeguards
B. Deferring all privacy decisions to external vendors
C. Collecting maximum data to improve future model accuracy
D. Sharing customer data broadly to enable faster innovation

41 A company adopts Google's AI principles verbatim but treats them purely as a public-relations statement, with no operational review process. Which critique best captures the fundamental flaw in this approach?

Understand why Google has put AI principles in place Hard
A. AI principles create measurable accountability only when embedded in governance, review, and decision-making processes rather than serving as aspirational statements
B. AI principles are only relevant for companies that build foundation models, not those that consume them
C. AI principles are legally binding contracts that require external auditors to enforce them across all jurisdictions
D. AI principles should never be published because doing so exposes proprietary model architectures to competitors

42 Google's AI principles include both objectives (e.g., 'be socially beneficial') and a list of applications it will not pursue. Why is the explicit exclusion list strategically significant?

Understand why Google has put AI principles in place Hard
A. It guarantees that no competitor can enter those excluded markets first
B. It sets concrete boundaries that constrain product decisions, making the principles enforceable rather than purely aspirational
C. It transfers all legal liability for misuse from Google to downstream developers
D. It ensures the principles satisfy every national regulation automatically

43 Consider a scenario where an AI feature technically satisfies every stated principle but still produces a discriminatory outcome for a subgroup not anticipated during design. What does this reveal about the role of AI principles?

Understand why Google has put AI principles in place Hard
A. Discriminatory outcomes are acceptable if all documented principles were followed
B. Principles guide intent but require ongoing testing and monitoring because they cannot pre-enumerate every real-world harm
C. Principles fully guarantee fair outcomes once a system passes the initial design review
D. The principles were incorrectly written and must be discarded entirely

44 Why does Google frame its AI principles as a living framework subject to revision rather than a fixed one-time policy?

Understand why Google has put AI principles in place Hard
A. Because competitors force annual changes through industry agreements
B. Because revising them frequently reduces the cost of legal compliance
C. Because a fixed policy would be impossible to publish on a website
D. Because AI capabilities, societal expectations, and identified harms evolve, requiring principles to adapt over time

45 An organization has a strong data-science team but no formal responsible AI practice. It repeatedly ships models that later require costly recalls due to bias and trust failures. Which analysis best explains the root cause?

Identify the need for a responsible AI practice within an organization Hard
A. Responsible AI is irrelevant here because the failures are purely technical accuracy issues
B. The data-science team lacks sufficient computing resources to train larger models
C. The models were trained on too little data, which a responsible AI practice cannot address
D. Absence of a structured responsible AI practice means ethical review is ad hoc and reactive rather than systematic and preventive

46 A leadership team argues that responsible AI can be handled entirely by a single compliance officer at the end of the pipeline. What is the strongest objection to this staffing model?

Identify the need for a responsible AI practice within an organization Hard
A. Responsibility must be distributed across roles and stages because harms originate throughout the lifecycle, not only at final review
B. Compliance officers legally cannot review machine learning systems
C. Responsible AI concerns only marketing and therefore belongs in that department
D. A single officer would be too expensive compared to hiring more engineers

47 Two competing startups build similar recommendation systems. One invests early in a responsible AI practice; the other defers it to 'after product-market fit.' Under what condition is the deferring startup's decision most likely to backfire?

Identify the need for a responsible AI practice within an organization Hard
A. When the product never reaches any users at all
B. When the responsible AI practice would have doubled the size of the engineering team
C. When investors specifically require an ethics certificate before any funding round
D. When embedded biases and trust issues become entrenched in the product and infrastructure, making later remediation expensive and disruptive

48 Which statement best distinguishes a mature responsible AI practice from mere regulatory compliance?

Identify the need for a responsible AI practice within an organization Hard
A. The two are identical because all ethical concerns are already codified in law
B. A responsible AI practice only tracks legally mandated documentation and audit trails
C. A responsible AI practice proactively addresses ethical values and stakeholder trust beyond the minimum required by law
D. Regulatory compliance always exceeds the scope of any internal responsible AI practice

49 An enterprise notices that responsible AI initiatives stall because engineers see them as blocking velocity. What organizational approach most effectively resolves this tension?

Identify the need for a responsible AI practice within an organization Hard
A. Removing all responsible AI checks to preserve development speed
B. Assigning responsible AI exclusively to an isolated ethics board that reviews finished products
C. Requiring engineers to attend a one-time training and then never revisiting the topic
D. Integrating responsible AI checks into existing workflows and tooling so they become part of normal development rather than an external gate

50 During problem framing, a team defines success solely as maximizing click-through rate. Later stages faithfully optimize this metric, yet the product amplifies sensational misinformation. At which stage did the responsible AI failure originate?

Recognize that decisions made at all stages of a project have an impact on responsible AI Hard
A. The problem-framing stage, where the objective definition failed to account for downstream societal harms
B. No stage is responsible because the model correctly optimized its objective
C. The data-collection stage, because more data would have prevented the issue
D. The deployment stage, since that is where users actually saw the content

51 A team collects a demographically skewed dataset but applies rigorous fairness auditing only at the evaluation stage. Why might this fail to produce a responsible outcome?

Recognize that decisions made at all stages of a project have an impact on responsible AI Hard
A. Because harms introduced during data collection can constrain what any later-stage audit can detect or correct
B. Because evaluation-stage auditing is always more accurate than data-stage auditing
C. Because skewed datasets have no effect once the model is trained
D. Because fairness auditing should only ever be performed after deployment

52 Consider a project where each individual stage decision seems locally reasonable, yet the deployed system behaves unfairly. What does this best illustrate about responsible AI?

Recognize that decisions made at all stages of a project have an impact on responsible AI Hard
A. Responsible AI requires evaluating cumulative and interaction effects across stages, not just judging each decision in isolation
B. Only the final deployment decision determines whether a system is responsible
C. Locally reasonable decisions always guarantee a globally fair system
D. Unfair behavior can only result from an obviously unethical single decision

53 A model performs equitably in testing but a deployment-stage decision to set a single global confidence threshold degrades accuracy for a minority subgroup. What principle does this scenario reinforce?

Recognize that decisions made at all stages of a project have an impact on responsible AI Hard
A. Testing equity guarantees deployment equity regardless of configuration
B. Only the training algorithm can introduce fairness problems
C. Deployment and configuration decisions are as consequential to responsible AI as modeling decisions
D. Threshold settings are purely cosmetic and never affect fairness

54 Why is documenting decisions and their rationale at every project stage important for responsible AI, beyond mere record-keeping?

Recognize that decisions made at all stages of a project have an impact on responsible AI Hard
A. It primarily serves to increase the volume of paperwork for audits
B. It replaces the need to test the model for bias or performance
C. It enables tracing how and where harms were introduced, supporting accountability and targeted correction
D. It shifts responsibility away from the team to the documentation itself

55 A team faces pressure to skip the data-representativeness review to meet a deadline, reasoning that later monitoring will catch any issues. What is the strongest reason this trade-off is risky for responsible AI?

Recognize that decisions made at all stages of a project have an impact on responsible AI Hard
A. Later monitoring is guaranteed to catch every possible representativeness problem
B. Skipping the review has no effect because data quality is fixed at collection
C. Monitoring is always technically impossible for representativeness issues
D. Monitoring detects harms only after they affect users, whereas an upstream review can prevent them from reaching users at all

56 Two hospitals adopt the same responsible AI framework but implement it differently: one prioritizes maximum recall for a diagnostic model, the other prioritizes precision. Which conclusion is most defensible?

Recognize that organizations can design AI to fit their own business needs and values Hard
A. Only the recall-maximizing hospital is acting responsibly since sensitivity always matters most
B. The framework was applied incorrectly since responsible AI mandates identical configurations
C. Neither is responsible because a single universal metric threshold must apply to all organizations
D. Both can be responsible because organizations tailor AI trade-offs to their specific context, values, and risk tolerance

57 An organization claims it can design AI 'to fit its own values' and uses this to justify weakening fairness safeguards for profit. Why is this reasoning a misapplication of the principle?

Recognize that organizations can design AI to fit their own business needs and values Hard
A. Organizations are never permitted to customize any aspect of their AI systems
B. Values-based customization applies only to user interface design, not model behavior
C. Tailoring AI to organizational values must operate within ethical and responsible boundaries, not override protections against harm
D. Profit motives automatically make any design choice responsible by definition

58 A financial firm and a creative-writing platform both deploy generative models but define acceptable-output policies very differently. What best explains why both approaches can be legitimate?

Recognize that organizations can design AI to fit their own business needs and values Hard
A. The financial firm must always adopt the platform's more permissive policy for consistency
B. Only heavily regulated industries are allowed to define custom output policies
C. Any difference in policy between organizations indicates one of them is irresponsible
D. Responsible AI allows context-specific values and risk profiles to shape acceptable-use boundaries for each organization

59 When an organization translates a broad principle like 'fairness' into its own concrete operational definition, what is the most important safeguard to prevent misuse of this flexibility?

Recognize that organizations can design AI to fit their own business needs and values Hard
A. Adopting whatever definition minimizes engineering effort regardless of impact
B. Delegating the definition entirely to the model to decide at runtime
C. Making the chosen definition transparent and justifiable to stakeholders so customization does not become a cover for harm
D. Keeping the definition confidential to protect competitive advantage

60 An organization wants its AI to reflect a specific value of 'user privacy above engagement.' Which design consequence most directly follows from genuinely embedding this value?

Recognize that organizations can design AI to fit their own business needs and values Hard
A. Ignoring privacy in the model but adding a privacy disclaimer to the interface
B. Maximizing data collection to improve personalization while labeling it as privacy-respecting
C. Treating privacy and engagement as identical objectives requiring no trade-off
D. Accepting reduced personalization or data collection even when it lowers engagement metrics
Previous MCQ
Unit 2
Next MCQ
Unit 4