Which of the following is NOT a method of explainability in AI?

The correct answer highlights that customer reviews are not a method of explainability in AI. Explainability in AI refers to techniques and tools that help interpret and understand how an AI model makes decisions. Heatmaps, saliency maps, and SHAP values are all methodologies designed specifically to provide insights into the model's behavior.

Heatmaps and saliency maps visually represent the areas of input data that most influence a model's predictions. For instance, they can highlight which parts of an image were considered most important by a neural network during classification tasks. SHAP (SHapley Additive exPlanations) values quantify the contribution of each feature to a particular prediction, providing a clear numerical justification for the model’s outputs.

In contrast, customer reviews do not serve this purpose. While they can offer feedback about user experience with an AI system, they do not provide a systematic or analytical approach to understanding the decisions or processes of the AI itself, making them unrelated to explainability in the context of AI.