NCA-GENL Question 210
Select 3A company is deploying an AI-based language model for customer service automation. During development, they discover that the model generates biased responses for specific demographics. Which actions best align with the ethical principles of trustworthy AI to address this issue?
- A
Conduct a bias audit and retrain the model using a more representative dataset.
- B
Deploy the model as-is, but include a disclaimer about potential biases in responses.
- C
Integrate fairness checks into the development pipeline to monitor for future biases.
- D
Prioritize model performance over fairness, as long as it meets accuracy benchmarks.
- E
Engage a diverse group of stakeholders to provide feedback on the model's outputs.
Show answer and explanation
Correct answers: A, C, E
Explanation
The principles of trustworthy AI emphasize fairness, accountability, and transparency. Addressing biases through audits, fairness checks, and stakeholder engagement ensures that the AI system operates ethically and inclusively. Deploying a biased model or ignoring fairness for performance violates these principles.
- A. Correct.
This option is correct because conducting a bias audit and retraining the model helps ensure fairness and addresses the issue of biased responses.
- B. Incorrect.
This option is incorrect because deploying the model without addressing biases violates the principles of accountability and fairness, even with a disclaimer.
- C. Correct.
This option is correct as integrating fairness checks is a proactive approach that aligns with the principles of transparency and fairness in trustworthy AI.
- D. Incorrect.
This option is incorrect because prioritizing performance over fairness undermines the ethical principles of equity and inclusivity in AI.
- E. Correct.
This option is correct because involving diverse stakeholders ensures accountability and helps identify biases that may not be apparent to the development team.