liveBook · Manning

Consistency rule

As explained in Causal AI

The consistency rule is a causal inference principle stating that a hypothetical outcome equals the observed outcome when the hypothetical condition matches the factual condition.

The consistency rule links a counterfactual intervention to the factual world. When the hypothetical condition matches the condition that actually occurred, the hypothetical outcome equals the observed outcome. Formally, P(Yₓ₌ₓ | X = x, Z = z) = P(Y | X = x, Z = z). section 8.2.5 section 8.2.4

Apply the rule by checking whether the factual and hypothetical conditions overlap without conflict:

  • Write the factual conditions and the hypothetical intervention separately. section 8.2.4
  • If the intervention sets a variable to the value already observed, remove that hypothetical condition from the notation. section 8.2.5 section 8.2.4
  • Retain conflicting interventions as genuine counterfactual questions; consistency does not collapse them. section 8.2.5 section 8.2.4

In-game purchases

Suppose a player actually has low side-quest engagement and $75 in-game purchases. The query P(Iᴱ⁼ˡᵒʷ | E = “low”, I = 75) asks about the same engagement level that is factually observed, so consistency reduces it to P(I | E = “low”, I = 75): the outcome is $75. By contrast, asking about high engagement remains a conflicting counterfactual. section 8.2.5

Figure 8.4 Elements of a conditional counterfactual hypothetical formalized in counterfactual notation
figure

Use the notation diagram to distinguish factual conditions, hypothetical conditions, and hypothetical outcomes before applying consistency. section 8.2.4

How Causal AI covers it

How deep it goes
Derives the rule formally and works through matching and conflicting intervention examples.
Examples
Uses side-quest engagement and in-game purchases, including an observed purchase amount of $75.
What it assumes
Assumes familiarity with conditional probability and introduces formal counterfactual notation through natural-language queries.
What you can do after
Evaluate counterfactual notation and simplify queries whose hypothetical conditions match observed conditions.
How it connects
Connects counterfactual reasoning, interventions, factual conditions, and hypothetical conditions.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Causal AI. The section links point to the parts of the book it draws on.