Machine Learning Basics
Which of the following describes explainable AI (XAI)?
- APublishing the number of parameters and layers of a model to inform users of its size and structure
- BAdding training data or enlarging the model to make prediction accuracy as high as possible
- CTechniques and ideas for showing, in a form people can understand, why a model made a given prediction and which features influenced it
Show answer and explanation
Answer: C. Techniques and ideas for showing, in a form people can understand, why a model made a given prediction and which features influenced it
Complex models such as deep learning models tend to become black boxes: even when they perform well, it is hard to see why they made a given prediction. XAI is the set of techniques and ideas that let people check the reasons for predictions, using methods that compute each feature's contribution to a prediction (such as SHAP and LIME) or heat maps that show which part of an image influenced the decision. It helps in fields that require decisions to be explained, such as medicine and lending, and in finding errors and biases in models.
- A. The number of parameters and the structure only describe the model's size; they do not explain individual predictions.
- B. Improving performance and being able to explain the reasons for predictions are separate issues.
- C. Correct. It shows the reasons for predictions in a form people can understand.