anki_GARP_Risk-AI-Full_Question_Bank Flashcards
1
Q
What does GOFAI stand for?
A
A
2
Q
Which of the following best describes reinforcement learning?
A
B
3
Q
Which AI technique is most associated with learning from labeled data?
A
C
4
Q
In deep learning, what is a common characteristic that distinguishes it from classical AI?
A
C
5
Q
Which of the following is NOT a type of machine learning?
A
D
6
Q
Recursion is a technique that involves:
A
B
7
Q
What type of AI model typically operates as a ‘black box’ due to its complexity?
A
C
8
Q
Which search method is characterized by exploring all possible options at one level before moving to the next level?
A
B
9
Q
Which of the following is a common application of unsupervised learning?
A
C
10
Q
The concept of ‘trial and error’ is most closely associated with which type of machine learning?
A
C
11
Q
Which AI technique is specifically designed to mimic human decision-making in game scenarios, such as chess or tic-tac-toe?
A
B
12
Q
Which of the following best defines the term ‘heuristic’ as used in AI?
A
A
13
Q
Which of the following is a characteristic of supervised learning?
A
A
14
Q
A neural network with multiple layers that can learn hierarchical representations of data is an example of:
A
C
15
Q
What is the main purpose of dimensionality reduction in AI?
A
A
16
Q
Which of the following best describes binary search?
A
B
17
Q
Depth-first search (DFS) is most suitable for which of the following scenarios?
A
C
18
Q
In reinforcement learning, what is the purpose of a reward function?
A
B
19
Q
What is the primary difference between breadth-first search (BFS) and depth-first search (DFS)?
A
B
20
Q
Which of the following is a characteristic of A* search?
A
A
21
Q
Which of the following is minimized in the K-means clustering algorithm to optimize clustering performance?
A
B
22
Q
In clustering, which metric evaluates the separation between clusters based on the average distance between points in one cluster and points in the nearest other cluster?
A
B
23
Q
When using a scree plot to determine the optimal number of clusters, what feature of the plot indicates this number?
A
B
24
Q
Which formula correctly represents the Variance Ratio Criterion (Calinski-Harabasz index) in clustering?
A
A
25
The silhouette score, which measures clustering quality, can range between which values?
B
26
Which of the following limitations is associated with the K-means clustering algorithm?
B
27
The “curse of dimensionality” affects clustering algorithms by:
C
28
In which situation would hierarchical clustering be preferred over K-means?
C
29
Which of the following is a primary reason for using DBSCAN over K-means clustering?
B
30
Which distance measure is used in Mahalanobis distance calculations for clustering analysis?
B
31
A bank uses AI models for detecting fraudulent transactions. Which of the following statements correctly describes inscrutability in AI models?
A
32
In March 1997, Deep Blue, an AI model developed by IBM, defeated Garry Kasparov. Which of the following best describes a characteristic of classical AI?
C
33
Which of the following best describes a societal risk scenario regarding the impact of AI?
B
34
One of the key breakthroughs of AI and machine learning is the development of deep learning. Which of the following is a key differentiating feature of deep learning over classical AI?
D
35
To tailor its products, a private wealth advisor surveyed potential customers. What is the most appropriate way to classify the survey results?
B
36
As part of cleaning and preparing a data set, which statement is correct regarding data scaling?
D
37
An analyst examines a scatter plot of data for K-means clustering with single linkage. Which of the following best identifies an issue with the data?
D
38
An analyst suspects that a polynomial regression is best for modeling salary vs. age. To select a model, the analyst uses forward stepwise regression. Based on AIC values, which model should be chosen?
A
39
A data scientist is working to improve the accuracy of a decision tree using a random forest technique. Which of the following best describes an advantage of using a random forest?
C
40
A software engineer is using a facial recognition system with an input matrix X and kernel matrix W. What is the correct feature map (F) given X and W?
A
41
An analyst is using an ensemble technique to improve a decision tree model. Which technique involves sampling with replacement and averaging multiple decision tree results?
D
42
An asset manager wants to classify stocks into categories using SVMs. Based on a boundary line equation, how should a stock with values 0.8 and -0.75 be classified?
A
43
An analyst working on text classification splits features into two sets for co-training. What is an advantage of co-training?
A
44
An analyst trained four models to classify consumer loans into Default or Non-Default. Which model correctly identifies the highest percentage of defaulted loans?
C
45
A newly hired analyst is asked to 'tokenize' a text passage for an NLP application. Which of the following best describes tokenization?
B
46
A data scientist is implementing a model architecture most likely representing which of the following?
B
47
Consider a medical diagnostic algorithm for a rare disease. Which measure would you use to assess the algorithm's performance?
A
48
A company is testing the fairness of a model predicting purchase decisions. How does demographic parity contribute to group fairness?
B
49
A bank implemented a model to detect fraudulent transactions in a dataset with only 20 fraud cases in 10,000 samples. What bias is most likely to occur?
C
50
A logistic model determines job suitability based on experience and education level. What approach reduces the influence of education level in predictions?
B
51
A bank uses a chatbot for customer inquiries, which occasionally provides incorrect information. What action should the bank take to address this issue?
D
52
A credit card company develops a model to analyze transaction patterns for fraud detection. Which action would increase the model’s explainability?
A
53
A sports analyst develops an AI system to predict league winners. Which technique best explains the marginal effect of each feature?
A
54
A bank uses a model to determine loan approvals. Concerned about transparency, customers question the decision process. What is the source of reputational risk?
C
55
A technology company wants to minimize risks related to customer information. What step should the company take?
B
56
Which of the following is considered an advantage of implementing a practical ethics framework?
A
57
A local bank conducts periodic validation of a Value at Risk (VaR) model. Which of the following is considered best practice for validation?
B
58
The Chief Risk Officer (CRO) oversees model risk governance for an investment firm. What is crucial when evaluating the model landscape?
D
59
A sovereign wealth fund reviews a banking organization’s model governance framework. Which describes appropriate roles in model governance?
D
60
Which team ensures proper model integration with IT systems in a large organization?
B
61
In financial institutions, why is frequent monitoring of machine learning models advantageous?
B
62
Compared to traditional statistical models, what is the likely data quality requirement for AI/ML models?
D
63
A bank’s new AI model outperformed an existing model in testing. What should be the next step?
D
64
A bank uses an AI-based credit scoring model with alternative data. What is an appropriate finding by the validation team?
B
65
Utilitarianism is one ethical theory in AI ethics. Which statement is correct about Utilitarianism?
A
66
A company board considers an AI ethics framework. What advantage does it provide?
D
67
Which best describes a characteristic of a deontological ethical framework?
B
68
Which of the following is correct regarding the principle of justice in AI ethics?
A
69
An AI system recommends treatments that conflict with patient preferences. Which ethics principle is violated?
C
70
A pharmaceutical model trained on data from specific demographics performs poorly in different regions. What is the likely cause?
D
71
A large bank has implemented a machine learning model for fraud detection. Which risk is most likely if the model is applied outside its original domain?
C
72
Which of the following best describes the purpose of model validation in a financial institution?
C
73
An AI-driven hiring tool scores candidates based on language use in applications. Which bias might arise if specific dialects are penalized?
B
74
In a text analysis task, a data scientist removes stop words to improve model accuracy. What is the primary purpose of this process?
D
75
In AI ethics, which term describes a model's ability to provide understandable reasons for its decisions?
B
76
A bank uses an AI model to classify loan applicants. What step can the bank take to enhance model transparency?
B
77
Which of the following accurately describes the concept of demographic parity in AI fairness?
B
78
A credit scoring model is used in several countries with different demographics. Which term best describes the need to ensure the model is fair across all groups?
A
79
A data scientist designs a model for text classification. Which technique improves explainability by showing the impact of each word on predictions?
C
80
In data preparation for machine learning, which scaling technique ensures that all features are in the same range without changing their relative differences?
B
81
A company is evaluating the effectiveness of an ethics framework. Which of the following best describes a primary benefit of such a framework?
B
82
When assessing model risk in financial institutions, what is a primary goal of model validation?
C
83
Which term describes the bias introduced when training data reflects outdated social patterns that persist in model predictions?
C
84
An AI system that uses personal data without user consent is likely violating which ethical principle?
A
85
A large organization uses an AI-driven HR tool that consistently favors certain demographics. Which action helps mitigate this algorithmic bias?
C
86
Which of the following is a primary benefit of increasing model explainability in high-stakes AI applications?
B
87
A financial institution aims to achieve fairness across demographic groups in loan approval models. Which fairness metric ensures equal opportunity across groups?
C
88
Which AI technique is most appropriate for clustering high-dimensional data with varying densities?
C
89
In AI ethics, which concept ensures that AI decisions do not disproportionately harm any particular group?
B
90
A model used in healthcare predicts treatments based on patient data. What is a primary concern if the model lacks transparency?
A
91
In machine learning, which of the following describes 'overfitting'?
A
92
What is a common technique to prevent overfitting in machine learning models?
C
93
In AI ethics, which term refers to the responsibility to ensure that AI systems are designed with accountability in mind?
D
94
A financial institution is assessing model risk. Which step is crucial for ensuring model alignment with real-world applications?
B
95
An AI system used in credit scoring has been found to disproportionately favor one demographic over others. What is a recommended mitigation technique?
B
96
Which of the following best defines 'predictive parity' in AI fairness?
A
97
A bank uses an AI model for risk assessment but faces challenges with model interpretability. What is one approach to improve interpretability?
B
98
Which of the following is a consequence of high dimensionality in data?
B
99
How are categorical data types typically processed in machine learning?
B
100
How are categorical variables commonly encoded for ML models?
B
101
How does data normalization help in machine learning?
B
102
How does data standardization differ from normalization?
C
103
How does label encoding differ from one-hot encoding?
C
104
In data analysis, which data type has both a meaningful zero and equal intervals?
D
105
In machine learning, what is the main function of a validation data subset?
B
106
In machine learning, why is data cleaning essential?
B
107
In principal component analysis (PCA), what do principal components represent?
B
108
In supervised learning, which data subset is used to adjust the model’s parameters?
A
109
What is a benefit of using semi-supervised learning over supervised learning?
B
110
What is a key difference between supervised and unsupervised learning?
A
111
What is the benefit of cleaning data before analysis?
D
112
What is the main purpose of splitting data into training, validation, and test sets?
C
113
What is the primary difference between machine-learning techniques and classical econometrics?
B
114
What is the primary goal of feature scaling?
B
115
What is the primary goal of principal component analysis (PCA)?
C
116
What is the primary use of validation data?
B
117
What is the purpose of applying transformations like log or square root to data?
B
118
What is the purpose of principal component analysis (PCA) in machine learning?
B
119
What is the purpose of splitting data into training, validation, and test sets?
C
120
What is the role of test data in machine learning?
C
121
What type of data is 'age' typically classified as in machine learning?
C
122
Which data preparation step is crucial before applying machine learning algorithms that rely on distances?
B
123
Which data subset is used to evaluate model performance after all training is completed?
C
124
Which data type includes categories with no inherent order?
B
125
Which method helps to reduce the dimensionality of a dataset while retaining important information?
C
126
Which technique is commonly used to encode binary categorical variables?
B
127
Which transformation is commonly applied to reduce skewness in data?
A
128
Which transformation technique is commonly applied to make data more normally distributed?
C
129
Which type of data preparation technique is often applied to handle missing values?
B
130
Why is it important to identify and address missing values in a dataset?
C
131
Why is one-hot encoding used for categorical variables?
B
132
Why might outliers be removed from a dataset?
C
133
Why might principal component analysis (PCA) be used in data preparation?
B
134
Which of the following is a primary purpose of unsupervised learning?
B
135
In the context of unsupervised learning, clustering can best be described as:
B
136
What is the primary difference between hierarchical clustering and partitional clustering methods?
C
137
Which of the following methods is specifically designed to handle non-spherical clusters in clustering analysis?
C
138
The K-means algorithm is often applied in clustering due to which of the following advantages?
B
139
Which of the following is the correct first step in the K-means clustering algorithm?
B
140
In K-means clustering, the process of recalculating centroids involves:
B
141
What is the purpose of the K-means++ initialization in K-means clustering?
B
142
When using hierarchical clustering with single linkage, the distance between clusters is calculated by:
B
143
Density-based clustering methods, like DBSCAN, are particularly useful in situations where:
D
144
Which of the following is minimized in the K-means clustering algorithm to optimize clustering performance?
B
145
In clustering, which metric evaluates the separation between clusters based on the average distance between points in one cluster and points in the nearest other cluster?
B
146
When using a scree plot to determine the optimal number of clusters, what feature of the plot indicates this number?
B
147
Which formula correctly represents the Variance Ratio Criterion (Calinski-Harabasz index) in clustering?
A
148
The silhouette score, which measures clustering quality, can range between which values?
B
149
Which of the following limitations is associated with the K-means clustering algorithm?
B
150
The “curse of dimensionality” affects clustering algorithms by:
C
151
In which situation would hierarchical clustering be preferred over K-means?
C
152
Which of the following is a primary reason for using DBSCAN over K-means clustering?
B
153
Which distance measure is used in Mahalanobis distance calculations for clustering analysis?
B
154
What does the intercept term represent in a simple linear regression model?
A
155
Which of the following methods is most commonly used for estimating parameters in linear regression?
A
156
In multiple linear regression, what does each coefficient estimate?
A
157
Which of the following is a common issue with regression models?
B
158
What is multicollinearity, and why can it be problematic?
A
159
What is the main purpose of Cook�s distance in regression analysis?
A
160
Why can�t linear regression be used for binary classification?
B
161
What transformation does logistic regression apply to ensure predictions are between 0 and 1?
B
162
In logistic regression, what is the role of the logit function?
C
163
Which classification method assumes that classes have multivariate normal distributions with different means but a shared covariance matrix?
B
164
What is the primary application of linear discriminant analysis (LDA)?
B
165
In the context of LDA, what does the discriminant function calculate?
B
166
What is the purpose of stepwise regression in model building?
B
167
Which criterion is often used in stepwise regression to determine the model performance?
B
168
How does heteroskedasticity affect a linear regression model?
B
169
What approach does the Heckman two-stage procedure address?
C
170
What type of regression model is suitable for multi-class classification with ordered categories?
C
171
What is the primary difference between regression and classification decision trees?
B
172
Which of the following is a purpose of pruning in decision trees?
C
173
What is a disadvantage of decision trees compared to other machine learning models?
B
174
What is the primary goal of using ensemble techniques like bagging and boosting?
C
175
In random forests, what technique is used to reduce correlation among individual trees?
C
176
How does boosting differ from bagging in terms of training trees?
B
177
Why is K-nearest neighbors (KNN) considered a �lazy learner�?
C
178
What is the primary factor that influences the choice of K in KNN?
C
179
In support vector machines, what is the main goal of the margin?
B
180
How do SVMs handle linearly inseparable data?
B
181
What is the purpose of an activation function in a neural network?
B
182
In a neural network with multiple classes, which activation function is often used in the output layer?
C
183
What is backpropagation used for in a neural network?
B
184
What is the primary purpose of autoencoders in machine learning?
B
185
How do autoencoders achieve dimensionality reduction compared to PCA?
B
186
When is semi-supervised learning most useful?
C
187
How does semi-supervised learning differ from supervised and unsupervised learning?
C
188
Which assumption implies that data points within a cluster are likely to share the same label?
B
189
What does the manifold assumption state?
B
190
What is the primary goal of self-training in semi-supervised learning?
B
191
What is a potential drawback of self-training?
C
192
In self-training, how are labels assigned to unlabeled data?
B
193
What is a key feature of co-training that differentiates it from self-training?
B
194
How does co-training help reduce overfitting?
B
195
What is the main characteristic of transductive methods in semi-supervised learning?
B
196
Which technique is an example of a transductive method in semi-supervised learning?
C
197
What is the purpose of inductive methods in semi-supervised learning?
B
198
Which of the following best describes the 'cluster-then-label' approach?
A
199
How does pre-training in semi-supervised learning help model performance?
B
200
In co-training, how are labels assigned to unlabeled data in each subset?
B
201
What is the main objective of reinforcement learning?
B
202
Which of the following best describes the output of reinforcement learning?
C
203
In the Multi-Arm Bandit problem, what is the role of the agent?
B
204
Which strategy focuses solely on the best-known action without exploring other options?
C
205
What is the purpose of the ?-greedy strategy in reinforcement learning?
C
206
In Markov Decision Processes (MDPs), what does the Markov property assume?
D
207
What does the discount factor in MDPs represent?
B
208
Which of the following methods is classified as a value-based approach in reinforcement learning?
D
209
How does Q-learning differ from policy-based approaches?
A
210
In policy gradient methods, what is the primary goal?
B
211
Which reinforcement learning method evaluates rewards only after completing an entire episode?
B
212
How does Temporal Difference learning differ from the Monte Carlo method?
A
213
What challenge in reinforcement learning does deep reinforcement learning address?
B
214
In deep reinforcement learning, what role does a neural network play?
B
215
What is one advantage of Temporal Difference learning over Monte Carlo methods in reinforcement learning?
A
216
What does the Ordinary Least Squares (OLS) method aim to minimize?
B
217
Which parameter estimation method uses a likelihood function to maximize the probability of observing the data?
C
218
Why is OLS often preferred over Maximum Likelihood Estimation (MLE) in linear regression?
B
219
What is the primary purpose of gradient descent in parameter optimization?
C
220
In the context of neural networks, what does backpropagation help determine?
C
221
What is a common issue with gradient descent that momentum helps to address?
A
222
What is the main consequence of overfitting a model?
B
223
What is one technique used to prevent overfitting by simplifying the model?
B
224
What does the bias-variance trade-off illustrate?
B
225
Which regularization method penalizes the absolute values of coefficients to encourage feature selection?
B
226
What is a primary distinction between LASSO and Ridge Regression?
B
227
Which regularization method combines the penalties of both Ridge and LASSO?
A
228
What is the purpose of cross-validation in model estimation?
C
229
How does stratified cross-validation improve upon regular k-fold cross-validation?
B
230
What is the main objective of a grid search in the context of machine learning?
B
231
What is the primary purpose of using Mean Squared Error (MSE) in model evaluation?
B
232
Which of the following is an advantage of Root Mean Square Error (RMSE) over MSE?
B
233
What limitation does Mean Absolute Error (MAE) overcome that is present in MSE?
C
234
Why might Mean Absolute Percentage Error (MAPE) be preferred in some evaluation contexts?
A
235
What does the accuracy metric indicate in a confusion matrix?
B
236
In model evaluation, which metric is highly sensitive to class imbalance?
C
237
Which metric helps in assessing the likelihood that a model's positive prediction is correct?
C
238
What does the recall metric indicate in classification?
A
239
When is the F1 score particularly useful in model evaluation?
A
240
What does the Receiver Operating Characteristic (ROC) curve illustrate in model evaluation?
B
241
What does a higher Area Under the Curve (AUC) value signify?
B
242
In evaluating models with different performance metrics, why might one prefer a model with lower MAPE over one with lower MSE?
A
243
What is a common use case for the confusion matrix in model evaluation?
B
244
How is the true negative rate (specificity) useful in certain applications?
B
245
Which evaluation measure is most intuitive when presenting model performance to non-technical stakeholders?
B
246
What is one common application of natural language processing (NLP) in finance?
B
247
What type of data does NLP typically work with?
B
248
What is the purpose of tokenization in text pre-processing?
B
249
Why are 'stop words' removed in NLP pre-processing?
C
250
What is 'stemming' in NLP pre-processing?
A
251
What is the Bag of Words (BoW) approach used for in NLP?
B
252
What does Term Frequency-Inverse Document Frequency (TF-IDF) measure?
B
253
Why might a high TF-IDF score for a word be significant?
B
254
In sentiment analysis using a dictionary approach, what is the primary purpose of the dictionary?
B
255
What is the primary difference between a generative and discriminative classifier in NLP?
A
256
What is one advantage of the Na�ve Bayes classifier in document classification?
C
257
Which NLP technique can improve the accuracy of a Bag of Words model by considering pairs or sequences of words?
A
258
How can the fit of an NLP model be evaluated when the data has labeled classes?
A
259
What does a high F1 score signify in an NLP classification task?
B
260
What is one challenge of evaluating an NLP model in unsupervised contexts (e.g., social media analysis)?
A
261
What key feature allows transformer models to capture long-range dependencies in text?
C
262
What is the primary function of large language models (LLMs) in generative AI?
C
263
Which type of neural network is designed to capture word dependencies across sequences?
A
264
What is the role of a discriminator in a Generative Adversarial Network (GAN)?
C
265
Why is a Variational Autoencoder (VAE) particularly suited for modifying existing images?
B
266
What distinguishes an autoencoding LLM from an autoregressive LLM?
B
267
In Word2Vec, which model uses the "fill-in-the-blank" approach to predict a missing word?
B
268
What benefit does Word2Vec provide over traditional Bag of Words?
B
269
What function does the encoder perform in the original transformer model?
C
270
How does the "temperature" parameter affect the output in generative AI?
B
271
In the context of LLMs, what is "prompt engineering"?
A
272
What is one common application of generative AI in finance?
A
273
How does the continuous skip-gram model in Word2Vec operate?
B
274
Which type of generative AI is known for using probabilistic models to reconstruct images based on text prompts?
B
275
What is the main purpose of reinforcement learning with human feedback (RLHF) in generative AI models like GPT?
B
276
What is the definition of algorithmic bias?
B
277
Which measure assesses whether predictions are equally distributed across groups, regardless of accuracy?
B
278
What does the concept of equal opportunity aim to ensure?
B
279
What is one common source of algorithmic bias related to data collection?
B
280
What is the primary difference between explainability and interpretability in AI?
B
281
Which XAI technique ranks the influence of input features on model outcomes?
C
282
What is a common limitation of using surrogate models in XAI?
C
283
How can AI systems impact human autonomy?
B
284
Which example best illustrates manipulation through AI?
B
285
What was demonstrated by the Facebook emotional contagion experiment?
D
286
What are primary causes of reputational damage related to AI?
A
287
What is meant by �existential risk� in AI?
B
288
How does global inequality relate to AI development and adoption?
C
289
What does �fairness through unawareness� mean in the context of AI?
B
290
Which of the following is a significant concern in using generative AI to manipulate public opinion?
B
291
What is one potential business benefit of implementing a practical ethics framework?
C
292
Which of the following ethical frameworks is focused on the consequences of actions?
C
293
Which principle in medical ethics emphasizes not inflicting harm on others?
C
294
How does deontology differ from consequentialism?
B
295
Which principle of AI ethics is primarily concerned with giving individuals control over their data?
C
296
What is 'data minimization' in the context of AI and privacy?
B
297
What type of bias arises when data does not represent the broader population?
A
298
How does the principle of beneficence apply to AI?
C
299
Which European regulation mandates transparency and limits personalization in large platforms?
B
300
Which U.S. initiative addresses AI in national security and safety through risk management?
B
301
What is the main goal of the EU AI Act?
A
302
Which principle of AI ethics emphasizes transparency in decision-making?
B
303
What is one of the main benefits of explainability in AI systems?
B
304
What is local explainability in AI?
B
305
Which term refers to allowing users to control and access their personal information within an AI system?
C
306
What is the primary purpose of a data governance framework?
C
307
Which aspect of data governance is most concerned with ensuring data accuracy and reliability?
B
308
What is 'data provenance' in the context of AI model governance?
B
309
Why is data classification essential in a data governance strategy?
B
310
What is an example of ensuring compliance in data governance?
A
311
What is one key responsibility of model governance?
C
312
Why are model validation activities necessary?
B
313
Which aspect of model governance involves evaluating potential model risks based on their impact?
C
314
What is the purpose of including AI/ML applications in the model inventory?
B
315
Which principle in AI model governance emphasizes the need for transparent data usage and transformation?
B
316
Who is primarily responsible for establishing a firm�s model governance framework?
C
317
What is one primary goal of ongoing model validation?
B
318
What does the term 'effective challenge' mean in model validation?
B
319
Which test examines a model's behavior in extreme scenarios or with unusual data?
B
320
What is one common outcome of poor model validation practices?
C
321
What is a common cause of misalignment in AI/ML model objectives?
B
322
Why might a model optimized for efficiency be misaligned in healthcare applications?
B
323
Which of the following describes the purpose of stakeholder engagement in mitigating misalignment risk?
C
324
What does the Value-Sensitive Design (VSD) principle primarily focus on?
B
325
Which framework can be used to align model objectives across different organizational teams?
C
326
How can constraint-based optimization aid in addressing misalignment in AI models?
B
327
Which real-world issue exemplifies the risks of a misaligned AI model in financial trading?
B
328
In healthcare, how can a misaligned diagnostic model affect patient outcomes?
B
329
What risk is introduced by relying solely on 'accuracy' as a success metric for a medical diagnosis AI model?
A
330
Why is ambiguity in success metrics particularly risky in safety-critical fields like healthcare?
B
331
What is a common cause of ambiguity in setting success metrics for AI/ML models?
B
332
Which of the following would be an effective mitigation technique for addressing ambiguous success metrics?
B
333
In the context of ambiguous success metrics, what is the purpose of scenario analysis?
B
334
Which metric is suggested for healthcare applications to account for patient safety?
B
335
How can cost-sensitive learning techniques help address ambiguity in success metrics?
B
336
What risk arises in criminal justice when using accuracy alone to evaluate a recidivism prediction model?
B
337
What is a primary cause of unintended consequences in AI models?
B
338
Why might a customer service chatbot optimized for engagement lead to unintended consequences?
A
339
How can unintended consequences arise from AI in financial applications?
B
340
What approach is recommended to mitigate unintended consequences in AI/ML models?
C
341
How can unintended consequences in healthcare AI systems impact patient outcomes?
B
342
What is one common cause of data quality issues in AI/ML models?
B
343
How can poor data quality affect an AI-driven predictive maintenance model?
B
344
What technique can help manage data quality by flagging issues like incomplete or duplicate entries?
B
345
Which method is effective for handling missing data when re-collection isn�t feasible?
A
346
What is a recommended metric for monitoring data quality?
B
347
Why is data auditing important in maintaining AI model quality?
B
348
What is a common cause of bias in AI models trained on historical data?
B
349
How does sampling bias typically occur in AI data collection?
B
350
What type of bias may human annotators introduce during data labeling?
B
351
Which technique is recommended to create a balanced dataset in the presence of underrepresented groups?
C
352
What is one purpose of bias auditing in AI/ML models?
C
353
What is one primary privacy risk associated with AI/ML data collection?
B
354
How does inadequate anonymization of data contribute to privacy risks?
B
355
Which technique is effective for protecting individual identities in a dataset?
B
356
What role does informed consent play in mitigating privacy risks in AI models?
B
357
What is a primary goal of strict data access controls in managing privacy risks?
B
358
What is one common cause of limited data diversity in AI/ML datasets?
B
359
How can limited data diversity affect model performance in healthcare applications?
B
360
Which technique is effective for artificially increasing data diversity when collection is challenging?
B
361
What is a recommended practice to maintain data diversity over time?
A
362
What real-world issue can arise in facial recognition systems due to limited data diversity?
A
363
What does data provenance in AI primarily refer to?
B
364
Which of the following is a common cause of data provenance risk?
B
365
How can inadequate documentation of data handling affect AI model reliability?
C
366
What is one effective technique for ensuring transparency in data provenance?
B
367
Why might using alternative data sources increase data provenance risks?
B
368
What does data integrity risk primarily involve in the context of AI/ML models?
B
369
What is a key cause of change management issues in AI/ML models?
B
370
Which method helps mitigate data integrity risks by tracking changes in data structure and content?
B
371
How can version control improve change management in AI/ML models?
B
372
Why is change impact analysis essential in managing data change risks?
B
373
What is a primary cause of data leakage risk in AI/ML systems?
B
374
How does poor data anonymization contribute to data leakage risks?
B
375
Which technique is recommended for secure data transfer to mitigate data leakage risks?
B
376
What is the role of encryption in managing data leakage risks?
B
377
Why is role-based access control (RBAC) important in preventing data leakage?
A
378
What is a primary cause of regulatory non-compliance in AI/ML systems?
B
379
How can using third-party data sources contribute to regulatory non-compliance risks?
C
380
Which of the following is a recommended approach to mitigate regulatory non-compliance?
B
381
What role does a Data Stewardship and Governance Committee play in regulatory compliance?
C
382
Why are consent management systems important for regulatory compliance?
B
383
What is a common cause of algorithmic bias in AI models?
B
384
How does training on historical data contribute to algorithmic bias?
B
385
What is a recommended approach to reduce algorithmic bias during model development?
B
386
Which technique can improve fairness by rebalancing underrepresented groups in the training data?
D
387
Why is it important to implement fairness constraints during model training?
B
388
What is a primary cause of opacity in complex AI/ML models?
B
389
Why is opacity in AI models a regulatory concern?
B
390
Which technique is recommended to improve explainability in complex AI models?
B
391
What is an effective approach to address opacity risk in high-stakes applications?
A
392
How does layer-wise relevance propagation (LRP) contribute to model transparency?
B
393
What is a common cause of inappropriate model selection in AI/ML projects?
A
394
How can misunderstanding the problem domain contribute to inappropriate model selection?
B
395
What is one effective way to evaluate model suitability before final selection?
A
396
Which technique can help identify the most appropriate model for complex problems?
A
397
Why is consulting with domain experts recommended during model selection?
B
398
What is a primary cause of data-driven overfitting in AI models?
B
399
How does overfitting affect model performance on unseen data?
B
400
What role does regularization play in reducing overfitting?
B
401
Which technique is effective for ensuring a model generalizes well to new data?
B
402
What is one purpose of using early stopping during model training?
B
403
What is overfitting in AI/ML models?
B
404
Which technique helps prevent overfitting by reducing model complexity?
D
405
How does underfitting differ from overfitting in AI models?
B
406
Which of the following is an indicator of overfitting in model training?
C
407
What is one approach to balance overfitting and underfitting in model selection?
B
408
What is a primary risk of using biased or inappropriate evaluation metrics in AI/ML models?
B
409
Why is accuracy often a poor metric in fraud detection models?
B
410
What is a recommended approach to address biased evaluation metrics in imbalanced datasets?
B
411
Which metric is most suitable for models in imbalanced datasets?
A
412
What is the purpose of using a multi-metric evaluation approach?
B
413
What does hyperparameter sensitivity refer to in AI/ML models?
B
414
How does manual hyperparameter tuning contribute to sensitivity risks?
B
415
Which method is recommended to improve hyperparameter selection systematically?
C
416
What is the benefit of cross-validation in hyperparameter tuning?
C
417
Why is early stopping useful in iterative training models to address hyperparameter sensitivity?
B
418
What is one primary risk of neglecting fairness testing in AI models?
B
419
How can fairness audits help address ethical risks in AI models?
B
420
What role does an ethical checklist play in model development?
B
421
Which fairness metric could be used to test whether an AI model is equitable across different groups?
A
422
Why is it important to apply bias mitigation techniques in AI models?
B
423
What is a primary cause of algorithmic bias in AI systems?
B
424
How can biased training data impact AI model outputs?
B
425
Which approach is effective for auditing an AI model for fairness?
A
426
What is one method for reducing algorithmic bias during model training?
B
427
What role do fairness constraints play in AI model training?
B
428
What is a primary cause of transparency and accountability risks in complex AI models?
B
429
How can lack of model documentation contribute to accountability risks?
C
430
What tool can help make complex AI models more interpretable and transparent to users?
B
431
Why are transparency guidelines important in AI governance?
A
432
What is the purpose of using surrogate models in highly complex AI systems?
B
433
What is an example of an ethical dilemma in AI model development?
A
434
Why do ethical dilemmas in AI often involve trade-offs between competing principles?
B
435
Which approach is recommended for resolving ethical trade-offs in AI?
B
436
What role does multi-objective optimization play in addressing ethical dilemmas in AI?
B
437
How does stakeholder engagement help in managing ethical trade-offs?
B
438
What risk to human autonomy can arise from fully automated decision-making in AI systems?
B
439
How does over-reliance on AI in decision-making affect human autonomy?
B
440
What is the purpose of incorporating Human-in-the-Loop (HITL) mechanisms in AI-driven decision-making?
B
441
Why are appeals and recourse processes important in AI decision-making?
B
442
How can transparency in decision rights support human autonomy in AI systems?
B
443
What is a significant safety risk associated with deploying AI/ML models in autonomous systems?
B
444
How can insufficient real-world testing impact safety in healthcare AI applications?
B
445
What approach is recommended to minimize risks to safety and well-being in AI deployments?
B
446
Why is continuous monitoring important for AI models deployed in high-stakes scenarios?
B
447
In which industry would an AI model�s failure to accurately detect obstacles pose a serious safety risk?
C
448
What is model drift in AI/ML systems?
B
449
Which factor commonly leads to model drift?
B
450
What is a common mitigation technique for managing model drift?
B
451
Why is periodic retraining important in preventing model degradation?
B
452
How can real-time monitoring help manage model drift?
B
453
What is a primary operational risk when scaling AI models across multiple environments?
B
454
How can scalability issues impact AI model deployment in large organizations?
B
455
What is a recommended approach to ensure scalability and reliability of AI models in diverse environments?
B
456
Why is infrastructure flexibility important for operational scalability of AI models?
B
457
Which operational risk arises when a model fails to handle increased data volume as it scales?
C
458
What is a primary security risk when deploying AI/ML models?
B
459
How does data poisoning affect the security of AI models?
B
460
Which technique can help mitigate risks associated with model inversion attacks?
A
461
Why is encryption essential for securing sensitive data in AI models?
B
462
What role does access control play in managing privacy risks for AI/ML models?
A
463
What is a primary risk of deploying AI models without sufficient user training?
B
464
Why might non-technical users misinterpret AI model outputs?
B
465
What technique can help users correctly interpret AI model outputs in high-stakes scenarios?
B
466
Which mitigation technique involves providing training sessions to ensure users understand AI model limitations?
B
467
Why is clear documentation of AI model outputs essential?
B
468
What is the primary objective of model risk management in AI/ML systems?
B
469
Which factor commonly causes model risk in deployed AI/ML systems?
B
470
What is a recommended practice to mitigate model risk through governance?
B
471
How do regular performance audits contribute to model risk management?
B
472
Why is assigning defined roles and responsibilities important in managing model risk?
A
473
What is bias accumulation in AI/ML models?
B
474
How can reliance on historical data contribute to bias accumulation?
A
475
Which technique is recommended for detecting bias accumulation in AI models?
B
476
What is a common mitigation technique for addressing bias accumulation post-deployment?
B
477
Why are fairness constraints essential in preventing bias accumulation?
B
478
What is a primary risk associated with documentation gaps in AI models?
C
479
How can incomplete documentation affect model maintenance over time?
B
480
What is one effective approach to mitigate documentation gaps?
B
481
Why is documentation important for model risk management in regulated industries?
B
482
What impact do documentation gaps have on model reproducibility?
B
483
What is a primary risk of failing to update an AI model with new data?
B
484
How does model drift affect the performance of AI models over time?
B
485
What is one recommended approach to prevent model performance degradation due to outdated data?
B
486
Why is it critical for AI models in dynamic environments to be updated with fresh data?
B
487
What impact does failure to update for new data have on high-stakes applications, like healthcare or finance?
B
488
What is a primary cause of AI model obsolescence?
B
489
How does obsolescence impact AI models in dynamic environments?
C
490
What is a recommended approach to mitigate the risk of obsolescence in AI models?
B
491
Why is model monitoring important in managing the risk of obsolescence?
B
492
What role does flexibility in model design play in preventing obsolescence?
A
493
What is a primary risk associated with legacy AI models?
B
494
How can legacy models impact an organization�s ability to innovate?
B
495
Which of the following is an effective approach to managing legacy model risks?
B
496
Why is technical debt often associated with legacy AI models?
B
497
What is one reason why legacy models may pose security risks?
B
