A data scientist creates a new feature by multiplying two existing features together. What is this technique called and when is it useful?