AIF-C01 Guidelines for Responsible AI — 25 Questions

Question 1

Which TWO actions are most aligned with responsible AI practices when deploying a model that makes decisions affecting individuals? (Choose 2)

Accepted Answer

Continuously monitor the model for fairness metrics. Options B and E are correct. Continuous monitoring for fairness metrics helps detect and mitigate bias over time. Providing meaningful explanations for model decisions supports transparency and accountability. Option A is wrong because collecting data without quality checks can introduce bias and reduce data quality, undermining fairness. Option C is wrong because a homogeneous development team can lead to blind spots and lack of diverse perspectives, which is contrary to responsible AI practices. Option D is wrong because using the most complex model does not necessarily ensure fairness and may harm explainability and transparency.

Answer

Collect as much data as possible without quality checks

Answer

Ensure the development team is homogeneous to avoid conflicts

Answer

Use the most complex model available for maximum accuracy

Question 2

A healthcare company must train a model on sensitive patient data while complying with privacy regulations. They want to add noise to the training process to prevent re-identification. Which technique should they implement?

Accepted Answer

Differential privacy. Differential privacy is the correct technique because it adds calibrated noise to the training process (e.g., via gradient clipping and noise injection in stochastic gradient descent) to ensure that the model's outputs do not reveal whether any individual's data was included in the training set. This provides a formal mathematical guarantee (ε-differential privacy) that limits the risk of re-identification, which is essential for complying with privacy regulations like HIPAA or GDPR when training on sensitive patient data.

Answer

k-anonymity

Answer

Federated learning

Answer

Homomorphic encryption

Question 3

Which of the following is a key principle of responsible AI according to AWS?

Accepted Answer

Transparency. Transparency is a core principle of responsible AI according to AWS, meaning that customers should be able to understand how an AI system makes decisions, including its inputs, outputs, and limitations. AWS emphasizes this through features like model cards in Amazon SageMaker, which document model performance, intended uses, and biases. This principle ensures that AI systems are explainable and auditable, building trust with users.

Answer

Complexity

Answer

Speed

Answer

Profitability

Question 4

A company wants to ensure accountability for its machine learning models by tracking all changes to the model and its training data. Which AWS feature should they use?

Accepted Answer

AWS CloudTrail. AWS CloudTrail is the correct choice because it records API activity across AWS services, enabling governance, compliance, and operational auditing. For machine learning models, CloudTrail logs all changes to model artifacts, training data, and SageMaker resources by capturing calls to the SageMaker API, such as CreateTrainingJob, UpdateEndpoint, or PutModelPackageGroupPolicy. This provides a complete audit trail for accountability, as required by responsible AI guidelines.

Answer

Amazon CloudWatch

Answer

Amazon SageMaker Experiments

Answer

AWS CodeCommit

Question 5

Which TWO actions should a data scientist take to evaluate fairness of a binary classification model using Amazon SageMaker Clarify? (Choose two.)

Accepted Answer

Use post-training bias metrics like Difference in Positive Proportions. Amazon SageMaker Clarify provides post-training bias metrics such as Difference in Positive Proportions (DPPL) to measure disparities in model predictions across sensitive groups. These metrics quantify whether the model's predictions are fair after training, which is a key step in evaluating fairness for a binary classification model.

Answer

Ensure the training dataset is balanced by resampling

Answer

Generate SHAP values for feature importance

Answer

Run a data quality monitoring job on unlabeled data

Question 6

Which TWO techniques provide interpretability for machine learning models at a local (per-prediction) level? (Choose two.)

Accepted Answer

SHAP values. SHAP (SHapley Additive exPlanations) values are a game-theoretic approach that explains individual predictions by computing the contribution of each feature to the prediction. For a given instance, SHAP assigns an importance value to each feature, showing how much it pushed the prediction away from the average. This makes it a local interpretability method because the explanation is specific to that single prediction, not the model overall.

Answer

Partial dependence plots

Answer

Confusion matrix

Answer

Permutation feature importance

Question 7

Refer to the exhibit. A developer is reviewing CloudWatch Logs for a deployed model and notices the same input appears multiple times with slightly different probabilities. What responsible AI concern does this pattern suggest?

Accepted Answer

The model is not robust; it produces inconsistent predictions for the same input.. The pattern of the same input producing slightly different probabilities indicates that the model's predictions are not deterministic for identical inputs. This violates the principle of robustness in responsible AI, which requires that a model should produce consistent outputs for the same input under the same conditions. Inconsistent predictions for identical inputs undermine trust and reliability, making the model non-robust.

Answer

The model is overfitting to the training data.

Answer

The model is exhibiting bias against a demographic group.

Answer

The input data is drifting from the training distribution.

Question 8

A government agency uses Amazon Rekognition for identity verification. They want to ensure the model is robust against adversarial attacks. What should they do?

Accepted Answer

Train the model with adversarial examples. Training the model with adversarial examples (a technique known as adversarial training) directly improves the model's robustness against adversarial attacks by exposing it to perturbed inputs during training, forcing the model to learn decision boundaries that are less sensitive to small, malicious perturbations. This is a fundamental approach in building resilient AI systems, especially for sensitive tasks like identity verification.

Answer

Use SageMaker Model Monitor to detect adversarial inputs

Answer

Implement input validation and use Amazon SageMaker Shadow Testing

Answer

Apply AWS Shield Advanced

Question 9

A social media company uses Amazon Comprehend to moderate user comments. They want to avoid censoring legitimate speech while catching hate speech. Which approach aligns with responsible AI governance?

Accepted Answer

Implement a human-in-the-loop review for borderline cases. A human-in-the-loop (HITL) review for borderline cases aligns with responsible AI governance by balancing automated detection with human judgment. Amazon Comprehend can flag comments with moderate confidence scores (e.g., 0.5–0.9) for manual review, ensuring that ambiguous or context-dependent hate speech is not censored while still catching clear violations. This approach mitigates false positives and respects free expression, which is a core tenet of responsible AI.

Answer

Use multiple models and average their scores

Answer

Use a single model with high confidence threshold

Answer

Rely solely on automated filtering

Question 10

Which TWO actions help ensure fairness in an AI system deployed on AWS? (Select two.)

Accepted Answer

Train the model on a representative dataset. Training an AI model on a representative dataset ensures that the data reflects the diversity of the real-world population the system will serve. A non-representative dataset can lead to biased predictions, as the model may underperform or discriminate against underrepresented groups. This is a foundational step in responsible AI to mitigate fairness issues before model deployment.

Answer

Enable AWS CloudTrail for audit

Answer

Use a single validation set

Answer

Encrypt data at rest using AWS KMS

Question 11

Which THREE practices support transparency in AI systems? (Select three.)

Accepted Answer

Provide documentation of model limitations. Documenting model limitations is a core transparency practice under responsible AI frameworks. It ensures users and stakeholders understand the boundaries of model performance, such as known biases, edge cases, or data distribution shifts, which is essential for informed decision-making and trust.

Answer

Implement explainability methods like SHAP

Answer

Use open-source models exclusively

Question 12

Which TWO actions can help mitigate bias in a face recognition model trained on AWS? (Select two.)

Accepted Answer

Ensure the training dataset is balanced across demographics. Options A and B are correct. Mitigating bias in a face recognition model requires a balanced training dataset (A) and regular evaluation of model performance across demographic subgroups (B). Option C (deploying in multiple regions) affects latency or availability, not bias. Option D (larger neural network) does not address data imbalance. Option E (Amazon Rekognition's content moderation) is for detecting inappropriate content, not for bias mitigation.

Answer

Deploy the model in multiple regions

Answer

Use a larger neural network

Answer

Use Amazon Rekognition's content moderation

Question 13

Refer to the exhibit. An AWS customer runs SageMaker Clarify to evaluate bias in their training data. The report shows multiple metrics with status 'violated'. What should the customer do next?

Accepted Answer

Use data augmentation to balance the dataset. Data augmentation directly addresses the root cause of bias violations in training data: imbalanced representation of sensitive groups. SageMaker Clarify metrics like Class Imbalance (CI) or Difference in Positive Proportions in Labels (DPPL) flag when certain demographic groups are underrepresented. Augmenting the dataset with synthetic samples or additional real data for the underrepresented group reduces these imbalances, which is the recommended first step before retraining the model.

Answer

Reduce the number of features

Answer

Retrain the model with more data

Answer

Ignore the metrics because thresholds are too strict

Question 14

Refer to the exhibit. An ML team finds that their training data is stored in two subfolders under s3://my-bucket/train/. They need to ensure that the dataset is balanced for training a classification model. What should they do?

Accepted Answer

Count the number of files in each subfolder and resample. The core requirement is to balance the dataset by ensuring an equal number of samples from each subfolder (class). Counting the files in each subfolder and then resampling (e.g., undersampling the majority class or oversampling the minority class) directly addresses class imbalance. This is a standard data preprocessing step before training a classification model, and it does not require any additional AWS services beyond the existing S3 storage.

Answer

Use AWS Glue to create a balanced dataset

Answer

Use Amazon Rekognition custom labels

Answer

Enable versioning on the bucket

Question 15

A company uses Amazon SageMaker to build a binary classification model for loan approvals. After training, the data science team wants to evaluate the model for potential bias against a protected group. Which AWS service should they use to compute bias metrics?

Accepted Answer

Amazon SageMaker Clarify. Amazon SageMaker Clarify is the correct service because it is specifically designed to detect bias in machine learning models and datasets. It provides built-in bias metrics (e.g., difference in positive proportion, disparate impact) for both pre-training and post-training evaluation, making it the appropriate tool for assessing potential bias against a protected group in a binary classification model.

Answer

Amazon SageMaker Model Monitor

Answer

Amazon SageMaker Debugger

Answer

Amazon SageMaker Experiments

Question 16

A data scientist is using Amazon SageMaker to train a model and wants to understand the contribution of each feature to individual predictions. Which technique should they use to generate local explanations?

Accepted Answer

SHAP values. SHAP (SHapley Additive exPlanations) values are the correct choice because they provide local explanations by decomposing a prediction into the additive contribution of each feature, based on cooperative game theory. This allows the data scientist to understand exactly how each feature influenced a specific individual prediction, unlike global methods that summarize behavior across the entire dataset.

Answer

Permutation feature importance

Answer

Global feature importance

Answer

Partial dependence plots

Question 17

A company deploys a deep learning model for image classification using Amazon SageMaker. They are concerned about adversarial attacks that could misclassify images with small perturbations. Which of the following is the most effective approach to improve model robustness?

Accepted Answer

Apply adversarial training. Adversarial training is the most effective approach because it explicitly augments the training dataset with adversarial examples—inputs crafted with small, intentional perturbations designed to fool the model. By training on these perturbed samples, the model learns to recognize and resist such attacks, directly improving its robustness against adversarial perturbations in image classification tasks on SageMaker.

Answer

Reduce training data size

Answer

Use early stopping during training

Answer

Increase model complexity

Question 18

A large enterprise has multiple teams deploying ML models on AWS. To ensure governance and accountability, they need to enforce that all models pass a fairness review before production deployment. Which SageMaker feature should they use to implement this approval workflow?

Accepted Answer

SageMaker Model Registry. SageMaker Model Registry is the correct choice because it provides a centralized catalog for managing ML models, including versioning, approval status, and metadata. It supports approval workflows by allowing you to define model groups, set approval statuses (e.g., PendingApproval, Approved, Rejected), and integrate with CI/CD pipelines to enforce that only approved models are deployed to production.

Answer

SageMaker Studio

Answer

SageMaker Experiments

Answer

SageMaker Model Monitor

Question 19

An e-commerce company uses an Amazon Lex chatbot to handle customer inquiries. They want to implement human oversight for sensitive interactions, such as when the chatbot cannot provide a confident response. Which AWS service should they integrate?

Accepted Answer

Amazon Augmented AI (A2I). Amazon Augmented AI (A2I) is the correct service because it provides built-in human review workflows for ML predictions, allowing you to route low-confidence responses from Amazon Lex to human reviewers for oversight. This directly addresses the requirement for human oversight on sensitive interactions where the chatbot cannot provide a confident response.

Answer

Amazon Rekognition

Answer

Amazon Comprehend

Answer

Amazon SageMaker Ground Truth

Question 20

Which of the following is NOT one of the core principles of responsible AI as defined by AWS?

Accepted Answer

Profitability. Profitability is not one of the core principles of responsible AI as defined by AWS. AWS defines six core principles: Fairness, Transparency, Robustness, Privacy, Security, and Explainability. Profitability is a business objective, not a principle for ensuring ethical and trustworthy AI systems.

Answer

Transparency

Answer

Fairness

Answer

Robustness