Google Professional Machine Learning Engineer Question 472
Select 3Google Cloud PlatformYou are part of a team developing a machine learning model for loan approval using Google Cloud. During the evaluation phase, you discover that the model disproportionately denies loans to applicants from certain minority groups, even when they meet the required financial criteria. Which actions align with Google’s Responsible AI practices to address this issue?
- A
Implement fairness metrics to evaluate model performance across different demographic groups.
- B
Fine-tune the model solely based on the majority group to improve overall accuracy.
- C
Analyze the dataset for imbalances and augment it with additional representative data from underrepresented groups.
- D
Deploy the model as is and monitor for bias in real-time post-deployment.
- E
Set up a process to regularly retrain the model with updated data to account for demographic shifts over time.
Show answer and explanation
Correct answers: A, C, E
Explanation
Google’s Responsible AI practices emphasize fairness, accountability, and inclusiveness. Addressing bias requires proactive measures such as evaluating fairness metrics, ensuring balanced datasets, and maintaining the model’s fairness over time through retraining. Simply deploying a biased model or focusing on the majority group does not align with these principles.
- A. Correct.
Fairness metrics are essential to identify and quantify bias across different demographic groups, a core principle of Google’s Responsible AI practices.
- B. Incorrect.
Focusing only on the majority group will exacerbate existing biases and contradict Responsible AI principles.
- C. Correct.
Balancing the dataset by augmenting it with representative data can help mitigate bias and improve fairness in the model.
- D. Incorrect.
Deploying the model as is without addressing the bias upfront is not aligned with Responsible AI practices, as it could perpetuate harm.
- E. Correct.
Regularly retraining the model helps ensure it adapts to changing demographics and remains fair over time, a key aspect of Responsible AI.