The Causal Revolution
Causality has become mathematical. Thus, the causal revolution, which shows us how to think rationally about cause and effect.
Mind Over Data
Science thrives upon symbols. There is a powerful, overlooked science that teaches us to distinguish fact from fiction. It’s called Causal Inference.
Causal inference is a mental model.
The question Why?
All of human civilization and progress is based upon our ability to seek cause and effect by asking why? The purpose is to extract causal knowledge.
- How effective is a disease treatment?
- Should I quit my job?
- Did the new tax law cause the sales increase or was it advertising?
All of these words: Preventing, cause, policy, should I? Words such as these eek a causal relationship. Science has neglected efforts to understand causal relationships. Historically. Now it’s time for causal inference. This is the purpose of The Book of Why.
AI, big data, and deep learning provides adventurous paths forward into the new millenium.
There is a gap between the vocabulary of causal language and the accepted means scientists use to discover cause. Causal inference closes this gap?
It’s based on probability theory, which is exciting if you’re read the Drunkard’s Walk.
Correlation is not causation!
This is science’s way of avoiding investigation of a kind of causality. Modern statistics also avoids questions of causality. A class of science attributes causal questions unscientific. Causal vocabulary was virtually prohibited in science for over 100 years.
Statistics preaches that correlation is NOT causation but fails to explain what causation IS. So, correlation is not something that we can’t define anyway.
To summarize statistics – it will only suggest how to things are correlated based on data, but never infer or investigate causation. Statistics organizations and summarizes data but does not interpret that data. Interpreting data is considered unscientific.
Data science is seen as the solution to all of our problems – the data economy. But it’s weak. For example, data can show you that people who took a certain medication recovered faster from an illness than people who did not take the medication. But it cannot tell you why this is the case.
Perhaps the medication caused the recovery. Maybe something else entirely caused the recovery in the group that took the medication. Causal inference, therefore, is needed to interpret data and assign cause. Statistics will never do that.
Smoking causes health problems?
As recently as 20 years ago, scientists could not answer this question, even though answers were desperately needed. But a movement happened – the Causal Revolution, which is a calculus of causation!
This is the point of the Book of Why by Judea Pearl.