Introduction To Causal Machine Learning Part 1: Why Causal?
In the introduction to the series, I must explain the limitations of Causal ML as well as a basic applied framework. Causal ML is not perfect but it is ready for prime time. Let's get started.
Causal Machine Learning is a difficult line of study which stands above all the other difficult lines of study in our field. I have spent over 5 years learning Causal ML and deploying causal models in production. I am still no expert, but I do not believe our field has experts in Causal ML yet.
Let’s begin with some important distinctions. Causal inference has frameworks and there are experts in casual frameworks. There are experts in experimental design and methodologies. In academia, causal is a living breathing field with a lengthy history.
Causal reasoning goes back to the ancient philosophers. There is still debate around the nature of reality, causality, and our perception of both. Physics to metaphysics have attempted to quantify reality and the laws which govern everything around us.
You must see both paths to understand the fundamentals of experimental design and causal frameworks. The philosophical side of science always strikes me as odd. We have no empirical definition of reality and that impacts the core of all experimentation.
It creates a footnote to everything I will describe going forward. All these relationships we discover exist without a defined framework or common point of reference. We are still naïve to the grand design and unaware of the unifying structure of reality. This has both macro and micro impacts.