Machine learning for social scientists

Given the knowledge and practice as a social scientist (e.g., the understanding of regression methods), what would be the most useful information about Machine learning?

  1. The goal is different. yhat or beta?
  2. Because of the different focus, there are different emphases. For instance, in ML, we focus more on preventing overfitting by employing dataset split, Cross validation, Regularization, and so on.
  3. ML is omnivorous. It doesn’t have to be a ‘statistical’ model. A good example would be the Genetic algorithm (see CGP Grey‘s ML video).

Athey2019machine is a good review from the perspective of economics.