NCA-GENM Question 209
Select 2A team is developing a generative AI model for medical image analysis. During the project, they realize that the dataset used for training has an overrepresentation of images from one demographic group. Which ethical principles of trustworthy AI should the team prioritize addressing in this scenario?
- A
Fairness and inclusivity
- B
Accountability and transparency
- C
Reliability and safety
- D
Privacy and data protection
- E
Human-centered values
Show answer and explanation
Correct answers: A, C
Explanation
In this scenario, fairness and inclusivity must be prioritized to ensure the model produces equitable outcomes across different demographic groups. Additionally, reliability and safety are crucial because a biased model could lead to unsafe or inaccurate conclusions in medical contexts, potentially harming patients. Addressing these principles helps ensure the model aligns with trustworthy AI standards.
- A. Correct.
Fairness and inclusivity are directly relevant because an overrepresentation of one demographic group in the dataset could lead to biased or unfair outcomes for underrepresented groups.
- B. Incorrect.
Accountability and transparency, while important in AI ethics, are not the primary concern in this scenario as the issue is related to dataset representativeness rather than the need for clear documentation or responsibility.
- C. Correct.
Reliability and safety are critical here because a biased model trained on an unbalanced dataset may produce unreliable results, especially for underrepresented groups, which could impact safety in medical decision-making.
- D. Incorrect.
Privacy and data protection are important in AI, but this scenario does not involve concerns about protecting individual data or ensuring privacy.
- E. Incorrect.
Human-centered values, while a broader ethical principle, do not directly address the specific issue of dataset bias and its impact on fairness or reliability.