D429 Introduction to AI for Computer Scientists - Set 2 - Part 1

Correct Answer: Partially observable environment

Option b is correct because failed sensors create partial observability, requiring the agent to reason with incomplete information about package weights.
Option a is incorrect because sensor failures mean the agent cannot fully observe all relevant package weight information needed for decisions.
Option c is incorrect because the environment involves dynamic package movement and sensor failures don't relate to static or deterministic properties.
Option d is incorrect because warehouse operations are sequential and continuous, not episodic, and sensor observability differs from discrete classifications.

Correct Answer: Online learning capability

Option b is correct because online learning allows the system to continuously update its model as new spam patterns emerge in real-time.
Option a is incorrect because static rules cannot adapt to new spam tactics without manual intervention to update the detection patterns.
Option c is incorrect because offline batch processing delays adaptation, allowing new spam tactics to succeed until the next batch update.
Option d is incorrect because fixed feature extraction cannot discover new characteristics of evolving spam tactics that emerge over time.

Correct Answer: Linear temporal logic

Option b is correct because linear temporal logic can represent and reason about sequences of events and future states in urban planning.
Option a is incorrect because static semantic networks cannot adequately represent the temporal evolution of population and transportation patterns over time.
Option c is incorrect because simple graph structures lack the temporal reasoning capabilities needed for projecting future transportation needs effectively.
Option d is incorrect because basic rule systems cannot capture the complex temporal dependencies between current usage and future growth patterns.

Correct Answer: Hybrid data fusion

Option c is correct because hybrid fusion combines structured customer metrics with insights from unstructured claim narratives for comprehensive risk assessment.
Option a is incorrect because using only structured data misses valuable risk indicators contained in detailed claim description narratives and explanations.
Option b is incorrect because text mining alone ignores crucial structured data like payment history, demographics, and quantitative risk factors.
Option d is incorrect because manual review doesn't scale effectively and lacks the systematic analysis capabilities of automated hybrid approaches.

Correct Answer: Name randomization blinding

Option a is correct because name randomization prevents the model from using name-based assumptions while preserving actual qualification data for assessment.
Option b is incorrect because removing all data would eliminate the ability to assess any candidate qualifications for recruitment decisions.
Option c is incorrect because amplifying correlations would worsen the bias problem rather than mitigating name-based discrimination in recruitment.
Option d is incorrect because adding more demographic data could introduce additional biases rather than solving the name discrimination issue.

Correct Answer: Bayesian belief updating

Option b is correct because Bayesian methods can combine prior seasonal patterns with new evidence about disruptions to update predictions dynamically.
Option a is incorrect because deterministic forecasting cannot account for the uncertainty introduced by unexpected weather events and other disruptions.
Option c is incorrect because moving averages smooth historical data but cannot explicitly model irregular events or update beliefs accordingly.
Option d is incorrect because fixed seasonal models cannot adapt to unexpected disruptions that deviate from regular seasonal patterns observed historically.

Correct Answer: Privacy-preserving activity detection

Option b is correct because activity detection can identify safety concerns like falls without invasive continuous visual monitoring of residents.
Option a is incorrect because continuous video recording violates privacy and dignity, creating discomfort for elderly residents in their living spaces.
Option c is incorrect because complete absence of monitoring fails to provide the safety benefits needed for vulnerable elderly residents' protection.
Option d is incorrect because limiting surveillance to public areas only misses critical safety events that occur in private resident rooms.

Correct Answer: Federated learning architecture

Option b is correct because federated learning trains models locally at each hospital and only shares model updates, preserving patient privacy.
Option a is incorrect because centralized pooling requires sharing sensitive patient data, violating privacy regulations and hospital data policies.
Option c is incorrect because public data sharing of medical records violates patient privacy laws and ethical guidelines for healthcare data.
Option d is incorrect because cloud consolidation still requires transferring sensitive patient data outside hospital systems, risking privacy breaches.

Correct Answer: Contextual semantic analysis

Option b is correct because semantic analysis examines meaning and context to distinguish educational historical content from harmful ideology promotion.
Option a is incorrect because keyword filtering cannot distinguish between historical documentation and promotion based on words alone without context.
Option c is incorrect because binary rules lack the nuance to differentiate between legitimate historical discussion and harmful content promotion.
Option d is incorrect because frequency-based detection counts occurrences without understanding the contextual purpose or framing of the content.

Correct Answer: Multi-sensor fusion validation

Option b is correct because fusing multiple sensors like cameras, lidar, and radar helps validate true obstacles versus visual artifacts.
Option a is incorrect because simple thresholding cannot reliably distinguish between real obstacles and visual artifacts like shadows or markings.
Option c is incorrect because single camera systems are susceptible to visual illusions that multi-sensor validation would correctly identify as false.
Option d is incorrect because color detection alone cannot determine whether something is a physical obstacle or just a road marking.

Correct Answer: Sentiment context modeling

Option b is correct because contextual sentiment models can detect subtle emotional cues beyond literal word meanings in customer expressions.
Option a is incorrect because keyword lists miss the contextual nuances that transform neutral words into expressions of frustration or sarcasm.
Option c is incorrect because frequency analysis counts words without understanding the emotional context or sarcastic intent behind their usage.
Option d is incorrect because simple polarity scoring treats all negative words equally, missing the subtlety of sarcasm and passive-aggressive communication.

Correct Answer: Existence uncertainty modeling

Option b is correct because existence uncertainty can model whether reliable delivery information actually exists or if estimates are fabricated.
Option a is incorrect because assuming perfect information ignores the reality that suppliers might provide false or unreliable delivery estimates intentionally.
Option c is incorrect because identity certainty doesn't address the uncertainty about whether the delivery information itself is truthful or accurate.
Option d is incorrect because deterministic planning cannot handle the uncertainty introduced by potentially unreliable supplier delivery information effectively.

Correct Answer: Reward shaping design

Option b is correct because reward shaping can provide graduated feedback that guides the system toward appropriately challenging problem generation.
Option a is incorrect because sparse rewards provide insufficient feedback for fine-tuning the delicate balance between challenge and achievability in problems.
Option c is incorrect because random rewards provide no meaningful signal about whether problem difficulty is appropriately calibrated for student learning.
Option d is incorrect because fixed reward schedules cannot adapt to individual student capabilities or provide nuanced difficulty adjustment feedback.

Correct Answer: Layered explanation system

Option b is correct because layered explanations can provide technical details for doctors while offering accessible summaries for patients' understanding.
Option a is incorrect because technical jargon excludes patients from understanding their own medical conditions and treatment recommendations effectively.
Option c is incorrect because simplified terms alone may lack the precision and detail that medical professionals need for clinical decision-making.
Option d is incorrect because statistical outputs without interpretation fail to provide meaningful explanations for either doctors or patients effectively.

Correct Answer: Adaptive baseline modeling

Option b is correct because adaptive baselines adjust to normal market conditions while detecting deviations that suggest potential manipulation patterns.
Option a is incorrect because fixed thresholds cannot distinguish between legitimate high volatility periods and actual manipulation in dynamic markets.
Option c is incorrect because historical averaging smooths out both normal and suspicious patterns, making manipulation detection less effective overall.
Option d is incorrect because volume tracking alone misses sophisticated manipulation techniques that maintain normal volumes while distorting prices strategically.

Correct Answer: Empathetic resource guidance

Option b is correct because providing empathetic support with professional resource recommendations respects both user needs and AI limitations appropriately.
Option a is incorrect because AI assistants should not provide diagnostic conclusions that require professional mental health expertise and clinical training.
Option c is incorrect because treatment recommendations exceed appropriate AI boundaries and could potentially cause harm without professional oversight involved.
Option d is incorrect because dismissing mental health concerns fails to provide needed support and could worsen the user's situation significantly.

Correct Answer: Value alignment constraints

Option b is correct because value alignment ensures the system's optimization remains consistent with human values and broader organizational goals.
Option a is incorrect because prioritizing speed could accelerate harmful single-minded optimization without considering negative consequences or side effects.
Option c is incorrect because unchecked pursuit directly enables the paperclip maximizer problem of optimizing one goal while destroying everything else.
Option d is incorrect because focusing solely on efficiency ignores the broader impact and value considerations that prevent harmful optimization behaviors.

Correct Answer: Exploration bonus inclusion

Option b is correct because exploration bonuses encourage the system to recommend diverse content beyond established user preference patterns occasionally.
Option a is incorrect because maximum similarity matching reinforces filter bubbles by only showing content similar to previous user choices consistently.
Option c is incorrect because pure exploitation strategies maximize immediate engagement without introducing the diversity needed to prevent filter bubbles.
Option d is incorrect because using only historical preferences perpetuates filter bubbles by never exposing users to different content types.

Correct Answer: Structured skill testing

Option b is correct because structured tests can evaluate specific technical skills directly without being confounded by writing or language abilities.
Option a is incorrect because essay evaluation heavily depends on writing ability, potentially disadvantaging technically skilled candidates with different language backgrounds.
Option c is incorrect because free-form interviews mix communication skills with technical knowledge, making it difficult to isolate technical competence alone.
Option d is incorrect because written applications conflate writing quality with technical skills, potentially creating bias against non-native speakers unfairly.

Correct Answer: Temporal consistency checking

Option b is correct because temporal consistency analysis can detect unnatural movements and transitions that indicate deepfake video manipulation techniques.
Option a is incorrect because color histograms analyze color distribution but miss the temporal artifacts that reveal sophisticated deepfake manipulations effectively.
Option c is incorrect because simple frame comparison lacks the sophistication to detect subtle temporal inconsistencies in advanced deepfake videos.
Option d is incorrect because resolution measurement alone cannot identify the complex manipulation artifacts present in high-quality deepfake videos today.