Correct Answer: Emotions can be difficult to interpret
Explanation:
Emotional signals vary between people and cultures, so AI interpretations may be uncertain or inaccurate.
Incorrect! Try again.
16How can AI support mental health services?
AI and mental health
Easy
A.By replacing all mental health professionals
B.By diagnosing every person with certainty
C.By making personal support unnecessary
D.By helping screen for concerns
Correct Answer: By helping screen for concerns
Explanation:
AI may assist with early screening, but qualified professionals should make clinical decisions.
Incorrect! Try again.
17Why must AI mental health tools be used carefully?
AI and mental health
Easy
A.They can replace emergency services
B.They never handle sensitive information
C.They understand every person perfectly
D.They may give inaccurate advice
Correct Answer: They may give inaccurate advice
Explanation:
AI tools can misunderstand situations, so users should not treat them as a complete replacement for professional care.
Incorrect! Try again.
18What type of information is especially sensitive in AI mental health applications?
AI and mental health
Easy
A.Names of common software programs
B.Personal health information
C.Public weather reports
D.General textbook definitions
Correct Answer: Personal health information
Explanation:
Mental health information is private and should be protected through strong privacy and security practices.
Incorrect! Try again.
19What does AI governance refer to?
AI and governance
Easy
A.The speed of internet connections
B.The construction of research offices
C.The design of computer keyboards
D.Rules for responsible AI use
Correct Answer: Rules for responsible AI use
Explanation:
AI governance includes policies, laws, and practices that guide the development and use of AI.
Incorrect! Try again.
20Which principle is commonly included in AI governance?
AI and governance
Easy
A.Uncontrolled deployment
B.Accountability
C.Permanent avoidance of evaluation
D.Unlimited secrecy
Correct Answer: Accountability
Explanation:
Accountability means that people and organizations remain responsible for the effects of AI systems.
Incorrect! Try again.
21A social research organization adopts an AI system that summarizes interview transcripts. Which change is most likely in the researchers' work?
AI and the future of work in social science fields
Medium
A.Researchers can spend more time validating themes and developing explanations
B.Researchers will no longer need to interpret interview findings
C.Researchers should accept all automatically identified themes as objective
D.Researchers will only need technical programming skills
Correct Answer: Researchers can spend more time validating themes and developing explanations
Explanation:
AI can reduce time spent on routine summarization, but human researchers are still needed to interpret context, assess meaning, and validate conclusions.
Incorrect! Try again.
22A university uses an AI tool to screen job applications for a social science research project. What is the most important human responsibility?
AI and the future of work in social science fields
Medium
A.Selecting applicants only from the largest demographic group
B.Checking whether the system produces unfair patterns across applicant groups
C.Removing all human review to improve processing speed
D.Assuming the system is neutral because it uses data
Correct Answer: Checking whether the system produces unfair patterns across applicant groups
Explanation:
AI systems can reproduce or amplify bias in historical data, so human oversight is needed to evaluate fairness and accountability.
Incorrect! Try again.
23Which skill would best help a sociologist remain effective as AI becomes more common in social research?
AI and the future of work in social science fields
Medium
A.Combining domain knowledge with data literacy and ethical judgment
B.Memorizing every output produced by a single AI system
C.Replacing interviews entirely with automatically generated surveys
D.Avoiding digital tools to preserve traditional research methods
Correct Answer: Combining domain knowledge with data literacy and ethical judgment
Explanation:
Future social science work will require understanding social contexts, evaluating data and algorithms, and making responsible decisions about technology.
Incorrect! Try again.
24An AI Ethics Analyst is asked to review a hiring algorithm that rejects applicants from a particular neighborhood at a higher rate. What should the analyst do first?
Career opportunities: AI Ethics Analyst
Medium
A.Investigate the data, criteria, and possible reasons for the disparity
B.Delete all neighborhood-related data without further analysis
C.Publish the names of rejected applicants to create pressure
D.Approve the system because its overall accuracy is high
Correct Answer: Investigate the data, criteria, and possible reasons for the disparity
Explanation:
A disparity may indicate bias, but it must be investigated by examining data quality, model features, decision rules, and relevant social context.
Incorrect! Try again.
25Which task is most directly associated with the role of an AI Ethics Analyst?
Career opportunities: AI Ethics Analyst
Medium
A.Installing computer networks for a research laboratory
B.Designing advertisements for a technology company
C.Evaluating risks related to privacy, fairness, and accountability
D.Writing fictional stories about future robots
Correct Answer: Evaluating risks related to privacy, fairness, and accountability
Explanation:
AI Ethics Analysts assess how AI systems affect people and institutions, especially regarding fairness, privacy, transparency, and responsibility.
Incorrect! Try again.
26A predictive policing system performs equally well overall but has more false positives in one community. What concern should an AI Ethics Analyst raise?
Career opportunities: AI Ethics Analyst
Medium
A.False positives matter only when the system uses images
B.The system may distribute burdens unfairly across communities
C.The system is automatically fair because its average accuracy is equal
D.Community differences should be ignored to protect efficiency
Correct Answer: The system may distribute burdens unfairly across communities
Explanation:
Overall performance can hide unequal error rates, and higher false positives for one group may lead to disproportionate surveillance or intervention.
Incorrect! Try again.
27A research assistant uses an AI tool to identify themes in focus-group transcripts. Which practice improves the reliability of the analysis?
Research Assistant with AI tools
Medium
A.Comparing AI-generated themes with human coding and original evidence
B.Using the first set of themes without reviewing the transcripts
C.Treating frequent words as proof of the most important theme
D.Removing comments that do not fit the AI-generated categories
Correct Answer: Comparing AI-generated themes with human coding and original evidence
Explanation:
Human review and comparison with source material help identify missing context, misleading patterns, and incorrect interpretations.
Incorrect! Try again.
28Before uploading confidential interview transcripts to an external AI service, what should a research assistant do?
Research Assistant with AI tools
Medium
A.Check consent, privacy rules, and the service's data-use policies
B.Replace participant names but ignore all other identifying details
C.Assume encryption removes the need for participant permission
D.Upload the files immediately because research data is always public
Correct Answer: Check consent, privacy rules, and the service's data-use policies
Explanation:
Confidential data requires careful review of informed consent, de-identification, institutional rules, and how the provider stores or uses data.
Incorrect! Try again.
29An AI assistant generates a citation for a journal article, but the research assistant cannot locate the article. What is the best response?
Research Assistant with AI tools
Medium
A.Replace the citation with a popular website on the same topic
B.Include the citation but mark it as probably accurate
C.Verify the source through trusted databases before citing it
D.Use the citation because academic-looking details are usually reliable
Correct Answer: Verify the source through trusted databases before citing it
Explanation:
AI tools can generate fabricated or inaccurate references, so every citation should be checked against reliable scholarly sources.
Incorrect! Try again.
30An NGO uses AI to identify households that may need food assistance. Which approach is most responsible?
AI in NGOs
Medium
A.Publish all household data so donors can inspect the model
B.Use the model as the only basis for distributing aid
C.Combine model results with community review and an appeal process
D.Exclude households with incomplete digital records automatically
Correct Answer: Combine model results with community review and an appeal process
Explanation:
AI can help prioritize support, but human review, local knowledge, and appeal mechanisms reduce errors and protect vulnerable groups.
Incorrect! Try again.
31An NGO working in disaster relief has internet access only in some areas. Which AI deployment plan is most suitable?
AI in NGOs
Medium
A.Use flexible tools that support offline work and later synchronization
B.Require continuous internet access before providing any assistance
C.Collect more personal data to compensate for weak connectivity
D.Replace field workers with a cloud-based chatbot
Correct Answer: Use flexible tools that support offline work and later synchronization
Explanation:
Offline-capable systems are more practical and inclusive in areas with limited connectivity, while still allowing information to be updated later.
Incorrect! Try again.
32An NGO wants to use an AI chatbot to answer questions about legal aid. Which safeguard is most important?
AI in NGOs
Medium
A.Providing clear limits, referral options, and human escalation
B.Allowing the chatbot to make binding legal decisions
C.Presenting the chatbot as a substitute for all legal professionals
D.Collecting sensitive information before explaining its purpose
Correct Answer: Providing clear limits, referral options, and human escalation
Explanation:
Legal guidance can have serious consequences, so users should know the system's limits and have access to qualified human support.
Incorrect! Try again.
33A company uses facial analysis to estimate employees' emotions during meetings. Which limitation should decision-makers consider most carefully?
Trends: Emotional AI
Medium
A.Facial expressions always reveal a person's true emotions
B.Employees cannot be affected by automated emotional judgments
C.Emotional expressions can vary across cultures and situations
D.Emotion models are accurate whenever enough cameras are used
Correct Answer: Emotional expressions can vary across cultures and situations
Explanation:
Emotional AI may misinterpret expressions because emotion is influenced by culture, context, individual differences, and the quality of available data.
Incorrect! Try again.
34A school plans to use voice analysis to identify students who are emotionally distressed. What is a major ethical concern?
Trends: Emotional AI
Medium
A.Voice recordings contain no information about individuals
B.Students will always understand how predictions are produced
C.The system may make sensitive inferences without valid consent
D.Emotional states can be measured with complete certainty
Correct Answer: The system may make sensitive inferences without valid consent
Explanation:
Inferring emotional or mental states from voice involves sensitive personal information and requires consent, transparency, and careful validation.
Incorrect! Try again.
35A mental health chatbot detects language suggesting that a user may be at immediate risk of self-harm. What should the system be designed to do?
AI and mental health
Medium
A.Delete the conversation to avoid responsibility
B.Continue casual conversation without changing its response
C.Offer crisis resources and connect the user with human support
D.Diagnose the user and prescribe medication automatically
Correct Answer: Offer crisis resources and connect the user with human support
Explanation:
High-risk situations require timely support and escalation to qualified professionals or emergency services rather than relying only on automated conversation.
Incorrect! Try again.
36A study finds that an AI mental health tool works well in one urban population. What should researchers do before recommending broad use?
AI and mental health
Medium
A.Remove human professionals to reduce inconsistent judgments
B.Assume the results apply equally to every population
C.Test performance with diverse groups and relevant clinical oversight
D.Measure success only by the number of conversations completed
Correct Answer: Test performance with diverse groups and relevant clinical oversight
Explanation:
Effectiveness may differ across languages, cultures, and populations, so broader validation and professional oversight are necessary.
Incorrect! Try again.
37Why can overreliance on an AI wellness app be harmful for some users?
AI and mental health
Medium
A.The app may provide confident but unsuitable advice for complex needs
B.Digital tools are incapable of storing any user information
C.Users always recognize when an automated recommendation is wrong
D.Wellness apps prevent people from seeking human assistance
Correct Answer: The app may provide confident but unsuitable advice for complex needs
Explanation:
AI tools may lack clinical understanding of an individual's situation and can produce inappropriate guidance if users treat them as professional care.
Incorrect! Try again.
38A government agency plans to use AI to determine eligibility for public benefits. Which policy would best support accountability?
AI and governance
Medium
A.Keeping the model completely secret from affected citizens
B.Allowing the model to change eligibility rules without oversight
C.Providing explanations, review procedures, and a way to appeal decisions
D.Measuring success only by the number of cases processed
Correct Answer: Providing explanations, review procedures, and a way to appeal decisions
Explanation:
People affected by government decisions need transparency, meaningful review, and opportunities to challenge errors or unfair outcomes.
Incorrect! Try again.
39Which governance approach is most appropriate for a high-risk AI system used in criminal justice?
AI and governance
Medium
A.Reliance on vendor claims without access to performance evidence
B.Risk assessment, independent auditing, and continuous monitoring
C.Permanent approval after a successful pilot test
D.Minimal oversight because public institutions are automatically trustworthy
Correct Answer: Risk assessment, independent auditing, and continuous monitoring
Explanation:
High-risk systems require ongoing evaluation because errors, bias, and social impacts may change after deployment.
Incorrect! Try again.
40A city is drafting rules for the use of facial recognition in public spaces. Which question should policymakers prioritize?
AI and governance
Medium
A.Whether the technology eliminates the need for police judgment
B.Whether residents can identify the brand of camera being used
C.Whether the system can be purchased at the lowest price
D.Whether its benefits justify privacy, fairness, and civil-liberty risks
Correct Answer: Whether its benefits justify privacy, fairness, and civil-liberty risks
Explanation:
AI governance should weigh public benefits against possible harms, including surveillance, discrimination, privacy loss, and reduced freedom.
Incorrect! Try again.
41A public-opinion research team uses an AI system to code interview transcripts. The system reduces coding time by 70% but systematically misclassifies dialectal expressions as hostile language. Which redesign best addresses the central sociotechnical risk?
AI and the future of work in social science fields
Hard
A.Replace human coders with a larger language model
B.Remove dialectal interviews from the training dataset
C.Allow participants to correct only the final published report
D.Use representative data, subgroup audits, and human review
Correct Answer: Use representative data, subgroup audits, and human review
Explanation:
The problem involves representational bias and unequal error rates. Diverse data, subgroup evaluation, and meaningful human oversight address both technical and institutional dimensions.
Incorrect! Try again.
42A university predicts that AI will eliminate most entry-level social science jobs because automated tools can summarize literature and analyze surveys. Which conclusion is most defensible?
AI and the future of work in social science fields
Hard
A.AI will affect only quantitative researchers, not qualitative researchers
B.AI adoption will have no employment effect if researchers remain accountable
C.AI will eliminate tasks but may increase demand for validation and interpretation
D.AI will replace professional judgment whenever prediction accuracy improves
Correct Answer: AI will eliminate tasks but may increase demand for validation and interpretation
Explanation:
Automation usually changes task composition rather than eliminating entire occupations. Interpretation, contextual judgment, ethical review, and accountability remain important.
Incorrect! Try again.
43A social science department wants to evaluate whether AI adoption has improved research quality rather than merely increased output. Which metric set is strongest?
AI and the future of work in social science fields
Hard
A.Publication count, staff satisfaction, and number of automated tasks
B.Accuracy alone, regardless of subgroup or downstream consequences
C.Validity, subgroup error rates, reproducibility, and stakeholder impact
D.Number of reports, model speed, and software subscriptions
Correct Answer: Validity, subgroup error rates, reproducibility, and stakeholder impact
Explanation:
Research quality includes whether findings are valid, reproducible, equitable across groups, and useful or safe for affected stakeholders.
Incorrect! Try again.
44An AI Ethics Analyst discovers that a welfare-screening model has equal overall accuracy for two demographic groups but denies eligible applicants more often in one group. What should the analyst investigate first?
Equal accuracy can conceal different false-positive and false-negative rates. In welfare screening, false negatives may deny essential support and therefore require direct fairness analysis.
Incorrect! Try again.
45A vendor claims that its hiring model is fair because it does not use race, gender, or disability status. Which response is most appropriate for an AI Ethics Analyst?
Career opportunities: AI Ethics Analyst
Hard
A.Accept the claim because protected attributes are excluded
B.Reject the system automatically because all prediction is unethical
C.Approve the system if its average accuracy exceeds human accuracy
D.Test for proxy variables and outcome disparities across groups
Correct Answer: Test for proxy variables and outcome disparities across groups
Explanation:
Features such as location, employment history, or institution can act as proxies. Fairness requires examining outcomes and error patterns, not only the input list.
Incorrect! Try again.
46A facial-emotion system is proposed for employee evaluation. Which recommendation best reflects an ethics assessment based on validity and power asymmetry?
Career opportunities: AI Ethics Analyst
Hard
A.Use it if employees sign a general workplace consent form
B.Permit it because emotional states are visible behavioral data
C.Require evidence that inferred emotions are valid and non-coercive
D.Deploy it only after increasing camera resolution
Correct Answer: Require evidence that inferred emotions are valid and non-coercive
Explanation:
Facial expressions do not reliably reveal inner emotional states, and workplace monitoring creates coercive conditions. Validity and voluntariness must be established before deployment.
Incorrect! Try again.
47A research assistant uses a generative AI tool to identify studies for a systematic review. The tool produces plausible but nonexistent citations. Which workflow is most rigorous?
Research Assistant with AI tools
Hard
A.Ask the same tool to rate the reliability of its citations
B.Use citations that contain realistic authors and journal names
C.Verify every citation against authoritative bibliographic databases
D.Keep only citations that appear in at least two AI-generated lists
Correct Answer: Verify every citation against authoritative bibliographic databases
Explanation:
Generative systems can hallucinate references. Independent verification using reliable databases and the original publications is necessary.
Incorrect! Try again.
48A researcher asks an AI assistant to summarize confidential interviews about political persecution. Which practice most effectively reduces privacy risk while preserving research utility?
Research Assistant with AI tools
Hard
A.Replace participant names but retain rare locations and dates
B.Remove direct identifiers and use an approved secure environment
C.Upload the files because the tool promises not to share them
D.Ask the model to forget the interviews after generating summaries
Correct Answer: Remove direct identifiers and use an approved secure environment
Explanation:
De-identification should address quasi-identifiers as well as names, and sensitive material should be processed only in an institutionally approved environment.
Incorrect! Try again.
49An AI tool assigns themes to interview data with high agreement against a researcher's initial coding. Why is this agreement insufficient evidence of validity?
Research Assistant with AI tools
Hard
A.Qualitative data cannot be analyzed with computational methods
B.AI systems are valid only when they produce numerical predictions
C.High agreement proves that the themes are objectively correct
D.Agreement may reproduce the researcher's unexamined coding assumptions
Correct Answer: Agreement may reproduce the researcher's unexamined coding assumptions
Explanation:
Agreement with one coder measures consistency with that coder, not construct validity. Researchers should compare alternative interpretations and examine disconfirming cases.
Incorrect! Try again.
50An NGO uses a predictive model to prioritize households for emergency aid. Historical data underrepresent undocumented migrants, although current need is high in that population. What is the greatest immediate concern?
AI in NGOs
Hard
A.Predictive models cannot process household-level administrative data
B.The model may amplify historical undercoverage into present exclusion
C.Humanitarian decisions are always less fair than automated decisions
D.The model will necessarily overfund undocumented migrants
Correct Answer: The model may amplify historical undercoverage into present exclusion
Explanation:
Models learn from available data. If past records omit a vulnerable group, predictions may systematically underprioritize that group.
Incorrect! Try again.
51An NGO evaluates an AI-supported food-distribution program during a drought. Which evaluation design best separates algorithmic effects from changing environmental conditions?
AI in NGOs
Hard
A.Compare this year's recipients with last year's recipients only
B.Survey staff opinions after deployment and treat agreement as impact
C.Measure the model's accuracy using its own recommended outcomes
D.Use a pre-specified comparison design with outcome and subgroup measures
Correct Answer: Use a pre-specified comparison design with outcome and subgroup measures
Explanation:
A credible evaluation needs a comparison strategy, defined outcomes, and subgroup analysis to distinguish program effects from external changes and distributional harms.
Incorrect! Try again.
52An NGO wants to use publicly available social-media posts to locate disaster survivors. Which governance condition is most important before deployment?
AI in NGOs
Hard
A.The posts must be written in a widely supported language
B.The NGO must assess consent, surveillance risks, and false positives
C.The model must be trained only on posts from previous disasters
D.The system must maximize the number of detected accounts
Correct Answer: The NGO must assess consent, surveillance risks, and false positives
Explanation:
Public availability does not eliminate ethical obligations. Unwanted identification, surveillance, and mistaken interventions can harm survivors.
Incorrect! Try again.
53A company claims that an AI can infer employee stress from voice pitch across cultures. Which finding would most seriously challenge the claim?
Trends: Emotional AI
Hard
A.The system requires more storage after additional training
B.Employees prefer written surveys to recorded interviews
C.Stress-related vocal patterns vary by language and social context
D.The model runs more slowly on mobile devices
Correct Answer: Stress-related vocal patterns vary by language and social context
Explanation:
If vocal cues depend strongly on culture and language, a model trained in one context may not validly infer stress in another.
Incorrect! Try again.
54A school proposes using emotional AI to identify disengaged students from webcam images during online classes. Which objection is strongest?
Trends: Emotional AI
Hard
A.Students cannot display any emotion during online instruction
B.Automated systems cannot classify more than two emotional categories
C.Facial expressions are imperfect indicators and monitoring may alter behavior
D.Webcam images are always too low-resolution for machine learning
Correct Answer: Facial expressions are imperfect indicators and monitoring may alter behavior
Explanation:
Emotional inference from appearance has uncertain validity, while surveillance can create stress, misclassification, and behavioral changes that undermine the intended measure.
Incorrect! Try again.
55A chatbot for mental-health support performs well in average user ratings but occasionally gives unsafe advice to users expressing suicidal intent. Which deployment policy is most responsible?
AI and mental health
Hard
A.Remove all crisis-related conversations from evaluation data
B.Add crisis detection, escalation pathways, and continuous safety audits
C.Deploy widely because average satisfaction is high
D.Allow the chatbot to provide unrestricted advice after a disclaimer
Rare but severe failures require risk-sensitive safeguards. Crisis escalation, human support, and ongoing monitoring are more appropriate than relying on average ratings.
Incorrect! Try again.
56An AI mental-health app predicts depression from phone-use patterns. Users consent to data collection, but the prediction is shared with employers for wellness planning. What is the primary ethical issue?
AI and mental health
Hard
A.Employers are always legally prohibited from receiving aggregate statistics
B.The prediction may be computationally less efficient than a clinical interview
C.Phone-use patterns cannot contain any information about mental health
D.Consent for personal support may not authorize employment-related secondary use
Correct Answer: Consent for personal support may not authorize employment-related secondary use
Explanation:
Consent must be specific and meaningful. A user agreeing to support does not necessarily authorize disclosure that could affect employment or create pressure.
Incorrect! Try again.
57A government agency adopts an opaque risk score to allocate inspections. Officials argue that accountability is unnecessary because the model is statistically accurate. Which governance requirement is most justified?
AI and governance
Hard
A.Replace the model with random inspection to eliminate all bias
B.Require explainability, appeal mechanisms, and responsibility assignment
C.Publish source code but provide no process for affected people
D.Permit secrecy if the model's accuracy exceeds a legal threshold
Correct Answer: Require explainability, appeal mechanisms, and responsibility assignment
Explanation:
Accuracy does not resolve due process or accountability concerns. People affected by consequential decisions need understandable reasons, review, and identifiable responsibility.
Incorrect! Try again.
58A city regulates a predictive-policing system only by requiring an annual accuracy report. Why is this policy inadequate?
AI and governance
Hard
A.Annual reporting always produces more data than agencies can store
B.Predictive policing has no possible statistical evaluation
C.Aggregate accuracy can conceal unequal burdens and feedback loops
Correct Answer: Aggregate accuracy can conceal unequal burdens and feedback loops
Explanation:
Police data reflect prior enforcement patterns. Aggregate accuracy may hide disparate impacts, while deployment can reinforce the same patterns that generated the data.
Incorrect! Try again.
59A national AI policy requires impact assessments only for systems classified as high risk before deployment. Which additional measure best addresses risks that emerge after deployment?
AI and governance
Hard
A.Evaluate only whether the system remains commercially profitable
B.Ban updates after the first approved release
C.Require monitoring, incident reporting, audits, and reassessment
D.Allow developers to self-certify all later system changes
Correct Answer: Require monitoring, incident reporting, audits, and reassessment
Explanation:
Real-world data, users, and institutions change. Post-deployment oversight is needed to detect drift, unexpected harms, and new risk pathways.
Incorrect! Try again.
60An international NGO deploys an AI system across countries with different privacy laws and cultural norms. Which governance strategy is most robust?
AI and governance
Hard
A.Let the technology vendor determine which rules are relevant
B.Use one universal consent statement translated literally
C.Combine baseline protections with context-specific legal and community review
D.Apply the least restrictive country's rules everywhere
Correct Answer: Combine baseline protections with context-specific legal and community review
Explanation:
A common minimum standard can protect rights, but local law, cultural expectations, and affected-community input are necessary for legitimate and safe implementation.
Incorrect! Try again.
Did this save you a night before the exam?
LPU Notes is free, and it stays free. Ads cover part of the server bill.
The rest comes out of a student's own pocket: the domain, the storage,
and keeping the site up through the weeks everyone needs it at once.
The payment button didn't load. An ad blocker or a filtered network is the usual reason.
to try again.
Nothing here is ever locked, and nothing unlocks. Chip in only if it was worth it.
What it pays for →