AI0-001 AI Governance and Ethics Practice Test 4 — 15 Questions

Question 1

A financial institution wants to use AI for loan approvals and must comply with fair lending laws. Which TWO practices should the institution adopt to mitigate bias and ensure compliance?

Accepted Answer

Apply fairness-aware machine learning techniques during model training. To mitigate bias, using fairness-aware algorithms and conducting disparate impact analysis are direct steps. Using only demographic data is illegal (redlining). Removing all features reduces model utility. A black-box model without explanation would hinder compliance.

Answer

Remove all features except credit score to avoid bias

Answer

Use a black-box model without explainability to protect intellectual property

Answer

Use only demographic features to ensure equal treatment

Question 2

An organisation is developing an AI system that will be used to evaluate teacher performance in public schools. To ensure responsible use, which TWO governance elements should be in place?

Accepted Answer

Develop a responsible use policy for the AI system. Responsible use policies and vendor AI assessment are key governance elements. Using only open-source models is not a governance requirement. Human oversight should be human-in-the-loop, not just monitoring. Publishing all training data may breach privacy.

Answer

Implement human-on-the-loop oversight where humans only monitor outputs

Answer

Publish the entire training dataset publicly for transparency

Answer

Require the system to be built only with open-source models

Question 3

A data scientist is training a resume screening model to rank job applicants. The training data includes historical hiring decisions from the past 10 years. The company wants to avoid unfair bias against underrepresented groups. Which type of bias is most likely present in the training data?

Accepted Answer

Historical bias. Historical bias occurs when the training data reflects past societal biases, such as underrepresentation of certain groups in hiring. The model learns these patterns, perpetuating unfairness. Selection bias arises from non-random sampling, confirmation bias from favoring information that confirms preexisting beliefs, and algorithmic bias from model design choices.

Answer

Algorithmic bias

Answer

Selection bias

Answer

Confirmation bias

Question 4

A financial institution is deploying an AI system to approve personal loans. To comply with the EU AI Act's high-risk AI requirements, the bank must ensure meaningful human oversight. Which implementation BEST satisfies this requirement?

Accepted Answer

Require a human to review and approve every loan decision before it becomes final. The EU AI Act requires that high-risk AI systems allow for human oversight, including the ability to override or reverse the system's decisions. A mandatory human review before final approval ensures that the human can intervene. The other options either do not provide effective oversight or allow for rubber-stamping.

Answer

Use a separate AI model to audit the primary AI's decisions weekly

Answer

Allow applicants to appeal AI decisions through a customer service process

Answer

Provide a dashboard showing the AI's confidence score for each application

Question 5

A data scientist needs to explain why a specific loan application was rejected by a tree-based model. The model is complex and not inherently interpretable. Which method should the data scientist use to provide a local explanation for this single prediction?

Accepted Answer

LIME. LIME (Local Interpretable Model-agnostic Explanations) is the correct choice because it is specifically designed to provide local explanations for individual predictions by approximating the complex model with a simpler, interpretable surrogate model around that specific instance. For a tree-based model that is not inherently interpretable, LIME can explain why a single loan application was rejected by perturbing the input and observing the changes in predictions, making it ideal for this use case.

Answer

SHAP values

Answer

Model cards

Answer

Attention visualization

Question 6

A company wants to train a language model on sensitive customer data without transferring the raw data to a central server. Which privacy-preserving technique should they use?

Accepted Answer

Federated learning. Federated learning is the correct technique because it trains a shared model across decentralized edge devices holding local data, without transferring raw customer data to a central server. Only model updates (gradients) are sent to the aggregation server, preserving data locality and reducing exposure. This directly addresses the requirement of avoiding raw data transfer while still enabling collaborative model training.

Answer

Differential privacy

Answer

Data minimisation

Answer

Anonymisation

Question 7

An AI governance team is implementing the NIST AI Risk Management Framework. They have identified a high-risk AI system and are in the 'Measure' function. Which activity is most appropriate for this function?

Accepted Answer

Conduct bias and fairness impact assessments on the model. In the NIST AI Risk Management Framework (AI RMF), the 'Measure' function focuses on assessing and analyzing risks associated with AI systems. For a high-risk AI system, conducting bias and fairness impact assessments is a core activity within this function, as it quantifies and evaluates potential harms related to fairness, accuracy, and transparency. This aligns with the framework's emphasis on quantitative and qualitative risk measurement before moving to risk treatment in the 'Manage' function.

Answer

Implement technical controls to mitigate identified risks

Answer

Document the system's intended purpose and data sources

Answer

Establish an AI ethics board to oversee risk decisions

Question 8

A company uses an AI system to generate marketing images. They are concerned about copyright ownership of the generated content. According to current US copyright law, who typically owns the copyright for AI-generated work?

Accepted Answer

No one; the work may be in the public domain. Under US copyright law, copyright protection requires human authorship. Works created entirely by AI without human creative input are generally not copyrightable. The user who provides prompts may not own the copyright if the work lacks human authorship. The AI developer or AI system itself cannot hold copyright.

Answer

The user who provided the input prompts

Answer

The AI system itself, as the creator

Answer

The company that owns the AI model

Question 9

A company is evaluating fairness metrics for a hiring model. They want to ensure that the model has similar true positive rates (TPR) across demographic groups. Which fairness metric should they use?

Accepted Answer

Equalized odds. Equalized odds requires that the true positive rate and false positive rate are equal across groups. Demographic parity requires equal selection rates. Individual fairness requires similar individuals to be treated similarly. Calibration ensures predicted probabilities match actual outcomes for each group. The scenario specifies TPR, which is part of equalized odds.

Answer

Calibration

Answer

Individual fairness

Answer

Demographic parity

Question 10

A company wants to reduce the carbon footprint of training large AI models. Which practice is MOST effective for achieving 'Green AI'?

Accepted Answer

Prune the model to reduce its size before training. Green AI practices focus on reducing computational and environmental costs. Using model pruning reduces model size and computational requirements. Using more GPUs increases energy consumption. Training with larger datasets increases compute. Using older hardware is often less energy-efficient.

Answer

Train on larger datasets to improve accuracy

Answer

Use more powerful GPUs to speed up training

Answer

Use older, less efficient hardware to save on manufacturing emissions

Question 11

A healthcare AI model is subject to GDPR because it processes data of EU patients. The model makes automated decisions about treatment plans. Under GDPR, patients have the right to obtain an explanation of the decision. Which approach BEST satisfies this 'right to explanation'?

Accepted Answer

Generate a SHAP-based explanation showing the contribution of each input feature to the decision. GDPR's right to explanation requires meaningful information about the logic involved in automated decisions. Providing a clear, understandable explanation using techniques like SHAP fulfills this. A simplified summary may be needed, but the explanation must be substantive. Pointing to model cards or a confidence score alone is insufficient.

Answer

Offer the patient the option to have the decision reviewed by a human

Answer

State that the decision was made by an AI system and cannot be explained

Answer

Provide a copy of the model's source code to the patient

Question 12

A data scientist is using differential privacy to protect individual privacy in a training dataset. Which TWO actions are correct implementations of differential privacy?

Accepted Answer

Set a privacy budget (epsilon) to limit information leakage. Setting a privacy budget (epsilon) is a core mechanism in differential privacy that quantifies and limits the amount of information leaked about any individual in the dataset. By controlling epsilon, the data scientist can formally bound the privacy loss, ensuring that the model's outputs do not reveal whether any specific individual's data was included in training.

Answer

Train the model on a small subset of data to reduce exposure

Answer

Remove all personally identifiable information (PII) from the dataset

Answer

Aggregate data into groups before training

Question 13

A data scientist discovers that a model trained to predict loan defaults is denying loans at a higher rate for a particular demographic group. Which type of bias is MOST likely present?

Accepted Answer

Historical bias. Historical bias occurs when the training data reflects past societal inequalities, leading the model to learn and perpetuate those patterns. In this case, if historical loan data shows higher denial rates for a demographic group due to past discriminatory practices, the model will replicate that bias in its predictions. This is the most likely cause because the model is not inherently biased but inherits bias from the data it was trained on.

Answer

Confirmation bias

Answer

Selection bias

Answer

Algorithmic bias

Question 14

A financial institution wants to deploy an AI system that automates loan approvals. Under the EU AI Act, this system would be classified as high-risk. Which of the following is a MANDATORY requirement for high-risk AI systems before market placement?

Accepted Answer

Establishing a risk management system throughout the system lifecycle. The EU AI Act requires a conformity assessment (often including third-party involvement) for high-risk AI systems to ensure compliance with requirements such as data governance, transparency, and human oversight. The other options are either voluntary, not required, or apply to lower-risk tiers.

Answer

Obtaining a CE marking of conformity through self-assessment only

Answer

Acquiring liability insurance for AI-related harms

Answer

Publishing the training dataset in an open repository

Question 15

A company is training a large language model and wants to reduce its carbon footprint. Which practice is MOST effective for reducing training energy consumption while maintaining model quality?

Accepted Answer

Use mixed-precision training and prune unnecessary parameters. Green AI practices include using more efficient hardware (like GPUs with lower power draw), model pruning, and early stopping. Using CPUs is slower and less efficient. Increasing batch size without tuning can hurt convergence. Using a larger model increases energy. Training on renewable energy reduces the carbon impact but does not reduce energy consumption itself.

Answer

Increase the batch size to the maximum the GPU memory allows

Answer

Use a larger model architecture to achieve higher accuracy faster

Answer

Train the model on CPUs instead of GPUs