AWS Certified AI Practitioner Practice Exam

What should an AI practitioner include in a report to provide transparency about an ML model?

1. Code for model training

2. Partial dependence plots (PDPs)

3. Sample data for training

4. Model convergence tables

The correct answer is: Partial dependence plots (PDPs)

In providing transparency about a machine learning model, incorporating partial dependence plots (PDPs) is essential. PDPs illustrate how the predicted outcome of a model changes with varying values of specific features while averaging the effects of other features. This visualization helps stakeholders understand the relationship between input variables and the model's predictions. Including PDPs enhances interpretability, allowing both technical and non-technical audiences to grasp how individual features influence the model's predictions. Transparency is vital for building trust in AI systems, especially when the decisions made by the model can have significant consequences. By presenting this information, practitioners can facilitate discussions about the model's behavior and validate that it operates as expected within its intended domain. While the other options may contribute to a deeper understanding of the model, they do not directly address the aspect of transparency as effectively as PDPs. For instance, merely providing code may not be interpretable for non-technical stakeholders, and sample data might not illustrate feature importance. Similarly, model convergence tables focus on the training process rather than offering insights into how features affect predictions. Thus, PDPs stand out as a vital tool for transparency in model reporting.