WIT - Learn

Model probing for understandable, reliable, and fair machine learning

Learn how to explore feature sensitivity, compare model performance, and stress-test hypotheticals.

Discover the basics

Learn about different spaces and how they are used in the What-If Tool: tabs, workspaces, modules and playgrounds.

Explore how the What-If Tool can help us learn about a model and dataset through a concrete example.

Learn how to edit a datapoint in the What-If Tool.

Learn the steps to find a counterfactual to a datapoint in the What-If Tool.

Learn how to customize visualizations from features in a loaded dataset using the Datapoints visualization.

Learn how to view partial dependence plots and what value they provide.

Explore the functionality of the Features Overview dashboard in the What-If Tool.

Set up the What-If Tool inside of TensorBoard.

Set up the What-If Tool inside of notebook environments.

Learn to use the What-If Tool with arbitrary python functions, and how to include feature attributions in the analysis.

A brief overview of performance metrics for classification-based models.

Understand the relationship between the ratio of the costs of false positives and false negatives and the optimization controls.

Learn how to add non-input features into the What-If Tool to analyze subgroups.

Explore datapoints in Features Overview to identify sources of bias.

Learn how to analyze feature attributions in the What-If Tool.

Qwiklabs Quest: Explore Machine Learning Models with Explainable AI.

Introductory level, hands-on practice with Explainable AI from Google Cloud.

Using the ‘What-If Tool’ to investigate Machine Learning models.

A community-contributed run-through of the features of the What-If Tool.

Introducing the What-If Tool

Get familiar with the What-If Tool with a three-part video series.

What if AI model understanding were easy?

See analytics-for-AI in action as Cassie Kozyrkov walks us through the What-If Tool.