ISQI ISQI-CT-AI Online Practice Questions and Exam Preparation

ISQI ISQI-CT-AI Online Questions & Answers

• Question 121:

  The stakeholders of a machine learning model have confirmed that they understand the objective and purpose of the model and have ensured that the proposed model aligns with their business priorities. They have also selected a framework and a machine learning model that they will be using. What should be the next step to progress along the machine learning workflow?

  A. Prepare and pre-process the data that will be used to train and test the model
  B. Tune the machine learning algorithm based on objectives and business priorities
  C. Agree on defined acceptance criteria for the machine learning model
  D. Evaluate the selection of the framework and the model
  A. Prepare and pre-process the data that will be used to train and test the model

  ---

  The machine learning workflow typically proceeds iteratively, but once the objectives, business priorities, framework, and model have been agreed upon, the next step is data preparation. The ISTQB CT-AI syllabus explains that data preparation is required to provide suitable input data for training an ML algorithm and for evaluating the resulting model.

  The syllabus notes that ML models cannot be trained or evaluated without properly prepared data and that data preparation is therefore a prerequisite for subsequent steps such as training, tuning, and evaluation. This includes collecting, cleaning, transforming, and splitting the data for training and testing purposes.

  This directly supports Option A as the correct next step in the machine learning workflow.

• Question 122:

  Which of the following problems would best be solved using the supervised learning category of regression?

  A. Determining the optimal age for a chicken's egg-laying production using input data of the chicken's age and average daily egg production for one million chickens
  B. Recognizing a knife in carry-on luggage at a security checkpoint using airport scanner images
  C. Determining whether an animal is a pig or a cow based on image recognition
  D. Predicting shopper purchasing behavior based on the category of shopper and the positioning of promotional displays within a store
  A. Determining the optimal age for a chicken's egg-laying production using input data of the chicken's age and average daily egg production for one million chickens

  ---

  The ISTQB CT-AI syllabus defines supervised learning as being divided into classification and

  regression problems.

  Regression is used when the model must predict a continuous numeric value, rather than a discrete category.

  In this scenario:

  Option A involves predicting an optimal numeric value (age) based on historical numerical data (age and egg production), which is a classic regression problem.

  The remaining options represent classification tasks:

  Recognizing a knife in luggage

  Distinguishing between animal types

  Predicting purchasing behavior categories

  These involve assigning inputs to discrete classes rather than predicting continuous values.

  This aligns with the syllabus definition:

  "Regression is used when the problem requires the ML model to predict a numeric output, for example predicting the age of a person based on their habits."

  (ISTQB CT-AI Syllabus v1.0, Section 3.1.1, Page 26 of 99)

• Question 123:

  A neural network has been designed and created to assist day traders in improving efficiency when buying and selling commodities in a rapidly changing market. Suppose the test team executes a test on the neural network in which each neuron is examined. For this network, the shortest path indicates a "buy" decision, and this will only occur when the one-day predicted value of the commodity is greater than the spot price by 0.75%. The neurons are stimulated by entering commodity prices, and testers verify that they activate only when the future value exceeds the spot price by at least 0.75%.

  Which of the following statements BEST explains the type of coverage being tested on the neural network?

  A. Threshold coverage
  B. Neuron coverage
  C. Sign-change coverage
  D. Value-change coverage
  A. Threshold coverage

  ---

  The ISTQB CT-AI syllabus defines threshold coverage as a structural coverage criterion for neural networks that requires each neuron to reach an activation value greater than a defined threshold during testing. The purpose is to ensure that neurons are sufficiently activated and that threshold-based decision logic is exercised.

  In this scenario, the testers explicitly verify that neurons activate only when the predicted future value exceeds the spot price by a specific threshold (0.75%). This directly matches the definition of threshold coverage, as the test checks neuron activation against a predefined activation threshold.

  The other options do not apply. Neuron coverage focuses on whether neurons are activated at all, not whether they exceed a specific threshold. Sign-change coverage and value-change coverage relate to changes in neuron activation signs or values, not to activation relative to a fixed threshold.

  Therefore, Option A is the correct answer.

• Question 124:

  Which two test procedures are BEST suited for CleverPropose system testing? (Choose Two)

  A. Back-to-back testing
  B. Adversarial testing
  C. Metamorphic testing
  D. Exploratory data analysis
  E. Pairwise testing
  A. Back-to-back testing
  C. Metamorphic testing

  ---

  The ISTQB CT-AI syllabus explains that AI-based decision-support systems benefit strongly from back-to-back testing and metamorphic testing when test oracle problems exist or when limited regression tests are available. In this scenario, CleverPropose replaces an older advisory system. Back-to-back testing (Option A) is appropriate because the outputs of the existing conventional system can serve as a reference, enabling comparison with the new AI-based system. This aligns with the syllabus guidance

  for situations where an AI system replaces a deterministic legacy system.

  Metamorphic testing (Option C) is also suitable, as described in Section 4.6 - Metamorphic Relations. When regression tests are limited and decision logic is complex, testers can define metamorphic relations such as "if customer income increases, the risk rating should not worsen." These relations allow meaningful validation even when exact expected outputs are unavailable.

  Exploratory data analysis (Option D) is not a system testing procedure. Pairwise testing (Option E) is not well suited for complex AI-based financial advisory systems. Adversarial testing (Option B) is primarily relevant for robustness or security-focused evaluations rather than core system testing of advisory functionality.

  Thus, Options A and C are the correct and syllabus-supported choices.

• Question 125:

  Which ONE of the following options does NOT describe a challenge for acquiring test data in ML systems?

  A. Compliance needs require proper care to be taken of input personal data
  B. Nature of data constantly changes with time.
  C. Data for the use case is being generated at a fast pace.
  D. Test data being sourced from public sources.
  C. Data for the use case is being generated at a fast pace.

  ---

  The ISTQB CT-AI syllabus describes several challenges related to acquiring test data for ML systems, including:

  Compliance and privacy requirements, which require careful handling of personal or sensitive data.

  Changing data over time (data drift), which affects data relevance and representativeness.

  Public data sources, which may introduce issues such as unknown quality, bias, or licensing constraints.

  In contrast, data being generated at a fast pace is not inherently a challenge for data acquisition. On the contrary, it often provides more opportunities for collecting sufficient training and test data, assuming storage, processing, and governance are properly managed.

  Therefore, Option C correctly identifies the statement that does not describe a challenge for acquiring test data in ML systems.

• Question 126:

  How can a tester check the system for bias as part of a review of data sources, acquisition, and preprocessing?

  A. During the review, it can uncover algorithmic bias by analysing the procedures used to obtain the training data.
  B. During the review of the preprocessing, the auditor can uncover whether the data has been influenced in a way that could lead to sample distortions.
  C. It may use the LIME method as part of its data collection review to detect inappropriate bias.
  D. As part of the review of preprocessing, it can reveal whether the data has been influenced in a way that could lead to algorithmic bias.
  B. During the review of the preprocessing, the auditor can uncover whether the data has been influenced in a way that could lead to sample distortions.

  ---

  Bias detection at the data level is performed by reviewing data acquisition and preprocessing steps, as explained in Section 2.3 - Data Quality and Bias of the ISTQB CT-AI syllabus. Sample bias can arise when data is distorted or when preprocessing introduces unintended shifts, for example through filtering, normalization, or labeling steps that disproportionately affect certain subsets of the data. Option B correctly reflects this approach: during preprocessing review, testers can identify whether the data has

  been influenced in a way that could lead to sample distortions.

  Option A is incorrect because algorithmic bias originates from the model and its learning process, not directly from data acquisition procedures. Option C is incorrect because LIME is an explainability technique applied to trained models, not a method used during data source or preprocessing reviews. Option D is incorrect because preprocessing affects sample bias rather than algorithmic bias.

  Thus, Option B correctly matches the syllabus guidance on detecting bias during reviews of data sources, acquisition, and preprocessing.

• Question 127:

  Which statement about using AI to analyze reported defects is correct?

  A. ML models trained with critical defect tickets can identify defects that cause serious consequences.
  B. ML models can support duplicate defect identification when checking defect criticality.
  C. ML models can identify categories for a reported defect during assignment.
  D. ML models identify developers who should handle a defect based on ticket content.
  C. ML models can identify categories for a reported defect during assignment.

  ---

  The ISTQB CT-AI syllabus (Section 5.3 - AI Support for Defect Analysis) explains that AI can categorize defect reports using natural language processing or classification models. Categorization helps route defects efficiently and determine which areas of the system are affected. Thus, Option C is correct: AI can identify defect categories, supporting assignment and triage.

  Option A is incorrect because machine learning models cannot reliably infer severity or real-world consequences without explicit domain-specific risk modeling. Option B is incorrect because duplicate defect detection is a separate activity from assessing defect criticality; the syllabus clearly distinguishes duplicate identification from severity analysis. Option D is incorrect because, while AI may support defect routing based on categories, the syllabus does not state that AI identifies specific developers

  responsible for fixing defects.

  Thus, Option C is the syllabus-accurate statement.

• Question 128:

  Which performance metric is BEST suited to assess the quality of trained models detecting fraudulent credit card transactions?

  A. Sensitivity
  B. Accuracy
  C. F1 value
  D. --
  C. F1 value

  ---

  The ISTQB CT-AI syllabus explains in Section 3.2 - Functional Performance Criteria of ML Models that accuracy becomes unreliable when class imbalance exists. In fraud detection, more than 99% of transactions are non-fraudulent, meaning the data set is extremely imbalanced. Because accuracy counts all correctly classified non-fraudulent transactions, it can appear artificially high even when the fraud detection performance is poor. Therefore, accuracy is not suitable for evaluating fraud detection

  systems.

  The syllabus further explains that sensitivity (recall) measures the proportion of correctly identified fraudulent transactions. This metric is important because missed fraudulent events can lead to significant financial loss. However, the client also requires that legitimate transactions be correctly identified, meaning false positives must be minimized to maintain customer satisfaction.

  The F1 score, defined as the harmonic mean of precision and recall, balances both aspects: precision limits false alarms affecting legitimate customers, while recall ensures that fraudulent transactions are detected. Section 3.2 emphasizes that when both false positives and false negatives have serious consequences and the data is highly imbalanced, the F1 value is the most appropriate metric.

  Thus, Option C is the correct choice.

• Question 129:

  Which of the following approaches would help overcome testing challenges associated with probabilistic and non-deterministic AI-based systems?

  A. Run the test several times to ensure that the AI always returns the same correct test result
  B. Decompose the system test into multiple data ingestion tests to determine if the AI system is getting a sufficient volume of input data
  C. Decompose the system test into multiple data ingestion tests to determine if the AI system is getting precise and accurate input data
  D. Run the test several times to generate a statistically valid test result to ensure that an appropriate number of answers are accurate
  D. Run the test several times to generate a statistically valid test result to ensure that an appropriate number of answers are accurate

  ---

  The ISTQB CT-AI syllabus explains that probabilistic and non-deterministic AI-based systems may produce different outputs for the same input. As a result, traditional single-execution testing is insufficient.

  The syllabus states that tests must be executed multiple times to obtain statistically valid results, allowing testers to evaluate whether the system produces an acceptable proportion of correct outcomes rather than identical outputs every time.

  Therefore, running the test several times to generate statistically valid results is the appropriate approach for testing probabilistic and non-deterministic AI-based systems, making Option D the correct answer.

• Question 130:

  An image classification system is being trained to classify human faces. The distribution of the training data is 70% ethnicity A and 30% ethnicities B, C, and D combined. Based ONLY on the information above, which of the following options BEST describes the situation of this image classification system?

  A. This is an example of expert system bias.
  B. This is an example of sample bias.
  C. This is an example of hyperparameter bias.
  D. This is an example of algorithmic bias.
  B. This is an example of sample bias.

  ---

  The ISTQB CT-AI syllabus defines sample bias as a bias that occurs when the training data is not representative of the population the model is expected to encounter in operation.

  In this scenario, ethnicity A is significantly overrepresented in the dataset compared to ethnicities B, C, and D. As a result, the model may learn patterns that perform better for ethnicity A and worse for the underrepresented groups.

  Expert system bias relates to biases introduced by explicitly defined rules, not data distribution.

  Hyperparameter bias concerns model configuration choices, not dataset composition.

  Algorithmic bias refers to bias arising from the algorithm itself, whereas this scenario describes bias originating from the training data.

  Therefore, sample bias (Option B) is the correct answer.