Your Model Is Accurate and Wrong: The Case for Causal Thinking in Data Science

Predictive accuracy and causal validity are different properties. Conflating them is one of the most common and consequential errors in applied data science.