Updating Beliefs with New Information
You'll learn: How conditioning changes our probability assessments when we observe new evidence—the mathematical foundation of learning from data.
The Core Idea
When you first assess a situation, you start with an initial probability based on what you know. But when new information arrives, a rational person updates their belief to reflect that evidence. Conditional probability is the mathematical language for this updating process.
Think of it like a detective story. Before finding any clues, each suspect might seem equally likely. But each new piece of evidence (a fingerprint, an alibi, a witness statement) should change your assessment of who's guilty. You're not starting from scratch—you're conditioning your original belief on the new facts.
How It Works
Suppose you believe there's a 30% chance of rain today (your prior belief). Then you look outside and see dark clouds gathering (new evidence). The probability of rain given those clouds is much higher—maybe 70%. The formula P(Rain | Dark Clouds) captures this updated belief.
Every time you calculate a conditional probability, you're mathematically updating a belief.
This is why conditioning is fundamental to data science: every dataset is "new information," and our job is to update beliefs (hypotheses, predictions) accordingly. When you train a model on data, you're essentially updating the probability of different outcomes given what you've observed.
The more evidence you gather, the more refined your probabilities become. This isn't guessing—it's systematic belief revision based on logic and evidence.
Key Takeaway: Conditional probability isn't just a calculation—it's the mathematical model of learning itself, showing how rational beliefs must change when new evidence appears.