Describe Artificial Intelligence workloads and considerations practice questions

Inclusiveness

Why wrong: Inclusiveness is the correct principle because it directs AI systems to empower every individual and community, including those speaking less-common languages. In this scenario, the corporation's goal is to serve users equitably across languages, which requires intentional data collection, design, and testing for linguistic diversity. This is not simply about avoiding harm, but about actively ensuring the system is usable and beneficial to all, which is the core mandate of inclusiveness.

Reliability and safety

Why wrong: Reliability and safety is incorrect because this principle ensures an AI system functions consistently and without harm, such as avoiding critical errors, system crashes, or unsafe behavior. The scenario concerns proactive support for diverse languages, which influences user reach and performance breadth, but does not directly involve system stability or physical/cyber safety. When a system performs poorly for some languages, the issue is a lack of inclusive design, not a reliability failure in the narrow sense of the system failing to operate as intended.

Transparency

Why wrong: Transparency is incorrect because this principle focuses on clearly disclosing an AI system's capabilities, limitations, and how decisions are made, not on expanding which languages the system can support. While a transparent system might document language coverage gaps, the scenario's proactive goal of addressing those gaps is a design and performance matter, not a communication or explainability issue. Transparency would be relevant if the question asked about communicating the system's language limitations to users, not about removing them.

Inclusiveness

Why wrong: Inclusiveness focuses on designing AI systems that are accessible to and empower diverse users, but it is not specifically about preventing discriminatory outcomes based on protected attributes like age or disability. While related, it is not the most direct principle for this scenario.

Privacy and security

Why wrong: Privacy and security addresses protecting personal data and ensuring system security, not directly relevant to avoiding discrimination in prioritization.

Reliability and safety

Why wrong: Reliability and safety ensures the system operates consistently and without harm, but does not specifically target bias or fairness in decision-making.

Accountability

Why wrong: Accountability is about taking responsibility for the system's operation. While related, the primary issue is the lack of transparency, which affects fairness. In this context, Fairness is the most directly violated principle because the decision cannot be examined for bias.

Reliability and Safety

Why wrong: Reliability and Safety focus on whether the system works reliably and safely. The scenario does not indicate any failure or safety issue.

Privacy and Security

Why wrong: Privacy and Security pertain to data protection. The scenario does not involve a privacy or security breach.

Classification

Why wrong: Classification predicts discrete categories (yes/no, cat/dog) — regression predicts continuous numeric values.

Clustering

Why wrong: Clustering groups similar items without predefined labels — regression predicts a specific numeric output.

Anomaly detection

Why wrong: Anomaly detection identifies unusual data points — regression predicts continuous values for normal predictions.

Detecting whether a customer is experiencing technical problems with a product

Why wrong: This option is incorrect because detecting whether a customer is experiencing a technical problem is a form of issue categorization or error analysis, not sentiment analysis. Sentiment analysis measures the emotional polarity of text (positive, negative, neutral), whereas identifying a specific technical issue requires intent recognition or topic classification to pinpoint the problem type. A negative sentiment score might accompany a complaint, but it does not tell you that the issue is technical or what the underlying problem is.

Predicting whether a customer will make a repeat purchase based on their browsing history

Why wrong: This option is wrong because predicting repeat purchases is a predictive analytics/machine learning task based on behavioral data such as browsing history, purchase frequency, and engagement patterns—not a text analysis task. Sentiment analysis works on written or spoken language to extract opinions or emotions; it does not infer future behavior from historical actions. Recommendation systems or churn prediction models would be appropriate for this scenario, whereas sentiment analysis would only be relevant if analyzing a customer's expressed feedback or review text.

Automatically translating customer reviews into multiple languages for international support teams

Why wrong: Translating customer reviews into multiple languages is machine translation, not sentiment analysis.

Accountability

Why wrong: Accountability concerns the assignment of responsibility to individuals and organizations for the AI system's design, development, and deployment, including governance, audit trails, and clear ownership. It ensures that when harm occurs, those responsible can be identified and held to account, but it does not by itself prevent discrimination from happening in the first place. A bank could have strong accountability structures, yet the underlying loan model might still encode bias; fairness is needed to proactively ensure equitable outcomes.

Inclusiveness

Why wrong: Inclusiveness is the design principle that aims to empower everyone by creating systems that are accessible and useful for a wide range of users, including people with disabilities or diverse cultural backgrounds. While important for user experience, inclusiveness does not directly address whether the AI's loan approval decisions are free from bias or discriminatory patterns. A system could be inclusive in its user interface yet still unfairly reject qualified applicants due to biased training data, so inclusiveness does not solve the discrimination problem.

Reliability and Safety

Why wrong: Reliability and safety focus on ensuring the AI system performs consistently and without harmful failures under expected conditions, such as avoiding outages, data corruption, or unsafe actions. These principles deal with the system's correctness and robustness over time, but they do not evaluate whether its decisions are fair across demographic groups. A loan approval model could be highly reliable and safe—operating without errors—while still systematically discriminating against certain applicants, so these principles are insufficient for addressing bias.

Fairness

Why wrong: Fairness in Responsible AI aims to eliminate algorithmic bias and ensure equitable treatment across different groups, such as ensuring a hiring or credit model does not disadvantage people based on gender, race, or other protected attributes. In a predictive maintenance scenario, the input features are typically sensor readings, machine telemetry, and usage logs—not demographic data—so the risk of discrimination against human groups is not the central concern. Fairness speaks to social impartiality, but the requirement here is about physical operational reliability, making it an incorrect choice.

Privacy & Security

Why wrong: Privacy & Security is a separate Responsible AI principle focused on protecting data confidentiality, ensuring secure access controls, and defending AI systems against adversarial attacks or unauthorized tampering. While a manufacturing AI system may process sensitive operational data, the requirement in question is about maintaining reliable prediction performance when environmental conditions change, not about safeguarding information assets. Even if the data were perfectly secure, the model could still fail under new factory conditions, so Privacy & Security does not address the core need.

Inclusiveness

Why wrong: Inclusiveness is the Responsible AI principle that requires AI to be designed for and accessible to people of all abilities, languages, and backgrounds, such as providing alternative interfaces, accommodating disabilities, or accounting for varying levels of digital literacy. The manufacturing scenario described is focused on predicting when an event (e.g., equipment failure) will occur, which is a function of machine state, not human interaction. Inclusiveness relates to human-centered accessibility, not to handling environmental variability in an industrial predictive system, so it does not apply to this requirement.

Model Interpretability

Why wrong: Model Interpretability techniques, such as SHAP values, LIME, or feature importance, explain why a specific prediction was made by attributing outcomes to input features. While these insights can reveal that a model relies heavily on a proxy for a sensitive attribute, they do not systematically compare error rates or prediction consistency across predefined protected groups. For fairness auditing, we need a quantitative, group-level comparison rather than a per-instance narrative.

Error Analysis

Why wrong: Error Analysis focuses on finding cohorts, data subsets, or 'error slices' where the model has particularly high error rates, often revealing patterns like geographic or product-type issues. Its slices are not inherently defined by protected attributes, and its goal is to improve overall accuracy or identify difficult cases. It does not enforce the specific fairness constraint of measuring bias across sensitive groups, so it is insufficient for addressing systematic demographic discrimination.

Data Balance Analysis

Why wrong: Data Balance Analysis inspects the training dataset to detect imbalances such as underrepresentation of a sensitive group or skewed label distributions across attributes. Although severe data imbalance can breed biased outcomes, this check occurs before model training and does not measure whether the final predictions actually favor or penalize any demographic segment. Predicting fairness requires evaluating post-training outputs, not merely examining input statistics.

Fairness

Why wrong: Fairness concerns the equitable treatment of patients across different demographic groups, ensuring the system does not introduce or amplify biases in diagnosis or treatment recommendations. While fairness is a critical requirement for any clinical AI, it is not the core issue indicated by the scenario, which centers on the absence of a human in the loop to take responsibility. A fully automated system could be fair and still violate accountability, because no human can be held accountable for its mistakes.

Reliability and safety

Why wrong: Reliability and safety focus on whether the system performs correctly and consistently under normal and adversarial conditions, such as avoiding misdiagnoses or maintaining safe operation. A fully automated system could in principle be designed to be highly reliable and safe, passing rigorous validation and monitoring, so the central problem is not necessarily a failure of reliability. The key deficiency is the lack of human oversight, which is directly addressed by the accountability principle.

Transparency

Why wrong: Transparency in AI refers to the degree to which the system's decisions and inner workings can be understood and audited by humans. In this scenario, a fully automated diagnosis system might still be transparent (e.g., with explainable inputs and confidence scores), so the absence of human involvement is not fundamentally a transparency violation. The principle that most directly addresses the lack of human oversight is accountability, because it requires a designated party to answer for outcomes and take corrective action.

Accountability

Why wrong: Accountability is not the right answer because it refers to assigning ownership and governance for the AI system's development, deployment, and consequences, not to communicating explanations to end users. An organization can define a clear chain of responsibility and still deploy an opaque model that cannot answer a customer's 'why' question. While accountability is essential for auditing failures and monitoring outcomes, it does not in itself create or guarantee any explanation of individual recommendations.

Fairness

Why wrong: Fairness is not the right answer because it centers on eliminating bias, ensuring equitable treatment across protected groups, and achieving parity in outcomes, rather than explaining an individual recommendation. The customer is not asking whether the decision was statistically non-discriminatory; they are asking for the specific reasons behind their personal result. Fairness is a critical attribute but is evaluated at the population or segment level, whereas transparency is what provides the case-specific narrative the customer expects.

Reliability

Why wrong: Reliability is not the right answer because it focuses on the AI system's ability to perform consistently and correctly across different inputs and operating conditions, such as avoiding crashes, latency spikes, or inconsistent scoring. A model can be extremely reliable—returning the same plausible output under normal and exceptional loads—yet remain a black box that offers no explanation for any particular recommendation. Reliability addresses system robustness and uptime, not the semantic justification of decisions requested by the customer.

Transparency

Why wrong: Transparency is about making the AI system's behavior, capabilities, and limitations understandable to stakeholders, such as through explainability techniques or clear documentation. It aims to build trust by revealing how predictions are made, not by safeguarding data. Since transparency does not involve enforcing encryption, authentication, or data retention policies, it is not the principle that protects patient information from unauthorized exposure.

Fairness

Why wrong: Fairness ensures the AI system does not produce biased outcomes or discriminate against individuals or groups based on protected attributes like race, gender, or socioeconomic status. Fairness techniques focus on mitigating algorithmic bias and achieving equitable treatment, which is crucial for healthcare equity. However, fairness has nothing to do with securing data against breaches or privacy invasions, making it irrelevant to the protection of patient data in this scenario.

Reliability and safety

Why wrong: Reliability and safety focuses on an AI system's ability to perform its intended function consistently and without causing physical or operational harm. This includes robustness against errors, drifts in performance, and fail-safe mechanisms for clinical decision support. While important, it does not govern who can access stored data or how that data is protected, so it does not address the specific concern of unauthorized access to sensitive information.

Inclusiveness

Why wrong: Inclusiveness is a design philosophy that aims to build systems usable by the widest possible range of people, including those with disabilities or differing needs. While patients with rare symptoms should be included in design and testing, the core problem here is not a lack of representation in the design process; it is that the model's decision-making is already biased against those patients. Fairness deals with the allocative harm in the outcome, whereas inclusiveness only addresses the breadth of human diversity considered during development.

Transparency

Why wrong: Transparency requires AI systems to be explainable, auditable, and communicative about how decisions are made, such as documenting that symptom frequency affects triage scores. However, making the bias visible does not remove it; a transparent system could clearly show that rare symptoms are penalized while still being unfair in practice. In this scenario, the missing element is equitable treatment, so transparency is an enabler for detecting the issue but not the principle that directly governs the biased triage outcome.

Accountability

Why wrong: Accountability refers to the governance and ownership of AI systems, including mechanisms that ensure an organization can answer for the system's behavior and remediate harm. Even with strong accountability in place, an AI that systematically disadvantages rare-symptom patients remains unfair; accountability determines who is responsible, not whether the system's decisions are morally or ethically correct. The question asks which principle is violated by the biased behavior itself, and that principle is fairness, not the post-hoc responsibility structure.

Transparency

Why wrong: Transparency focuses on clearly disclosing what the AI system can and cannot do, including its capabilities, limitations, and the data or logic behind its decisions. While transparency helps users understand or challenge outcomes, it does not by itself establish a responsible party for the system's actions. A transparent system could describe its workings in detail yet still have no designated person or process for oversight, auditing, or redress. Thus transparency is necessary but not sufficient, because it does not address human control or answerability.

Reliability and Safety

Why wrong: Reliability and safety emphasize whether the AI system performs consistently, avoids unexpected errors, and prevents physical or digital harm. These qualities address the system's technical robustness and behavior, but they do not specify who is accountable for its operation. An AI system can pass rigorous safety testing and still lack a clear human owner responsible for supervising live deployment and answering for outcomes. Accountability, in contrast, centers on organizational responsibility and human governance rather than purely technical performance.

Privacy and Security

Why wrong: Privacy and security are concerned with protecting personal data from unauthorized access, complying with data-protection laws, and defending the system against adversarial attacks or misuse. These controls safeguard the data and infrastructure, but they do not establish who is answerable for the AI system's decisions or require human oversight of its operation. A system can be strongly encrypted and privacy-compliant while still running without any human accountable for its output. Therefore privacy and security address data protection, not governance or liability, making them the wrong answer for this scenario.

Transparency

Why wrong: Transparency concerns the degree to which an AI system's behavior, decisions, and limitations are openly communicated and explainable to stakeholders, for example through model documentation or interpretability tools. Although revealing the model's reliance on zip code might help someone notice the bias, the transparency principle itself does not mandate equitable outcomes; a fully transparent model can still be clearly and openly biased. The violation in this scenario is not a lack of explanation but the discriminatory effect of the predictions, so transparency is not the best fit.

Reliability and Safety

Why wrong: Reliability and Safety focus on whether the AI system performs consistently, robustly, and without causing physical or operational harm, such as failing under unusual inputs or producing unsafe actions. A zip-code-driven bias is not primarily a malfunction in accuracy or robustness; the model may be reliably and predictably producing the same biased result every time. While unreliable systems can also cause harm, the core issue here is disparate treatment across groups rather than system failure or unsafe operation.

Privacy and Security

Why wrong: Privacy and Security govern the proper handling, protection, and authorized use of personal data, including encryption, access controls, and compliance with data-protection regulations. Using zip code as a feature is not inherently a data breach or unauthorized processing; the problem is the model's discriminatory impact, not how the data is stored or accessed. Even if the data are perfectly secured, the fairness violation persists, making privacy and security an incorrect classification for this scenario.

Accountability

Why wrong: Accountability is a governance principle that assigns responsibility to individuals or organizations for the AI system's outcomes, including monitoring and remediation after deployment. While important for maintaining trust and legal compliance, accountability does not inherently influence the chatbot's design or training data to make it more inclusive. A chatbot could be fully accountable—with clear human oversight—yet still fail to serve non-English speakers because that principle alone does not drive dataset diversity or localization efforts.

Privacy and security

Why wrong: Privacy and security are critical for protecting user data, but they address data handling, encryption, and consent, not the diversity of users the system can serve. A chatbot might have robust privacy safeguards yet only be trained on data from a single linguistic group, leaving other populations poorly supported. Thus, privacy and security are orthogonal to inclusiveness; they do not directly influence whether the system performs well for people of different languages or cultural backgrounds.

Transparency

Why wrong: Transparency focuses on making the chatbot's decisions and behavior understandable to users and stakeholders, often through explainability tools and documentation. While transparency helps users trust the system and allows developers to identify biases, it does not actively prevent exclusion. A chatbot could be highly transparent—showing exactly why it recommends a product—yet still not understand cultural nuances in certain regions, so transparency is a supporting principle rather than the primary one for equitable multilingual service.

Fairness

Why wrong: Fairness is not the core issue here because fairness is concerned with systematic and disparate impacts across protected groups—such as race, gender, or age—measured through metrics like demographic parity and equalized odds. A deep neural network may be entirely unfair if trained on biased historical data, but the question's scenario describes an inability to explain decisions, not evidence of unequal outcomes. While the two principles interact (you often need transparency to audit fairness), a model can be opaque yet still produce statistically balanced results, making explainability the primary deficiency.

Reliability & Safety

Why wrong: Reliability & Safety is a different pillar that focuses on whether an AI system operates consistently and robustly under normal and adverse conditions, including handling data drift, edge cases, adversarial inputs, and system failures without causing harm. The described gap is not that the bank's model fails unpredictably or produces unsafe outputs, but that the DNN's internal reasoning is not transparent. A model can be technically reliable—highly accurate and stable—yet still unable to provide a human-understandable explanation for any single prediction, so this option does not address the explainability problem.

Privacy & Security

Why wrong: Privacy & Security is unrelated to the explainability failure described, because this principle governs how personal and financial data is collected, stored, processed, and protected from unauthorized access, leakage, or misuse. A deep neural network could be fully transparent—with every decision explained—while still violating privacy through insecure data handling, or, conversely, it could keep data completely secure and remain a black box. The issue in the scenario is that the bank cannot explain the model's decisions to users, which is a transparency obligation, not a data-protection or access-control matter.

Fairness

Why wrong: Fairness is incorrect because it focuses on preventing algorithmic bias that leads to discriminatory outcomes for different demographic groups or individuals based on protected attributes. A warehouse robot's lighting-induced malfunction is an environmental and physical robustness problem that affects operations uniformly, irrespective of a person's race, gender, age, or any protected characteristic. There is no disparate impact on a particular group, so fairness is not the operative responsible-AI principle here.

Privacy and Security

Why wrong: Privacy and Security are incorrect because they pertain to protecting data confidentiality and maintaining system integrity against cyber threats, unauthorized access, or data breaches. A robot malfunctioning due to lighting conditions is not a result of external hacking, malicious input, or improper handling of personal data—it is a failure of physical and operational correctness. While a warehouse robot may capture images or sensor data, the described issue has no connection to data misuse or security vulnerabilities, so this principle is irrelevant to the problem.

Transparency

Why wrong: Transparency is incorrect because it concerns making AI decisions interpretable and understandable to humans, such as documenting model behavior or providing explanations for a classification outcome. A robot that fails to navigate due to poor lighting is not suffering from a lack of explainability; even if it could clearly state why it failed, the malfunction still occurs. The issue is physical/operational robustness, not the absence of interpretable rationale, so transparency does not address this failure mode.

Fairness

Why wrong: Fairness is concerned with actively mitigating bias and ensuring equitable outcomes across demographic groups, often through algorithmic techniques like reweighting or adversarial debiasing, but it does not by itself mandate telling users about existing disparities. A system could be fair in its design yet still lack transparency if it hides its limitations. The requirement to disclose performance differences is an act of openness that falls under transparency, with fairness instead focusing on eliminating the disparities themselves.

Accountability

Why wrong: Accountability is not the primary principle here because it focuses on assigning responsibility for AI outcomes through governance structures, audit trails, and human oversight, rather than mandating public disclosure of performance limitations. While an accountable organization may eventually reveal limitations, the act of openly publishing them is specifically a transparency requirement, not an accountability one. Accountability answers the question of who is responsible, whereas transparency answers the question of what is disclosed.

Privacy and Security

Why wrong: Privacy and Security are distinct principles that concentrate on protecting sensitive data and system integrity through mechanisms like encryption, access controls, and differential privacy, not on disclosing model limitations. Publishing a known performance gap does not inherently compromise data confidentiality or system security, nor is it a security control. Thus, while privacy and security are critical for ethical AI, they do not directly address the need to inform users about a model's weak spots.

Privacy and security

Why wrong: Privacy and security concern how data is collected, stored, and protected from unauthorized access, including encryption, access controls, and compliance with data protection laws. Although issuing parking tickets involves license plate data, nothing in the scenario suggests a data breach, unauthorized disclosure, or insecure processing. The core issue is the system's unexplained decision, which has nothing to do with safeguarding personal data.

Inclusiveness

Why wrong: Inclusiveness addresses whether the AI system fairly serves people of all backgrounds, abilities, and circumstances, such as ensuring accessibility and avoiding disparate treatment. The scenario contains no information about bias against any group or about certain drivers being disproportionately affected. The problem is the absence of an explanation for a ticketing decision, which is not a matter of inclusive design.

Reliability and safety

Why wrong: Reliability and safety mean the AI should perform its function accurately and without harmful errors, such as falsely ticketing a vehicle that should not be ticketed, and should remain safe under real-world conditions. While an inability to explain a decision could coexist with a reliable system, the described problem is specifically that the decision is opaque or unjustifiable. Transparency is the principle that directly addresses this lack of meaningful human understanding of the AI's reasoning.

Reliability and Safety

Why wrong: Reliability and Safety concern whether the AI system performs its intended function consistently and without causing physical, financial, or operational harm due to errors, crashes, or unexpected behavior. While a biased lending model could be seen as unreliable, the core issue here is not a malfunction or safety failure but rather the learned discriminatory patterns from training data. The system may be highly reliable in terms of uptime and accuracy, yet still produce unfair results, which is a distinct problem.

Privacy and Security

Why wrong: Privacy and Security involve protecting personal and financial data from unauthorized access, leakage, or misuse through encryption, access controls, and compliance with data protection regulations. The loan recommendation bias described is not a data breach or confidentiality failure; the system may fully protect user data yet still render unfair decisions based on biased historical labels. Therefore, while privacy and security are critical, they address a different risk than the fairness violation at issue here.

Inclusiveness

Why wrong: Inclusiveness focuses on designing AI systems that are accessible and usable by people of all abilities, backgrounds, and languages, such as supporting screen readers, providing multilingual interfaces, or accommodating diverse user needs. The bias in loan recommendations does not stem from a lack of accessibility or exclusion from use, but rather from disparate treatment of protected classes in the decision-making logic. Thus, improving inclusiveness would not directly remedy the discriminatory outcomes described.