AUSTIN, TEXAS - Before continuing with this article, be sure to first take a look at the previous one, Causation: The most misunderstood concept in AI, which covered the following:

- The difference between ‘useless’ correlations in AI versus ‘useful’ correlations/associations, as well as the three factors that determine a useful correlation.

- The different requirements for predictive and prescriptive models.

- The factors that determine causality at a simple level.

This article will analyze a model for increasing levels of knowledge and understanding, and then describe a specific example.

Seeing, doing, and imagining: Three levels of proof for prediction and prescription

This paradigm is attributed to Judea Pearl¹ in his work The Book of Why.² Pearl described a ‘ladder of causation,’ where increasing evidence provides insight into one’s understanding of why something occurred and with what level of deterministic mechanism behind it.

Level 1: Seeing

This level mainly deals with patterns of association, as well as the human activities of seeing and observing.

Questions:

What if I see _____?

How are the variables related?

How would seeing X change my belief in Y?

Examples:

What does a symptom tell me about a disease?

What does a survey tell me about election results?

Level 2: Doing

This level is about intervention, as well as the human activities of doing something and seeing what happens next.

Questions:

What if I