SCENARIO V Flashcards
Please use the following scenario to answer the next FOUR questions.
Moxy Inc., headquartered in New York, is a significant player in the global medical device market. In a strategic
move, the company’s CEO declared AI governance and responsible AI as top priorities for the forthcoming year.
To spearhead this initiative, Moira, the chief privacy officer of Moxy Inc., is appointed to lead the AI governance
strategy. Recognizing the multidisciplinary nature of this task, Moira assembles an AI governance committee
comprising experts from various departments, with each bringing a wealth of knowledge from technical, legal
and ethical fields.
Under Moira’s guidance, the committee aims to address key concerns at the intersection of AI development
and privacy. A particular point of contention is the principle of data minimization — a privacy cornerstone —
against the backdrop of AI’s insatiable data needs. This inherent tension underlines the challenges in
maintaining privacy without stifling innovation. Beth, Moxy’s chief data scientist and Moira’s choice to lead the AI governance committee, embarks on defining
the pillars of responsible AI. These include ensuring accuracy, accountability, fairness, human-centric design
and transparency, among others. The overarching goal is to foster an environment where AI systems are not
only transparent, reducing biases, but also understandable to the end users, elucidating the “how” and “why”
behind AI decisions.
A practical application of these principles is seen in a project undertaken by the research and development
department. The team is developing an algorithm to offer personalized health recommendations, augmenting
its flagship medical device. This project exemplifies the challenges in creating high-quality, trustworthy AI
solutions, especially in a domain as sensitive as health care.
During the inaugural meeting of the AI governance committee, Beth faces resistance. The developers and data
scientists express satisfaction with their existing testing and validation methods, skeptical of the need for
change. Beth is now tasked with navigating these challenges, advocating for the adoption of new practices that
align with responsible AI governance.
- Which privacy-enhancing technology would be the most effective for Moxy’s developers to use in training
their datasets to understand group patterns without compromising individual privacy?
A. Differential privacy.
B. Federated learning.
C. Zero-knowledge proofs.
D. Homomorphic encryption.
- The correct answer is A. Differential privacy prevents the disclosure of individual information by introducing
statistical noise to the dataset, allowing the analysis of group patterns while preserving individual privacy.
Body of Knowledge Domain VI, Competency E
- To assess the AI system’s trustworthiness, which validation approach should Beth recommend to the
development team?
A. NIST AI RMF.
B. Model cards.
C. OECD tools and metrics.
D. Homomorphic encryption.
- The correct answer is B. Model cards serve as an effective validation tool for trustworthy AI by documenting
model performance across various conditions, including edge cases and unseen data, to facilitate a
comprehensive evaluation of the AI system’s reliability.
Body of Knowledge Domain VI, Competency E
- To standardize explainability and transparency, Beth encourages developers to complete model cards.
Which key aspect should these model cards emphasize?
A. Big data.
B. Model details.
C. Cluster analysis.
D. Linear regression model.
- The correct answer is B. The primary component of a model card is model details, which encompass the
model’s purpose, design, performance metrics and ethical considerations, providing a comprehensive
overview of the model’s functionality and application.
Body of Knowledge Domain VI, Competency E
