CT-AI ISTQB Certified Tester AI Testing Exam Questions and Answers

In a neural network, the activation value of a neuron is determined by a combination of inputs from the previous layer, the weights of the connections, and the bias at the neuron level. Here’s a detailed breakdown:

Inputs for Activation Value:

Activation Values of Neurons in the Previous Layer:These are the outputs from neurons in the preceding layer that serve as inputs to the current neuron.

Weights Assigned to the Connections:Each connection between neurons has an associated weight, which determines the strength and direction of the input signal.

Individual Bias at the Neuron Level:Each neuron has a bias value that adjusts the input sum, allowing the activation function to be shifted.

Calculation:

The activation value is computed by summing the weighted inputs from the previous layer and adding the bias.

Formula: z=∑(wi⋅ai)+bz = \sum (w_i \cdot a_i) + bz=∑(wi​⋅ai​)+b, where wiw_iwi​ are the weights, aia_iai​ are the activation values from the previous layer, and bbb is the bias.

The activation function (e.g., sigmoid, ReLU) is then applied to this sum to get the final activation value.

Why Option A is Correct:

Option A correctly identifies all components involved in computing the activation value: the individual bias, the activation values of the previous layer, and the weights of the connections.

Eliminating Other Options:

B. Activation values of neurons in the previous layer, and weights assigned to the connections between the neurons: This option misses the bias, which is crucial.

C. Individual bias at the neuron level, and weights assigned to the connections between the neurons: This option misses the activation values from the previous layer.

D. Individual bias at the neuron level, and activation values of neurons in the previous layer: This option misses the weights, which are essential.

Detecting biases in the ML pipeline involves various tests to ensure fairness and accuracy throughout the ML process.

Testing the distribution shift in the training data for inappropriate bias (A): This involves checking if there is any shift in the data distribution that could lead to bias in the model. It is an important test but not the most direct method for detecting biases.

Test the model during model evaluation for data bias (B): This is a critical stage where the model is evaluated to detect any biases in the data it was trained on. It directly addresses potential data biases in the model.

Testing the data pipeline for any sources for algorithmic bias (C): This test is crucial as it helps identify biases that may originate from the data processing and transformation stages within the pipeline. Detecting sources of algorithmic bias ensures that the model does not inherit biases from these processes.

Check the input test data for potential sample bias (D): While this is an important step, it focuses more on the input data and less on the overall data pipeline.

Hence, the most likely useful test to help detect different kinds of biases in the ML pipeline isB. Test the model during model evaluation for data bias.

Bias detection at thedata levelis performed by reviewingdata acquisition and preprocessing steps, as explained in Section2.3 – Data Quality and Biasof the ISTQB CT-AI syllabus. Sample bias arises when data is distorted or when preprocessing introduces unintended shifts—for example, by filtering, normalization, or labeling steps that disproportionately affect subsets of the data. OptionBcorrectly reflects this: reviewers can identify whether preprocessing steps have altered the dataset in a way that introducessample distortions. This aligns perfectly with syllabus guidance on reviewing data pipelines for bias sources.

Option A is incorrect because algorithmic bias originates from themodel, not data collection procedures. Option C is incorrect because LIME is anexplainabilitymethod applied post-model, not in data reviews. Option D incorrectly states “algorithmic bias,” but preprocessing affectssample bias, not algorithmic bias.

Thus, OptionBcorrectly matches the syllabus’ definition of how bias can be detected during data-related reviews.

The ISTQB CT-AI syllabus lists several characteristics that make it difficult to ensure safety in AI-based systems. Section2.8 – Safety and AIexplicitly names the characteristics that complicate safety assurance:complexity,non-determinism,probabilistic behavior,self-learning,lack of transparency, andlack of robustness. Among these,complexityis a core challenge because modern AI systems—particularly those using deep learning—have highly non-linear behavior, large numbers of parameters, and intricate interactions that are hard to predict.

Option B (Complexity) directly aligns with the syllabus and is therefore correct.

Option A (Determinism) is the opposite of AI behavior; AI is oftennon-deterministic, and determinism doesnotmake systems unsafe. Option C (Interpretability) does impact trust and explainability, but the syllabus positions it as a transparency challenge, not the primary difficulty in ensuring safety. Option D (Robustness) is a desired quality, not a reason safety is hard; alackof robustness would be a challenge, not robustness itself.

Thus,complexitybest reflects the syllabus’ explicit safety-related difficulty.

The syllabus defines clustering as:

“Clustering: This is when the problem requires the identification of similarities in input data points that allows them to be grouped based on common characteristics or attributes. For example, clustering is used to categorize different types of customers for the purpose of marketing.”

(Reference: ISTQB CT-AI Syllabus v1.0, Section 3.1.2, page 26 of 99)

The syllabus describes exploratory testing as especially useful when there are poor specifications or test oracle problems:

“Exploratory testing is especially useful when there are poor specifications or test oracle problems, which is often the case for AI-based systems.”

(Reference: ISTQB CT-AI Syllabus v1.0, Section 9.6, page 70 of 99)

The syllabus describes automation bias as:

"A type of bias caused by a person favoring the recommendations of an automated decision-making system over other sources."

This is also known as complacency bias, where testers accept automated system outputs without questioning them.

(Reference: ISTQB CT-AI Syllabus v1.0, Glossary, Page 89 of 99)

The ISTQB CT-AI syllabus explains in Section1.6 – Machine Learning Approachesthatsupervised learningis appropriate when labeled data exists and the goal is to predict an output based on known historical examples. Predicting a customer’spurchase probabilityis aclassificationtask when the output corresponds to discrete categories such as“likely to purchase”vs.“not likely to purchase.”The syllabus gives similar examples in describing classification as the process of assigning instances to predefined classes based on learned patterns in labeled data. Because the retail company wants to determine whether a customer will make a purchase based on marketing actions, classification is the most appropriate choice .

Option A (regression) predicts continuous numeric values and is less suitable because the task centers oncategorical likelihood, not estimating exact monetary values. Option C (unsupervised learning) is used when labels are not available—here, the company has labeled purchase histories. Option D (reinforcement learning) requires an interactive environment with reward-driven behavior, which is not applicable to this scenario.

Thus,supervised learning (classification)is the most suitable approach according to the syllabus.

 Factors Affecting ML Functional Performance:The data pipeline, quality of the labeling, and biased data are all factors that significantly affect the performance of machine learning models. The number of classes, while relevant for the model structure, is not a direct factor affecting the performance metrics such as accuracy or bias.

 Reference:ISTQB_CT-AI_Syllabus_v1.0, Sections on Data Quality and its Effect on the ML Model and ML Functional Performance Metrics​​​​.

The syllabus explains that ML models can be used to analyze reported defects and suggest which developers are best suited to fix them based on historical data about defect assignment and resolution speed:

"Assignment: ML models can suggest which developers are best suited to fix particular defects, based on the defect content and previous developer assignments."

(Reference: ISTQB CT-AI Syllabus v1.0, Section 11.2, page 78 of 99)

The ISTQB CT-AI syllabus describes theML data preparation workflowin Section2.2 – Data Preparation. Data preparation consists ofdata gathering,cleaning,transformation, andsampling. The syllabus emphasizes that one significant challenge duringdata gatheringis combining data frommultiple heterogeneous sources, which often differ in structure, quality, and format. Ensuring the resulting dataset is accurate, complete, and representative can be complex, making this a critical challenge in the ML workflow. This aligns directly with OptionC.

Option A is incorrect because erroneous data correction is part ofcleaning, not transformation. Option B contradicts the syllabus: while automation can help,not all steps should be automateddue to the need for expert oversight, especially in detecting subtle data quality issues. Option D is incorrect because sampling continues to involve risk—particularly around representativeness—and the syllabus emphasizes caution, not complacency.

Thus, OptionCis the only statement that accurately reflects the syllabus.

 A. Technological singularity

Technological singularity refers to a hypothetical point in the future when AI surpasses human intelligence and can continuously improve itself without human intervention. This scenario involves capabilities far beyond those described in the question.

 B. Narrow AI

Narrow AI, also known as weak AI, is designed to perform a specific task or a narrow range of tasks. It does not have general cognitive abilities and cannot learn multiple tasks by itself without retraining.

 C. Super AI

Super AI refers to an AI that surpasses human intelligence and capabilities across all fields. This is an advanced concept and not aligned with the description of having cognitive abilities similar to a young child.

 D. General AI

General AI, or strong AI, has the ability to understand, learn, and apply knowledge across a wide range of tasks, similar to human cognitive abilities. It aligns with the description of AI that can learn multiple tasks and operate in new environments without needing retraining.

The syllabus explains the fundamental distinction betweenconventional systemsandAI-based systems using supervised machine learningin Section1.3 – AI-Based and Conventional Systems. A conventional system relies on human-programmed logic—such as branches, conditions, and explicit rules—to interpret input data. The system behaves exactly as specified by its developers.

In contrast,AI systems using supervised learning automatically learn patternsfrom labeled data. The syllabus states that“patterns in data are used by the system to determine how it should react in the future… The AI determines on its own what patterns or features in the data can be used”. This aligns directly with Option C: an AI system identifies relevant diagnostic patterns in X-ray images during training, whereas a conventional system requires human experts to explicitly program those patterns.

Option A is incorrect because AI outputs are typicallylessexplainable, not more. Option B is incorrect because both systems can use thesame X-ray images; ML does not require structurally different images. Option D is oversimplified and not fully accurate; while training data is central to ML, AI systems also include architecture, algorithms, and preprocessing—not just data.

Thus,Option Cis the correct and syllabus-aligned answer.

The syllabus states:

“A/B testing can be used to test updates to an AI-based system where there are agreed acceptance criteria, such as ML functional performance metrics, as described in Chapter 5. A/B testing is used to compare the updated variant with the previous variant.”

(Reference: ISTQB CT-AI Syllabus v1.0, Section 9.4, page 68 of 99)

The syllabus states that while pairwise testing is effective at finding defects by reducing the number of test cases needed, the resulting test suite can still be extensive and require automation:

"Even the use of pairwise testing can result in extensive test suites… automation and virtual test environments often become necessary to allow the required tests to be run."

(Reference: ISTQB CT-AI Syllabus v1.0, Section 9.2, Page 67 of 99)

The question refers to a problem where data used for an object detection ML system was labelled incorrectly. This issue is most closely related to "accuracy issues." Here's a detailed explanation:

Accuracy Issues: The primary goal of labeling data in machine learning is to ensure that the model can accurately learn and make predictions based on the given labels. Incorrectly labeled data directly impacts the model's accuracy, leading to poor performance because the model learns incorrect patterns.

Why Not Other Options:

Security Issues: This pertains to data breaches or unauthorized access, which is not relevant to the problem of incorrect data labeling.

Privacy Issues: This concerns the protection of personal data and is not related to the accuracy of data labeling.

Bias Issues: While bias in data can affect model performance, it specifically refers to systematic errors or prejudices in the data rather than outright incorrect labeling.

The syllabus describes that data pre-processing includes cleaning (e.g., fixing or removing invalid data) and transforming data (e.g., changing data types such as numbers stored as strings).

"Transformation: The format of the given data is changed… converting categorical data into numerical data, changing image formats…"

(Reference: ISTQB CT-AI Syllabus v1.0, Section 4.1.1, Page 34 of 99)

The syllabus defines overfitting as:

“Overfitting is when the ML model learns the training data so well that it is unable to generalize to accommodate new data.”

This occurs when the model memorizes the training data, including noise, instead of learning the general patterns.

(Reference: ISTQB CT-AI Syllabus v1.0, Section 3.5.1, page 31 of 99)

According to Section4.6 – AI Behaviour Testing (AIB Testing)of the ISTQB CT-AI syllabus, AIB testing is used to evaluate changes in the functional behavior of self-learning systems. The core principle iscomparing pre-change and post-change model behavior using the same test inputs, so that any difference in outputs can be attributed to the model’s learning and not to differences in input data. This directly corresponds to OptionC.

Option A is incorrect because the absence of a test oracle does not justify generating new test cases; AIB relies onreusing identical inputsto detect behavioral drift. Option B is invalid because using different inputs prevents meaningful comparison. Option D is incorrect because comparing with an unrelated non-self-learning system does not allow evaluation of the same model’s behavioral evolution.

Thus, OptionCaccurately represents the correct application of AIB testing: assessing model behavior changes by running identical test inputs before and after learning updates.

The ISTQB CT-AI syllabus emphasizes that testing AI-based systems requirescross-functional collaborationandexperience-based testingwhen parts of the team lack domain knowledge. In this scenario, the ML expert understands ML and dataset preparation but lacks knowledge ofcamera system behavior, the device’s operational data pipeline, and end-user workflows. The remainder of the team understands the domain and system testing but not ML. Section4.4 – Human Factors and AI Testingand4.3 – System Testing of AI Componentshighlight that when domain understanding is unevenly distributed,experience-based testing conducted by the full team(testers, developers, domain experts) is the most effective approach. This ensures that AI outputs align with actual user expectations and system behavior. OptionCaligns exactly with this principle.

Option A is too limited and does not address the need to validate ML integration. Option B is incorrect because reusing old test cases overlooks AI-specific risks in the operating data pipeline. Option D is useful but focuses only on data representativeness, not system-level user validation. Therefore,Option Cis the best, syllabus-aligned test approach.

To calculate the accuracy, recall, and specificity from the confusion matrix provided, we use the following formulas:

Confusion Matrix:

Actually Rotten: 45 (True Positive), 8 (False Positive)

Actually Fresh: 5 (False Negative), 42 (True Negative)

Accuracy:

Accuracy is the proportion of true results (both true positives and true negatives) in the total population.

Formula: Accuracy=TP+TNTP+TN+FP+FN\text{Accuracy} = \frac{TP + TN}{TP + TN + FP + FN}Accuracy=TP+TN+FP+FNTP+TN​

Calculation: Accuracy=45+4245+42+8+5=87100=0.87\text{Accuracy} = \frac{45 + 42}{45 + 42 + 8 + 5} = \frac{87}{100} = 0.87Accuracy=45+42+8+545+42​=10087​=0.87

Recall (Sensitivity):

Recall is the proportion of true positive results in the total actual positives.

Formula: Recall=TPTP+FN\text{Recall} = \frac{TP}{TP + FN}Recall=TP+FNTP​

Calculation: Recall=4545+5=4550=0.9\text{Recall} = \frac{45}{45 + 5} = \frac{45}{50} = 0.9Recall=45+545​=5045​=0.9

Specificity:

Specificity is the proportion of true negative results in the total actual negatives.

Formula: Specificity=TNTN+FP\text{Specificity} = \frac{TN}{TN + FP}Specificity=TN+FPTN​

Calculation: Specificity=4242+8=4250=0.84\text{Specificity} = \frac{42}{42 + 8} = \frac{42}{50} = 0.84Specificity=42+842​=5042​=0.84

Therefore, the correct combinations of accuracy, recall, and specificity are 0.87, 0.9, and 0.84 respectively.

The ISTQB CT-AI syllabus explains that reusing pre-trained models is strongly related tosimilarity between the original task and the new task. Section1.8 – Pre-trained Models and Transfer Learningstates that reuse is effective when the new task is similar to the original one, such as adapting a cat-classifier to classify dog breeds. The syllabus warns about risks related toinput differences,data preparation inconsistencies,inherited shortcomings, andexplainability issues. These are legitimate objections (matching options A, B, and C) because large differences in image inputs or patterns can undermine transfer learning; misclassification risk can increase; and explainability often decreases when reusing pre-trained models .

However,output differences are NOT a valid concernhere. Both the leaf-based and flower-based ML models classifythe same plant species, meaning theiroutputs are identical. The syllabus does not identify output mismatch as a transfer-learning risk. Real risks concerninputs,bias inheritance,model transparency, andtraining differences—not output labels. Therefore, OptionDdescribes an objection that can be safelyneglected, because output classes are the same and do not hinder reuse.

From the confusion matrix:

True Positives (TP) = 15

False Positives (FP) = 5

False Negatives (FN) = 15

True Negatives (TN) = 65

Accuracy= (TP + TN) / Total

= (15 + 65) / 100

=80%

Precision= TP / (TP + FP)

= 15 / (15 + 5)

= 15 / 20

=75%

Section3.2 – Functional Performance Criteriain the syllabus explains accuracy and precision exactly these ways when evaluating ML classification performance.

Option B is therefore the only correct pair of values.

 Labelling Approaches:Among the options provided, pre-labeled datasets require the least time and effort because the data has already been labeled, eliminating the need for further manual or automated labeling efforts.

 Reference:ISTQB_CT-AI_Syllabus_v1.0, Section 4.5 Data Labelling for Supervised Learning, which discusses various approaches to data labeling, including pre-labeled datasets, and their associated time and effort requirements​​.

The syllabus states:

"Transparency: This is considered to be the ease with which the algorithm and training data used to generate the model can be determined."

Access to the training data is an example of transparency.

(Reference: ISTQB CT-AI Syllabus v1.0, Section 2.7, page 24 of 99)

According to the ISTQB CT-AI syllabus,metamorphic testingworks by applyingmetamorphic 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 MR while keeping the expected output direction predictable.

Test Alowers v_mean compared to Test 1, with similar acceleration values. This directly satisfies the MR that lower speed → higher score.

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

Test Dmodifies 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.

ThusTest B cannot be considered a metamorphic follow-up, which makesOption Ccorrect.

 A. Using a pixel comparison of the GUI before and after the change to check the differences.

Pixel comparison is a traditional method and does not involve AI. It compares images at the pixel level, which can be effective but is not an intelligent approach. It is not considered an AI usage and is the least likely usage of AI for detecting GUI changes.

 B. Using computer vision to compare the GUI before and after the test object changes.

Computer vision involves using AI techniques to interpret and process images. It is a likely usage of AI for detecting changes in the GUI.

 C. Using vision-based detection of the GUI layout changes before and after test object changes.

Vision-based detection is another AI technique where the layout and structure of the GUI are analyzed to detect changes. This is a typical application of AI.

 D. Using a ML-based classifier to flag if changes in GUI are to be flagged for humans.

An ML-based classifier can intelligently determine significant changes and decide if they need human review, which is a sophisticated AI application.

 Technology Most Typically Used to Implement AI:Genetic algorithms are a well-known technique used in AI. They are inspired by the process of natural selection and are used to find approximate solutions to optimization and search problems. Unlike search engines, procedural programming, or case control structures, genetic algorithms are specifically designed for evolving solutions and are commonly employed in AI implementations.

 Reference:ISTQB_CT-AI_Syllabus_v1.0, Section 1.4 AI Technologies, which identifies different technologies used to implement AI​​.

The syllabus defines reward hacking as:

"Reward hacking can result from an AI-based system achieving a specified goal by using a 'clever' or 'easy' solution that perverts the spirit of the designer’s intent."

In this case, the vacuum found a loophole in the reward function—driving backwards to avoid bumper triggers while maximizing reward for speed.

(Reference: ISTQB CT-AI Syllabus v1.0, Section 2.6, page 24 of 99)

TheAI Effectis clearly defined in theISTQB Certified Tester AI Testing Syllabus v1.0under Section1.1 – Definition of AI and AI Effect. The document explains that society’s understanding of what qualifies as “AI” changes over time. Technologies once considered AI—such as expert systems from the 1970s and 1980s or early chess-playing systems—are no longer viewed as AI today. This phenomenon is explicitly labeled the“AI Effect,”described as“the changing perception of what constitutes AI.”The syllabus states that as AI capabilities become routine or widely implemented, they often stop being perceived as true artificial intelligence .

Options B, C, and D do not capture this definition. While AI learning from data (B) is a property of ML, it does not describe the shifting perception of AI. Option C describes a technological achievement, not a perceptual shift. Option D references a historical AI milestone (Deep Blue defeating Kasparov) that the syllabus specifically uses as an example of technology that is no longer considered AI due to the AI Effect. Therefore, onlyOption Aaccurately reflects the AI Effect as defined by the ISTQB syllabus.

Section2.2 – Data Preparationand4.1 – Challenges in Testing AI-Based Systemsdescribe difficulties in obtaining and managing large, representative datasets. AI-based systems requirerealistic, diverse, and representativedata reflecting real-world variations. The syllabus emphasizes that assembling such datasets is time-consuming, resource-intensive, and often constrained by availability, privacy, or domain complexity. Option B directly corresponds to these documented challenges.

Option A is incorrect: using the same implementation risksdefect masking, not preventing it; the syllabus warns against this practice. Option C is incorrect because real-world data naturally evolves, and the syllabus notes thatdriftis normal; expecting stable input data contradicts operational reality. Option D is incorrect: although data privacy is important, the syllabus does not claim that artificially generated data always requires legal approval, nor that sanitization/encryption is mandatory for synthetic data.

Thus,Option Baccurately reflects syllabus-defined difficulties in producing representative test data.

The syllabus lists multiple reasons for mislabelled data, including the lack of domain knowledge:

"Lack of required domain knowledge may lead to incorrect labelling."

(Reference: ISTQB CT-AI Syllabus v1.0, Section 4.5.2, page 38 of 99)

Back-to-back testing is a method where the same set of tests are run on multiple implementations of the system to compare their outputs. This type of testing is typically used to ensure consistency and correctness by comparing the outputs of different implementations under identical conditions. Let's analyze the options given:

A. Comparison of the results of a current neural network model ML model implemented in platform A (for example Pytorch) with a similar neural network model ML model implemented in platform B (for example Tensorflow), for the same data.

This option describes a scenario where two different implementations of the same type of model are being compared using the same dataset. This is a typical back-to-back testing situation.

B. Comparison of the results of a home-grown neural network model ML model with results in a neural network model implemented in a standard implementation (for example Pytorch) for the same data.

This option involves comparing a custom implementation with a standard implementation, which is also a typical back-to-back testing scenario to validate the custom model against a known benchmark.

C. Comparison of the results of a neural network ML model with a current decision tree ML model for the same data.

This option involves comparing two different types of models (a neural network and a decision tree). This is not a typical scenario for back-to-back testing because the models are inherently different and would not be expected to produce identical results even on the same data.

D. Comparison of the results of the current neural network ML model on the current data set with a slightly modified data set.

This option involves comparing the outputs of the same model on slightly different datasets. This could be seen as a form of robustness testing or sensitivity analysis, but not typical back-to-back testing as it doesn’t involve comparing multiple implementations.

Based on this analysis, optionCis the one that describes a situation of back-to-back testing the least because it compares two fundamentally different models, which is not the intent of back-to-back testing.