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

ISQI ISQI-CT-AI Online Questions & Answers

• Question 1:

  An ML engineer is trying to determine the correctness of a new open-source implementation X of a supervised regression algorithm. R-squared (R? is one of the functional performance metrics used to assess the quality of the model. Which ONE of the following would be an APPROPRIATE strategy to achieve this goal?

  A. Add 10% of the rows randomly, create another model, and compare the R?scores of both models.
  B. Train multiple models by changing the order of input features and verify that the R?scores of these models vary significantly.
  C. Compare the R?score of the model obtained using two different implementations that use two different programming languages while applying the same algorithm and the same training and test data.
  D. Drop 10% of the rows randomly, create another model, and compare the R?scores of both models.
  C. Compare the R?score of the model obtained using two different implementations that use two different programming languages while applying the same algorithm and the same training and test data.

  ---

  To verify the correctness of an implementation, the ISTQB CT-AI syllabus recommends back-to-back comparison using the same algorithm, the same training data, and the same test data. Comparing outputs or performance metrics across independent implementations allows defects in a specific implementation to be identified.

  Option C applies exactly this principle: two independent implementations of the same supervised regression algorithm are compared under identical conditions. If the resulting R?scores are consistent (within expected numerical tolerances), this provides strong evidence that the new implementation X is correct.

  Option A and Option D change the training dataset by adding or removing data. Variations in R?caused by dataset changes do not validate implementation correctness. Option B focuses on robustness with respect to feature ordering; while feature order should not affect results in a correct implementation, this does not directly validate correctness against a trusted reference.

  Therefore, Option C is the most appropriate strategy for verifying the correctness of the new implementation.

• Question 2:

  Arihant Meditation is a start-up using AI to aid people in achieving deeper and more effective meditation based on the analysis of various factors, such as the time and duration of meditation, pulse, blood pressure, EEG patterns, and others. Their model accuracy and other functional performance parameters have not yet reached the desired level.

  Which ONE of the following factors is NOT a factor affecting the ML functional performance?

  A. The data pipeline
  B. The quality of the labeling
  C. Biased data
  D. The number of classes
  D. The number of classes

  ---

  The ISTQB CT-AI syllabus explains that several factors directly influence machine learning functional performance, including the data pipeline, the quality of data labeling, and the presence of biased data. Problems in the data pipeline can introduce noise or inconsistencies, poor labeling quality can mislead the learning process, and biased data can distort model outcomes and reduce accuracy or fairness.

  The number of classes, however, is primarily a characteristic of the problem formulation and model design. While it may influence model complexity or architecture, it is not considered a direct factor affecting functional performance metrics such as accuracy, precision, recall, or bias in the sense described by the syllabus.

  Therefore, Option D is the correct answer.

• Question 3:

  The activation value output for a neuron in a neural network is obtained by applying a computation to the neuron. Which ONE of the following options BEST describes the inputs used to compute the activation value?

  A. Individual bias at the neuron level, activation values of neurons in the previous layer, and weights assigned to the connections between the neurons.
  B. Activation values of neurons in the previous layer, and weights assigned to the connections between the neurons.
  C. Individual bias at the neuron level, and weights assigned to the connections between the neurons.
  D. Individual bias at the neuron level, and activation values of neurons in the previous layer.
  A. Individual bias at the neuron level, activation values of neurons in the previous layer, and weights assigned to the connections between the neurons.

  ---

  In a neural network, the activation value of a neuron is computed using three components:

  Activation values of neurons in the previous layer, which provide the input signals Weights assigned to the connections, which scale the input signals An individual bias at the neuron level, which shifts the summed input The computation is typically expressed as:

  

• Question 4:

  Which statement about testing to prevent data poisoning and adversarial attacks is correct?

  A. Regression testing can be used to verify data sourcing policies to ensure the source of training data.
  B. The adversarial examples identified during adversarial testing must not be added to the training data so that they do not poison the model.
  C. Adversarial testing consists of using adversarial attacks to identify vulnerabilities so that they can be eliminated.
  D. Using AIB testing to identify data poisoning can better identify outliers than exploratory data analysis.
  C. Adversarial testing consists of using adversarial attacks to identify vulnerabilities so that they can be eliminated.

  ---

  The ISTQB CT-AI syllabus explains in Section 4.5 - Testing AI-Specific Risks that adversarial testing is a structured test activity in which testers apply adversarial attacks --- crafted or perturbed inputs --- to intentionally expose weaknesses in the machine learning model. The purpose is to identify vulnerabilities that could be exploited through data poisoning, evasion attacks, or input manipulation. Option C correctly reflects this syllabus definition: adversarial testing is about using attacks to locate weaknesses so that they can be removed or mitigated.

  Option A is incorrect because regression testing does not verify data sourcing policies; it verifies unchanged functionality after modifications. Option B is incorrect because adversarial examples are often added to training data sets to improve robustness (a practice known as adversarial training), rather than being excluded. Option D is incorrect because AIB testing is not described as superior to exploratory data analysis in identifying outliers; both approaches serve different purposes, and exploratory data analysis remains essential for data quality assessment.

  Thus, Option C is consistent with the syllabus-defined purpose of adversarial testing.

• Question 5:

  AI-enabled medical devices are used nowadays to automate certain parts of medical diagnostic processes. Since these are life-critical processes, the relevant authorities are considering introducing suitable certifications for these AI-enabled medical devices. This certification may involve several facets of AI testing (I-V).

  A. Autonomy II. Maintainability III. Safety IV. Transparency
  B. Side effects Which ONE of the following options contains the three MOST required aspects to be satisfied for the certification of AI-enabled medical devices?
  C. Aspects II, III, and IV
  D. Aspects I, II, and III
  E. Aspects III, IV, and V
  F. Aspects I, IV, and V
  C. Aspects II, III, and IV

  ---

  For AI-enabled medical devices operating in life-critical contexts, the ISTQB CT-AI syllabus emphasizes that certification must focus primarily on aspects that directly impact patient safety and regulatory trust.

  Safety (Aspect III) is essential to ensure that the AI system does not cause harm to patients.

  Transparency (Aspect IV) is required so that medical professionals and regulators can understand, assess, and justify the AI's diagnostic decisions. Side effects (Aspect V) must be identified and controlled, as unintended or emergent behaviors can have serious consequences in medical applications.

  Autonomy and maintainability are relevant quality characteristics but are secondary compared to safety-critical concerns such as preventing harm, ensuring explainability, and managing unintended effects.

  Therefore, Aspects III, IV, and V are the three most required aspects for certification in this scenario.

• Question 6:

  A tourist calls an airline to book a ticket and is connected to an automated system that is able to recognize speech, understand requests related to purchasing a ticket, and provide relevant travel options. When the tourist asks about the expected weather at the destination or potential impacts on operations due to a tight labor market, the only response from the automated system is, "I don't understand your question." This AI system should be categorized as:

  A. General AI
  B. Narrow AI
  C. Super AI
  D. Conventional AI
  D. Conventional AI

  ---

  The ISTQB CT-AI syllabus defines Narrow AI (also known as weak AI) as AI systems designed to perform specific tasks within a clearly defined domain, without the ability to generalize beyond that scope.

  In this scenario, the system can:

  recognize speech,

  understand ticket-purchasing requests,

  provide travel options.

  However, it fails to handle questions outside its designed domain, such as weather forecasting or labor market impacts. This behavior is a defining characteristic of Narrow AI.

  General AI would be able to understand and reason across domains.

  Super AI would exceed human-level intelligence.

  Conventional AI refers to rule-based or deterministic systems and is not the appropriate classification here, as the system clearly uses AI techniques such as speech recognition and natural language understanding.

  Therefore, the correct classification is Narrow AI (Option B).

• Question 7:

  Which of the following are challenges specifically related to testing AI-based systems? (Choose two)

  A. Non-deterministic behavior
  B. Requirement traceability
  C. Lack of explainability
  D. Syntax errors in source code
  A. Non-deterministic behavior
  C. Lack of explainability

  ---

• Question 8:

  A car insurance company is using a new AI service to reward defensive driving behavior among its policyholders. The driving behavior is recorded in a rating number (score).

  The AI service determines this score from the following input values:

  Reference speed v_max in km/h

  Average speed v_mean in km/h

  Average acceleration a_pos in m/s?

  Average braking deceleration a_neg in m/s?

  The more defensive the driving behavior is (slow driving, low acceleration, low braking deceleration), the higher is the score.

  Three initial test cases (Test 1 to Test 3) are used for testing the AI service. In addition, new test cases A-D are proposed.

  

  Which of the new tests is NOT a follow-up test case for metamorphic testing?

  A. Test A is not a follow-up test case.
  B. Test C is not a follow-up test case.
  C. Test B is not a follow-up test case.
  D. Test D is not a follow-up test case.
  C. Test B is not a follow-up test case.

  ---

  According to the ISTQB CT-AI syllabus, metamorphic testing works by applying metamorphic relations (MRs): predictable input transformations that should lead to predictable output changes. From the initial test data, clear relations emerge for defensive driving scoring. The score increases when:

  v_mean decreases,

  a_pos decreases,

  a_neg becomes less negative,

  and decreases when the opposite occurs.

  A valid metamorphic follow-up test must modify inputs in a direction consistent with at least one metamorphic relation while keeping the expected output direction predictable.

  Test A lowers v_mean compared to Test 1, with similar acceleration values. This directly satisfies the metamorphic relation that lower speed leads to a higher score.

  Test C increases both acceleration and braking intensity compared to Test 2, making the reduced score range (30-70) consistent with more aggressive driving.

  Test D modifies acceleration and braking magnitudes in ways consistent with Test 3's defensive-driving scoring boundaries.

  Test B, however, changes multiple variables in contradictory directions:

  v_mean increases (worse),

  a_pos increases (worse),

  a_neg becomes less negative (better).

  Because these changes conflict, the expected score trend becomes unpredictable, violating the premise of a metamorphic follow-up test.

  Thus, Test B cannot be considered a metamorphic follow-up test, which makes Option C correct.

• Question 9:

  Which of the following activities belong to data preprocessing? (Choose three)

  A. Cleaning invalid data
  B. Transforming data formats
  C. Augmenting data
  D. Defining acceptance criteria
  A. Cleaning invalid data
  B. Transforming data formats
  C. Augmenting data

  ---

• Question 10:

  Which AI-specific test objective and acceptance criterion should be selected MOST LIKELY for testing GPT_Legal?

  A. Test objective: Evidence of functional safety Acceptance criterion: The system recognizes failures in the transmission of information and data with the DPMA system and the evaluation system by means of self-tests.
  B. Test objective: Evidence of evolution Acceptance criterion: The quality of the research results does not deteriorate with further training.
  C. Test objective: Evidence of compatibility Acceptance criterion: The system can exchange information with the DPMA system and the evaluation system.
  D. Test objective: Evidence that the data is free from inappropriate bias Acceptance criterion: The DPMA's analysis data is statistically compared to data from other sources.
  B. Test objective: Evidence of evolution Acceptance criterion: The quality of the research results does not deteriorate with further training.

  ---

  The ISTQB CT-AI syllabus introduces AI-specific quality characteristics, including evolution, functional safety, compatibility, and bias-related data quality. Section 5.1 - AI-Specific Test Objectives explains that evolution refers to an AI system's capability to continue improving or at least maintain its performance as it undergoes additional training.

  GPT_Legal is explicitly described as a self-learning system expected to continuously reduce false positives, achieve weekly accuracy improvements of 10%, reach and maintain 90% accuracy, and adapt to new environments such as a patent law firm or a corporate legal department. This aligns directly with the syllabus definition of evidence of evolution, which focuses on ensuring that model quality does not degrade as new training data is introduced. Option B therefore best matches both the system

  characteristics and the acceptance criterion.

  Option A is not appropriate because functional safety is relevant for safety-critical domains, which do not apply to patent search and legal drafting. Option C addresses compatibility, which is important but not the primary AI-specific objective in this scenario. Option D relates to bias, which is a relevant concern but not central to the described goals of continuous learning and performance improvement.

  Thus, Option B is the most appropriate AI-specific test objective and acceptance criterion.