ITE302c C3

b

How do AI and other data-driven technologies use probability? 1 point a. By determining the objective likelihood of some event happening b. By providing a model of belief about the likelihood of some event happening c. By guaranteeing that some event will occur with 100% likelihood d. By estimating the likelihood of some event happening without input data

a

How might a non-stochastic decision tree be considered a black box algorithm? 1 point a. If the model is far too complex to be readily understood by human beings. b. If the model is expressed in an annotated visual form. c. If the algorithm is contained within an embedded system. d. If the algorithm serves as part of a data recorder.

a

In which of the following could STRIDE or VAST classifications be applied? 1 point a. Threat modeling or analysis tools b. Black box mitigation methods c. Bias mitigation techniques d. Network optimization tools

a

Question 1 What does the acronym RACI stand for? 1 point a. Responsible, Accountable, Consulted, Informed b. Redesignable, Accessible, Collected, Informative c. Relatable, Agreeable, Commendable, Irascible d. Responsible, Accessible, Consulted, Improvised

c

Question 1 What does the acronym SIEM stand for in a cybersecurity context? 1 point a. Social, interpersonal, empathic, mental b. Selective information extrapolation methods c. Security information and event management d. Serial input to externalized modulation

abd

Question 1 Which of the following are tasks in the overall risk management process? (Select three.) 1 point a. Identification b. Mitigation c. Deployment d. Analysis e. Elimination

c

Question 1 Which of the following best describes a marketing persona? 1 point a. A character that encapsulates a brand's personality b. An individual who assists with the activity of marketing c. A description of a type of character based upon demographics, habits, and interests d. The emotional impression that a customer has about a product or category of goods

d

Question 1 Which of the following is an explainability risk of self-learning models? 1 point a. Self-learning models are black boxes whose decisions cannot be understood. b. Self-learning models are unable to generate decisions as output. c. Self-learning models can corrupt data or otherwise make it unreadable. d. Self-learning models may introduce new data you're unaware of.

a

Question 1 Which of the following types of bias emerges when the same data is sampled over and over again, limiting a model's perspective? 1 point a. Reinforcement bias b. Confirmation bias c. Temporal bias d. Implicit bias

cd

Question 10 Which of the following are likely examples of a leverageable third party data-driven or data-related product? (Select two.) 1 point a. Management oversight b. Product design lifecycle c. Compiled models d. Data annotation and augmentation tools

ac

Question 10 Which of the following describe a process of cross-correlation of data? (Select two.) 1 point a. Comparing multiple data points or sets b. Comparing people of different demographic groups c. Discovering new information via inference d. Analysing an individual's emotional stimulus and response patterns

c

Question 10 Which of the following metrics is used to evaluate a linear regression machine learning model? 1 point a. Goodhart's Law b. Receiver operating characteristic (ROC) c. Cost function d. Accuracy

d

Question 2 Which of the following best describes the use case for zero-knowledge protocols (ZKPs)? 1 point a. Being made aware of when the battery in a product is past its zero point b. Choosing what to do when you have zero knowledge of a subject c. Handling divide-by-zero errors in computing software d. Proving that something exists, but not the specifics

a

Question 2 Which of the following describes extrajudicial judgment? 1 point a. A ruling or scoring made outside of a court. b. Unusual interpretation of law outside of its typical meaning. c. Applying law from one jurisdiction within another. d. Transferring defendants from one jurisdiction to another for trial.

b

Question 2 Which of the following is a transparency risk of closed source software? 1 point a. Closed source software cannot be reviewed by the organization that developed it. b. Closed source software may not be accessible to independent auditors. c. Closed source software includes obfuscated code that cannot be read by a human. d. Closed source software cannot be shared with business partners.

ac

Question 2 Which of the following sources does a SIEM system typically pull data from? (Select two.) 1 point a. System event logs b. Personnel emails c. Network intrusion detection alerts d. Encrypted personal data

d

Question 2 Why is a model that has been overfitted to its training data a source of fairness risk? 1 point a. Because the model includes too much noise. b. Because the model is too complex. c. Because the model has a temporal bias. d. Because the model won't generalize to the entire population.

a

Question 3 A disruptive user on a social media site is subjected to a shadow ban. Which of the following might be an indication of that shadow ban? 1 point a. A user logs in one day to find that they are able to post, but receive no direct responses. b. A user logs in one day to find that their posting privileges have been revoked. c. A user logs in one day to find a message from a site administrator saying that the user has been banned. d. A user logs in one day to find that they have been asked to discontinue their disruptive behavior, or they will be banned.

cd

Question 3 Which of the following elements constitute PII? (Select two.) 1 point a. Cookies and tokens b. Device IDs c. Passport numbers d. Telephone numbers

ac

Question 3 Which of the following elements might be considered moderately affected in an Impact Level 2 Algorithmic Impact Assessment? (Select two.) 1 point a. The health and well-being of individuals or communities. b. A brief delay in the conveyance of a commercial transaction. c. The rights of individuals or communities. d. A loss of life resulting from specific circumstances.

d

Question 3 Which type of edge case is caused by data outside the normal distribution? 1 point a. Errors b. Overfitting c. Noise d. Outliers

a

Question 4 Which of the following describes automation bias? 1 point a. Bias of people towards overly trusting in automation. b. Bias against automation by people worried about their jobs. c. Bias enabled or perpetuated by machine learning. d. Bias that is subconscious, committed during a state of automata.

ac

Question 4 Which of the following elements should be communicated with end users when negotiating ethical access to a user's data? (Select two.) 1 point a. The intention behind a certain activity or request b. An offer of money or other incentives to relinquish user's rights c. A request for consent to perform an action d. The monetary value of that data

b

Question 4 Which of the following is a function of exploratory data analysis (EDA)? 1 point a. To evaluate the methods used to collect the data. b. To evaluate the quality of data before it is used to train a model. c. To evaluate the decisions made by the model after training on the data. d. To evaluate the organization's project management structure for the AI project.

a

Question 4 Which of the following is an effort to turn black box AI models into models whose decisions are easier to understand? 1 point a. Explainable AI b. Self-learning models c. Human-in-the-loop d. Neural networks

bd

Question 4 Which of the following statements are correct about a variable that is normally distributed? (Select two.) 1 point a. The tails of a normal distribution are denser than the center. b. The variable's distribution, when graphed, exhibits a symmetrical bell shape. c. Less than half of all measurements fall within one standard deviation of the mean. d. The mean, median, and mode of all measurements is the same, and all are located at the center of the distribution.

d

Question 4 Which team typically oversees penetration test operations and adherence to the rules of engagement (ROE)? 1 point a. Purple team b. Blue team c. Red team d. White team

b

Question 5 How can persona modeling be used to identify potential biases in a machine learning model? 1 point a. The persona may show that specific users are a source of bias. b. The persona may represent groups of people that could be susceptible to bias. c. The persona may identify noise and other edge cases that lead to bias. d. The persona may reveal overfitting issues in a model that result from bias.

a

Question 5 Which of the following describes the technique of homomorphic encryption? 1 point a. A method of performing operations on encrypted data. b. A method of swapping public and private keys between hosts. c. A method of mitigating biased algorithms through enhanced secrecy. d. A method of cryptography that hides data inside other data.

a

Question 5 You want to solicit feedback from users who are seeking explanations about your AI products and services. So, you decide to place a form on your marketing website that users can fill out. Which of the following is the best way to structure that form for the purpose of collecting useful feedback? 1 point a. Provide a large, unrestricted text box where users can enter their thoughts at length. b. Provide a series of yes or no questions with radio buttons where users select their respective answers. c. Provide a series of open questions with a single-line entry field where users can provide short answers. d. Provide a drop-down menu of feedback categories that users can select from, with a promise that you'll follow up via email.

a

Question 6 In a classification model that determines whether or not a customer qualifies for a coupon, a significantly lower percentage of males qualified than females. Which of the following types of discrimination does this outcome potentially represent? 1 point a. Disparate impact b. Disparate non-impact c. Disparate mistreatment d. Disparate treatment

c

Question 6 In the following scatter plot, the GrossIncome variable is plotted against the Revenue variable. What type of correlation does this plot suggest? a. There is a weak positive correlation between both variables. b. There is a strong negative correlation between both variables. c. There is a strong positive correlation between both variables. d. There is a weak negative correlation between both variables.

d

Question 6 Which of the following could Nmap, Wireshark, or Metasploit be applied to for ethical purposes? 1 point a. Business continuity planning b. Persona modeling c. Criminal exploitation d. Attack simulations

bc

Question 6 Which of the following describe aspects of secondary data usage that might increase ethical risks? (Select two.) 1 point a. Encryption of data makes it harder to process b. Data applied to secondary purposes may expose private aspects or diminish personal dignity c. Additional organizations and processors may now have access to the data d. Secondary data may be more expensive

c

Question 6 You're the recipient of some data collected by a third party that you'll use in training an AI model. As per your business agreement, the third-party discloser of the data requests that you explain how you plan to use the data. At what point in the process should you provide this explanation? 1 point a. At the point of evaluating the model. b. At the point of training the model. c. At the point of collecting the data. d. At the point of planning the AI system.

d

Question 7 What does the initialism CIA stand for in a data security context? 1 point a. Compliance, integrity, accountability b. Communicability, interpretation, accessibility c. Conformity, influence, adaptation d. Confidentiality, integrity, availability

ad

Question 7 Which of the following are potential ethical benefits of conducting pilot testing prior to the release of emerging technologies? (Select two.) 1 point a. It reveals any gaps in accountability or auditability. b. It certifies that a product's developers are legally accountable. c. It demonstrates that the system is functional as early as possible in order to satisfy senior management and investor stakeholders. d. It minimizes the impact of any ethical issues which may emerge.

ac

Question 7 Which of the following are potential training data inadequacies that you should communicate to your users in support of transparency? (Select two.) 1 point a. Any known bias in the sample data. b. The amount of time it took to collect the data. c. Any missing values in the dataset and how they were handled. d. The relatively large size of a dataset used in training.

a

Question 7 Which of the following describes the purpose of a STEEPV analysis? 1 point a. To perform a strategic analysis of how external environments impact business operations. b. To perform a strategic analysis of how internal office politics impact business operations. c. To perform a strategic analysis of user needs and behaviors. d. To perform a strategic analysis of how bias can manifest in AI products.

bc

Question 7 Which of the following provide a user with more meaningful choices regarding how their data is being used? (Select two.) 1 point a. The format that the data is stored in b. Granular or modular control over data usage c. Opt-out/opt-in mechanisms d. Whether the data is used by a person of a certain gender

d

Question 8 SSL, TLS, and SSH are all forms of... 1 point a. Computer forensics tools b. Media codecs c. Version control systems d. Encryption protocols

ab

Question 8 Which of the following are best practices for incorporating inclusive design in AI projects? (Select two.) 1 point a. Leverage customer input to reduce bias. b. Consider bias a spectrum. c. Keep machines and humans separate. d. Solve for many, extend to one.

ad

Question 8 Which of the following are important steps to document during an auditing process? (Select two.) 1 point a. Specify and document what precisely is being audited. b. Specify a purpose for the audit or an agenda to be covered. c. Provide significant advance notice before an audit. d. Document the audit procedures to be followed.

c

Question 8 Which of the following is a technique used by the Python tool Alibi to produce a subset of features that will usually result in the same model prediction? 1 point a. Local Interpretable Model-Agnostic Explanations b. Perturbation c. Anchor Explanations d. Shapley value

ad

Question 8 Which of the following refer to legal regulations or industry frameworks for protecting data? (Select two.) 1 point a. HIPAA b. CE mark c. ISO 9000 d. NIST Privacy Framework

a

Question 8 You have a dataset of customers that includes each customer's gender, location, and other personal attributes. The label you are trying to predict is how much sales revenue each customer is likely to generate for the business based on these attributes. What type of machine learning outcome is this problem suited for? 1 point a. Regression b. Clustering c. Classification d. Dimensionality reduction

d

Question 9 Which of the following Python tools is designed to integrate with Google's Cloud AI platform? 1 point a. SHAP b. LIME c. ELI5 d. What-If

bc

Question 9 Which of the following are common key performance indicators (KPIs) within dashboards of data-driven systems? (Select two.) 1 point a. Earnings before interest, taxes, depreciation, and amortization (EBITDA) b. Performance metrics, such as accuracy and cost functions c. Model training time d. Operating cashflow forecast

ab

Question 9 Which of the following are prerequisites for high-level digital security and trust mechanisms? (Select two.) 1 point a. Authentication b. Authorization c. Reputation d. Kindness

c

Question 9 Which of the following describes the AI Fairness 360 project? 1 point a. A checklist for machine learning practitioners to follow when training fair AI models. b. An AI product that represents a completely fair model for AI practitioners to use as a benchmark. c. An open source library that evaluates models for bias and provides mitigation tactics to reduce that bias. d. A global initiative that promotes fairness in AI through seminars, conferences, and other community-driven activities.

d

Question 9 You're training a model to classify whether or not a bridge is likely to collapse given several factors. You have a dataset of thousands of existing bridges and their attributes, where each bridge is labeled as having collapsed or not collapsed. Only a handful of bridges in the dataset are labeled as having collapsed—the rest are labeled as not collapsed. Given your goal of minimizing bridge collapse and the severe harm it can cause, which of the following metrics would be most useful for evaluating the model? 1 point a. Confusion matrix b. Accuracy c. Precision d. Recall

c

What does the acronym CSIRT stand for in a security context? 1 point a. Computer security immediate readiness tools b. Computer science, Internet, real-time interaction, technology c. Cybersecurity incident response team d. Conformance strategy, immutability, reliability, trustworthiness

b

What is the primary advantage of radioactive data tracing over past techniques that modify input in order to determine whether or not that input was used in training? 1 point a. Radioactive data tracing makes the modification perceptible to human beings. b. Radioactive data tracing doesn't impact a model's performance. c. Radioactive data tracing targets language-based input. d. Radioactive data tracing can be used to modify the label.

ac

Which of the following are likely benefits of having defined organizational policies? (Select two.) 1 point a. It supports consistency of behavior and unification of culture. b. It supports improved sales and marketing results. c. It supports accountability and makes it more explicit. d. It supports profitable research innovations and intellectual property.

ac

Which of the following are likely to be found within a penetration test rules of engagement (ROE)? (Select two.) 1 point a. What methods of attack are legitimate b. The appropriate length and complexity of employee passwords c. What systems may be targeted d. How long CCTV camera recordings should be maintained

cd

Which of the following are typically true when comparing the LIME tool with the SHAP tool? (Select two.) 1 point a. LIME supports fewer types of models than SHAP. b. LIME supports more programming languages than SHAP. c. LIME outputs results faster than SHAP. d. LIME is less accurate than SHAP.

bd

Which of the following benefits can be gained from establishing baseline system behaviors? (Select two.) 1 point a. Holding employees more accountable b. Tracking deviance from norms c. Optimizing costs of operation d. Restoring compromised functionality through a system rollback

b

Which of the following is not a valid risk response technique? 1 point a. Accept b. Ignore c. Transfer d. Avoid

ab

Which of the following risk analysis methods use words like "likely," "unlikely," and "rare" to describe the likelihood of risk, and words like "low," "medium," and "high" to describe the impact of risk? (Select two.) 1 point a. Qualitative analysis b. Semi-quantitative analysis c. Quantitative analysis d. Semi-qualitative analysis

b

Which of the following statements accurately describes variance? 1 point a. Variance measures the shape of the tails in a distribution relative to the center. b. Variance measures how far a data example is from the mean. c. Variance measures the error between predicted values and actual values. d. Variance measures how much a variable's distribution differs from a normal distribution.