1
From Conditional Probability to Bayesian Decision-Making: Mathematical Tools for Causal Reasoning
MATH1003SA-PEP-CNLesson 2
00:00
Given Evidence (B)Trace the Cause (A|B)Causal Inference Black Box
Imagine you are a digital archaeologist. When you encounter a corrupted communication code (outcome $B$), your task is to infer the original instruction sent by the source (cause $A$). This logic of reasoning backward from 'effect' to 'cause' lies at the heart of how modern artificial intelligence handles uncertainty.

Starting from the definition of conditional probability $P(B|A)$, we can not only calculate the evolution of sequential events but also decompose global complexity into weighted sums of local conditions throughthe Law of Total Probabilityinto weighted sums of local conditions. AndBayes' Theoremis the crown of this theory—it allows us to continuously revise our prior knowledge (prior) based on new information (posterior), enabling dynamic cognitive evolution.

The Logical Three-Step Leap in Probability Theory

Step One: Local Dependence (Multiplication Rule)
When the occurrence of event $B$ depends on $A$, their joint probability is no longer a simple product, but $P(AB) = P(A)P(B|A)$. This is especially critical in sampling without replacement.

Step Two: Structural Decomposition (Law of Total Probability)
For complex macro-level events $B$, we project them onto different backgrounds $A_i$. The Law of Total Probability $P(B) = \sum P(A_i)P(B|A_i)$ tells us: the overall probability equals the expected value of local conditional probabilities.

Step Three: Causal Reverse Inference (Bayes' Theorem)
This is the formula of wisdom. It revises the 'prior probability' $P(A_i)$ (experience before the experiment) through the 'likelihood' $P(B|A_i)$ into the 'posterior probability' $P(A_i|B)$.

The Law of Total Probability is predictive—reasoning forward from cause to effect—while Bayes' Theorem is decision-making—reasoning backward from effect to cause. Together, they form the mathematical foundation of modern risk management and medical diagnosis.
$$P(A_i | B) = \frac{P(A_i)P(B | A_i)}{\sum_{k=1}^n P(A_k)P(B | A_k)}$$