D429 Introduction to AI for Computer Scientists - Set 2 - Part 2

Correct Answer: Fine-grained recognition models

Option b is correct because fine-grained recognition specializes in distinguishing between visually similar subcategories like different insect species effectively.
Option a is incorrect because binary classification oversimplifies the complex taxonomy of insects, missing important distinctions between beneficial and harmful species.
Option c is incorrect because coarse category detection lacks the precision needed to distinguish between visually similar beneficial insects and pests.
Option d is incorrect because color-based sorting fails since many beneficial insects and pests share similar coloration patterns in nature.

Correct Answer: Multi-criteria optimization

Option b is correct because multi-criteria optimization can balance factors like severity, resource availability, and response time across multiple emergencies.
Option a is incorrect because random selection ignores critical factors like severity and available resources when lives depend on proper prioritization.
Option c is incorrect because first-call priority might allocate all resources to less severe incidents while more critical emergencies await assistance.
Option d is incorrect because static ranking cannot adapt to the dynamic nature of multiple simultaneous disasters with changing severity levels.

Correct Answer: Jurisdiction-aware processing

Option b is correct because jurisdiction-aware processing can apply appropriate privacy rules based on each customer's location and applicable laws dynamically.
Option a is incorrect because applying the lowest privacy standard violates stricter regulations in many jurisdictions, risking legal penalties and fines.
Option c is incorrect because universal handling either violates some laws or unnecessarily restricts functionality where regulations are less stringent.
Option d is incorrect because manual review doesn't scale effectively and introduces human error in complex multi-jurisdictional compliance requirements.

Correct Answer: Adaptive interaction modeling

Option b is correct because adaptive modeling can adjust interaction styles based on individual children's responses and therapeutic progress over time.
Option a is incorrect because rigid scripts cannot respond appropriately to the varied and unpredictable behaviors of children with autism spectrum conditions.
Option c is incorrect because random behaviors would confuse children who benefit from predictable, structured interactions in therapeutic settings consistently.
Option d is incorrect because minimal engagement fails to provide the rich interactive experiences needed for effective autism therapy interventions.

Correct Answer: Analogical reasoning frameworks

Option b is correct because analogical reasoning can identify patterns across domains, generating novel hypotheses while maintaining connections to established knowledge.
Option a is incorrect because free association without constraints generates hypotheses that lack grounding in scientific principles and existing research foundations.
Option c is incorrect because random combinations produce mostly nonsensical hypotheses without the structured reasoning needed for scientific validity and testing.
Option d is incorrect because strict deductive logic alone cannot generate novel hypotheses, only deriving conclusions from existing premises mechanically.

Correct Answer: Balanced dataset sampling

Option b is correct because balanced sampling ensures the model learns features relevant to all demographic groups, improving cross-ethnic accuracy equally.
Option a is incorrect because focusing on majority groups perpetuates bias, resulting in poor recognition accuracy for underrepresented ethnic populations significantly.
Option c is incorrect because training on a single demographic creates models that fail to generalize across different ethnic facial features.
Option d is incorrect because minimal diversity in training data leads to biased models that perform poorly for many ethnic groups.

Correct Answer: Semantic similarity search

Option b is correct because semantic search understands legal concepts and relationships, finding relevant precedents even with different terminology used.
Option a is incorrect because exact string matching misses relevant cases that discuss similar legal principles using different words or phrasing.
Option c is incorrect because date filtering alone doesn't identify topically relevant precedents, missing important cases based solely on timing.
Option d is incorrect because random sampling provides no mechanism for identifying legally relevant precedents from the vast document collection available.

Correct Answer: Intrinsic motivation rewards

Option b is correct because intrinsic rewards can make learning inherently engaging while still ensuring coverage of required educational content effectively.
Option a is incorrect because rigid enforcement without considering engagement leads to student disinterest and poor learning outcomes despite curriculum coverage.
Option c is incorrect because external scores alone don't address engagement, potentially creating test-focused learning without deep understanding or interest.
Option d is incorrect because punishment-based approaches decrease motivation and create negative associations with learning, harming long-term educational outcomes.

Correct Answer: Ensemble method combination

Option b is correct because ensemble methods combine multiple models' strengths, detecting diverse fraud patterns across different transaction types effectively.
Option a is incorrect because single models often miss fraud patterns specific to certain transaction types or financial instruments they weren't optimized for.
Option c is incorrect because rule-based detection cannot adapt to new fraud techniques or identify complex patterns across multiple transaction types.
Option d is incorrect because manual pattern matching doesn't scale to the volume and complexity of modern financial transactions requiring analysis.

Correct Answer: Probabilistic imputation methods

Option b is correct because probabilistic imputation estimates missing values based on patterns while quantifying uncertainty in treatment recommendations appropriately.
Option a is incorrect because ignoring missing values can lead to incorrect treatment recommendations that don't account for important patient factors.
Option c is incorrect because deleting incomplete records would eliminate most real-world patient data, making the system practically unusable clinically.
Option d is incorrect because replacing missing values with zeros creates false data that could lead to dangerous treatment recommendations.

Correct Answer: Dynamic replanning algorithms

Option b is correct because dynamic replanning can quickly compute alternative routes when unexpected obstacles appear during drone flight operations.
Option a is incorrect because static planning cannot respond to unexpected obstacles that weren't present when the original route was calculated.
Option c is incorrect because fixed waypoint navigation lacks flexibility to deviate when obstacles block the predetermined waypoint sequence path.
Option d is incorrect because predetermined paths force drones to halt or crash when unexpected obstacles appear along the planned route.

Correct Answer: Medical knowledge verification

Option b is correct because verification against authoritative medical sources ensures generated content aligns with established health information standards consistently.
Option a is incorrect because unrestricted generation could produce dangerous medical misinformation that might cause serious harm to users' health.
Option c is incorrect because prioritizing speed over accuracy in medical content generation risks producing harmful misinformation quickly without verification.
Option d is incorrect because minimal filtering allows potentially dangerous medical advice to reach users who might follow harmful recommendations.

Correct Answer: Temporal feature creation

Option b is correct because temporal features capture how customer behavior evolves over time, indicating engagement trends and future value.
Option a is incorrect because raw data without engineering misses important patterns like purchase frequency acceleration or declining engagement over time.
Option c is incorrect because single snapshots lose the temporal dynamics that indicate whether customers are becoming more or less valuable.
Option d is incorrect because static averages obscure important trends and patterns in customer behavior that predict lifetime value changes.

Correct Answer: Pareto optimization frontier

Option b is correct because Pareto optimization identifies solutions where improving one objective requires sacrificing another, revealing optimal trade-offs clearly.
Option a is incorrect because maximizing single objectives ignores the necessary balance between environmental and economic considerations in urban planning decisions.
Option c is incorrect because random goal selection provides no systematic way to balance competing objectives or find optimal compromise solutions.
Option d is incorrect because processing goals sequentially might optimize early goals while severely compromising later objectives in planning decisions.

Correct Answer: Confidence threshold system

Option b is correct because confidence thresholds can identify when the AI's uncertainty indicates human expertise would provide better assistance quality.
Option a is incorrect because never escalating leaves customers frustrated with complex issues the AI cannot adequately resolve without human intervention.
Option c is incorrect because random escalation disrupts customer service flow without any relationship to actual problem complexity or AI capabilities.
Option d is incorrect because always escalating defeats the purpose of AI assistance and overwhelms human agents with simple, manageable queries.

Correct Answer: Stylistic pattern analysis

Option b is correct because stylistic analysis can identify satirical writing patterns and humor markers that distinguish satire from deceptive misinformation.
Option a is incorrect because fact checking alone cannot determine intent, treating satirical exaggeration the same as deliberate deceptive misinformation.
Option c is incorrect because blocking entire sources would eliminate legitimate satire, which serves important social commentary and entertainment functions.
Option d is incorrect because keyword systems cannot understand context and tone that differentiate humorous satire from harmful fake news.

Correct Answer: Distributed consensus protocols

Option b is correct because distributed consensus allows robots to coordinate effectively while maintaining resilience if individual robots fail or disconnect.
Option a is incorrect because centralized control creates a single failure point that could disable the entire robot team in dangerous environments.
Option c is incorrect because independent operation prevents the coordination needed for complex disaster response tasks requiring robot collaboration effectively.
Option d is incorrect because master-slave architecture makes the system vulnerable to failure if the master robot becomes damaged or disconnected.

Correct Answer: Counterfactual explanations

Option b is correct because counterfactual explanations show applicants what specific changes would result in loan approval, providing actionable feedback effectively.
Option a is incorrect because neural network weights are incomprehensible to loan applicants and provide no actionable information for improvement strategies.
Option c is incorrect because statistical coefficients alone don't translate into understandable actions applicants can take to improve their applications.
Option d is incorrect because black box predictions provide no explanation at all, leaving applicants confused about denial reasons and improvements.

Correct Answer: Irregular sampling models

Option b is correct because irregular sampling models are specifically designed to handle sensor data arriving at varying frequencies without information loss.
Option a is incorrect because assuming fixed intervals ignores valuable sensor readings that don't align with predetermined time slots, losing information.
Option c is incorrect because deleting data to achieve regular intervals throws away valuable sensor information critical for failure prediction.
Option d is incorrect because simple interpolation can introduce artifacts and false patterns that don't reflect actual equipment behavior accurately.

Correct Answer: Response time patterns

Option b is correct because analyzing response times can reveal suspiciously fast answers indicating guessing rather than thoughtful problem solving approaches.
Option a is incorrect because score totals alone cannot distinguish between lucky guesses and actual knowledge demonstration in student responses.
Option c is incorrect because checking single answers doesn't reveal patterns that distinguish systematic guessing from genuine knowledge-based responses effectively.
Option d is incorrect because grade averaging obscures the specific response patterns that indicate guessing versus actual understanding of material.