Which of the following is NOT a recommended approach to mitigating bias in CPMAI?

• A. Data diversification and reweighting, algorithmic fairness techniques, and ongoing auditing.
• B. Continuous monitoring and auditing after deployment.
• C. Regularly updating data and models.
• D. Relying on a single data source without diversification.

Answer: (D)

Mitigating bias in CPMAI relies on broad data coverage and ongoing governance. Using diverse data sources, applying reweighting to balance underrepresented groups, employing algorithmic fairness techniques, and conducting continuous auditing and monitoring after deployment are all established practices. Regularly updating data and models helps address data drift and keeps fairness considerations current. Relying on a single data source without diversification, on the other hand, tends to embed the biases present in that narrow source, limits generalization to different contexts, and can lead to biased outcomes when the model encounters new or underrepresented situations. This approach undermines fairness and reliability, so it isn’t a recommended practice.