Ask HN: How to think machine learning like a scientist?

I read a lot of deep learning papers but don’t get the process. What are some example papers to learn to think like a scientist? Are there any how-to guides to think like a scientist in the context of machine learning?