AI Testing Free Sample Questions

Methods and Techniques for the Testing of AI-Based Systems · Metamorphic Testing Application

1. A

   If a source test case with an applicant's income of $50,000 results in a specific risk score, then a follow-up test case with the same applicant but an income of $50,001 should result in exactly the same risk score.

2. B

   If a source test case passes, duplicate the applicant profile exactly for a follow-up test case; the output risk score must be identical.

3. C

   If a source test case with an applicant's income of $60,000 and 2 credit cards results in a risk score 'X', a follow-up test case with the same applicant but an income of $90,000 and 2 credit cards should result in a risk score less than or equal to 'X'.

4. D

   Submit a source test case and a follow-up test case with randomized input values; the difference in the output risk scores should be proportional to the difference in the randomized inputs.

Testing AI-Specific Quality Characteristics · Testing for Inappropriate Bias

1. A

   Perform white-box testing on the model's algorithm to ensure it does not contain any logic referencing demographic data.

2. B

   Analyze the model's predictions across different demographic subgroups (e.g., gender, ethnicity) and check for statistically significant differences in selection rates.

3. C

   Review the training data for proxy variables, such as zip codes or attendance at specific universities, that strongly correlate with demographic groups.

4. D

   Conduct A/B testing by deploying two different versions of the model to the hiring team and comparing their feedback.

5. E

   Verify that the model's predictions are deterministic by submitting the same resume multiple times and ensuring the output is identical.

Using AI for Testing · AI for Test Case Generation

1. A

   The AI tool will generate too many redundant test cases, overloading the pipeline.

2. B

   The AI tool will struggle to achieve high code coverage compared to manually written tests.

3. C

   The AI tool will not know the expected outcome for the generated tests, creating a significant test oracle problem.

4. D

   The AI tool will require a GPU, which is not available in their standard CI/CD runners.

Quality Characteristics for AI-Based Systems · Explainability (XAI)

1. A

   Reviewing the model's source code and the weights of the neural network to understand its internal logic.

2. B

   Using a model-agnostic technique like LIME (Local Interpretable Model-Agnostic Explanations) to generate explanations for individual predictions by analyzing how small input perturbations affect the outcome.

3. C

   Creating a simpler, transparent proxy model (like a decision tree) that approximates the DNN's behavior and using its rules as the explanation.

4. D

   Documenting the entire training dataset and the data pipeline, as this provides transparency into how the model was created.

ML - Neural Networks and Testing · Coverage Measures for Neural Networks

1. A

   True

2. B

   False

ML - Data · Data Labelling

1. A

   Assigning the entire dataset to a single, highly-trained internal expert to ensure maximum consistency and accuracy, despite the time it will take.

2. B

   Using an AI-assisted labeling tool to perform a first pass, and then having the internal team review and correct the AI's suggestions.

3. C

   Outsourcing the labeling to a crowdsourcing platform where each image is labeled by multiple (e.g., three) different workers, and using the majority vote as the final label.

4. D

   Using an unsupervised clustering algorithm to group similar images, and then manually labeling just a few images from each cluster to represent the whole cluster.

Methods and Techniques for the Testing of AI-Based Systems · Adversarial Attacks and Data Poisoning

1. A

   Introduce a small number of mislabeled images into the training dataset (e.g., images of 'Stop' signs labeled as 'Speed Limit 60') and observe if the trained model's accuracy degrades significantly or learns a backdoor.

2. B

   Use pairwise testing to combine different environmental conditions like rain, fog, and nighttime to ensure the model performs well in all scenarios.

3. C

   Apply metamorphic testing by slightly rotating or brightening an image of a stop sign and verifying the classification remains 'Stop Sign'.

4. D

   Generate adversarial examples by applying subtle, almost imperceptible pixel-level noise to a valid image of a stop sign and checking if it causes a misclassification (e.g., to 'Green Light').

5. E

   Perform A/B testing by comparing the current model's performance against a previous version on a standard benchmark dataset.

Machine Learning (ML) – Overview · Forms of ML

1. A

   Supervised Learning (Classification)

2. B

   Supervised Learning (Regression)

3. C

   Unsupervised Learning (Association)

4. D

   Reinforcement Learning

Test Environments for AI-Based Systems · Virtual Test Environments

1. A

   It allows the use of AI-specific processors like TPUs, which are not available in physical drones.

2. B

   It eliminates the need for any physical hardware testing before deployment, reducing costs by 100%.

3. C

   It enables the safe, repeatable, and scalable testing of rare and dangerous edge-case scenarios (e.g., multiple drones converging, sudden obstacles).

4. D

   It provides better explainability for the AI model's decisions than testing in the real world.

ML Functional Performance Metrics · Selecting ML Functional Performance Metrics

1. A

   Accuracy

2. B

   Precision

3. C

   Recall (Sensitivity)

4. D

   Specificity (True Negative Rate)