1. The Core Idea

Bayes’ Theorem is a mathematical way to update your beliefs when you get new evidence.
It tells you how to reason from what you already know (your prior) to what’s newly observed (your evidence).

$$P(H\mid E)=\frac{P(E\mid H),P(H)}{P(E)}$$

Where:

• $P(H\mid E)$ = Probability that the hypothesis is true given the evidence (posterior).
• $P(E\mid H)$ = Probability of the evidence if the hypothesis were true (likelihood).
• $P(H)$ = Probability of the hypothesis before seeing the evidence (prior).
• $P(E)$ = Probability of the evidence occurring at all (normalizing constant).
  Expanded denominator (law of total probability): $$P(H\mid E)=\frac{P(E\mid H),P(H)}{P(E\mid H),P(H)+P(E\mid \neg H),P(\neg H)}$$

---

2. Odds Form (Likelihood Ratio)

You can express Bayes’ Theorem in terms of odds:

$$\frac{P(H\mid E)}{P(\neg H\mid E)}=\frac{P(H)}{P(\neg H)}\times \frac{P(E\mid H)}{P(E\mid \neg H)}$$

Where:

• The left side = posterior odds (belief after seeing evidence)
• The first ratio = prior odds (belief before seeing evidence)
• The second ratio = Bayes factor (likelihood ratio) — how strongly the evidence supports one side

---

3. Example: Medical Test

A disease affects 1 in 1,000 people ($P(H)=0.001$).
A test detects it correctly 99% of the time ($P(+\mid H)=0.99$)
but gives 1% false positives ($P(+\mid \neg H)=0.01$).

$$P(H\mid +)=\frac{0.99\times 0.001}{(0.99\times 0.001)+(0.01\times 0.999)}\approx 0.09$$

Even though the test is 99% accurate, a positive result means only a 9% chance you actually have the disease.
This is because the disease is rare — base rates matter.

---

4. Example: DNA Match (Prosecutor’s Fallacy)

A DNA test has a 1 in 1,000,000 chance of matching by accident.
In a city of 10 million, that means about 10 random matches could exist.
Prior odds someone is guilty: $1:9,999,999\approx 10^{-7}$
Likelihood ratio (match accuracy): $\frac{1}{10^{-6}}=10^6$

$$\text{Posterior odds}=10^{-7}\times 10^6=0.1$$ $$P(\text{guilt}\mid \text{match})=\frac{0.1}{1+0.1}\approx 0.09$$

So even a perfect-looking DNA match only raises guilt probability to 9% if the prior chance was tiny.

---

5. Common Fallacies Bayes Guards Against

a. Base-Rate Neglect

Ignoring how rare the event is.

“The test says positive, so it’s almost certain I have the disease.”
Fix: Always include the prior probability $P(H)$.

Rare events need extraordinary evidence.

b. Transposed Conditional

Confusing $P(E\mid H)$ with $P(H\mid E)$.

“If a terrorist, likely to have this profile — so anyone with this profile is likely a terrorist.”

Fix: Bayes distinguishes these clearly. You need both $P(H)$ and $P(E\mid \neg H)$.

c. Prosecutor’s Fallacy

Equating “chance of match if innocent” with “chance of innocence given a match.”

“DNA match probability is 1 in a million, therefore guilt is nearly certain.”

Fix: Use Bayes. Consider population size and prior odds.

d. Misuse of Improbability

People think a very unlikely event must have a special cause.
Fix: In a world full of possible outcomes, some improbable event will always happen.
Improbability ≠ evidence of design or conspiracy.

---

6. How to Use Bayes’ Theorem in Arguments

To Guard Against Fallacies

• Always ask: What’s the prior probability?
• Consider: How likely was this evidence anyway?
• Distinguish correlation from causation.
• Update beliefs gradually, unless evidence is overwhelming.

To Attack Weak Arguments

• Challenge missing priors (base-rate neglect).
• Point out when people swap $P(E\mid H)$ with $P(H\mid E)$.
• Question “extraordinary claims” that rely on weak priors.
• Highlight cherry-picked data that inflates $P(E\mid H)$ but ignores $P(E\mid \neg H)$.

---

7. Key Insight

Bayes’ theorem formalizes what good reasoning already does:

• Start with prior knowledge.
• Evaluate how strongly new evidence fits.
• Update beliefs proportionally.
  It’s a method of rational humility: never absolute certainty, only refined probability.

---

Quick Reference (Cheat Sheet)

---

Takeaway:

Bayesian thinking is a shield against overconfidence and fallacious reasoning.
It forces every claim to face the same question:
“How much should this evidence really change my mind?”

---

Excellent question — and one that gets to the heart of why Bayes’ Theorem matters far beyond math.

You don’t need numbers or equations to “use” Bayes’ reasoning — you just need to think like Bayes. The math simply formalizes what good thinkers and debaters already do intuitively.

Here’s how to apply Bayesian reasoning in ordinary conversation or argument — in clear, everyday language.

---

Steps to Critique Argument

1. Step One: Start With a Prior

Before reacting to new information, ask yourself:

“How likely was this claim before I heard this new evidence?”
This is your prior belief — your sense of how plausible something is given what you already know about the world.
Example:
Someone says, “My friend’s uncle cured his cancer with herbal tea.”

• Your prior knowledge: cancer is complex; spontaneous recovery is extremely rare; herbal cures usually don’t pass clinical tests.
• So the prior probability that this claim is true is very low.
  That doesn’t mean it’s impossible — it just sets your starting point.

---

2. Step Two: Ask How Likely the Evidence Is If the Claim Were True

Now you ask:

“If this claim were true, would this kind of evidence make sense?”
In the example:

• If herbal tea really cured cancer, you’d expect many independent reports, clinical trials, and medical studies — not one anecdote.
  So the evidence (a single story) doesn’t fit what you’d expect if the claim were actually true.

---

3. Step Three: Ask How Likely the Evidence Is If the Claim Were False

Next question:

“Would I also see this kind of evidence if the claim were false?”
In this case:

• Yes — it’s common for people to report coincidences or to mistake remission for a cure.
  That means the evidence (a story) is not special — it could easily appear even if the claim isn’t true.
  This is crucial: if the evidence fits both “true” and “false” worlds, it’s weak evidence.

---

4. Step Four: Update — But Only a Little

After weighing those two sides, ask:

“Given what I’ve learned, how much should I really change my mind?”
In most cases, the answer will be: “Not much.”
Anecdotes, rumors, or social media posts rarely justify major belief changes.
But a replicated scientific study or multiple independent witnesses might.

This is Bayesian updating without numbers — it’s proportional belief change.

5. Step Five: Keep Updating As New Evidence Comes In

You don’t have to decide once and for all.
Each new piece of credible evidence nudges you slightly in one direction or another.
That’s how science and good reasoning progress — not by sudden conversions, but by gradual refinement.

---

6. Everyday Uses (No Math Required)

---

7. In Arguments

When someone presents a claim:

1. Ask for priors:
   “Has this ever happened before? How often?”
   This stops people from assuming every event is 50/50.
2. Ask for alternatives:
   “What’s another reason we might be seeing this?”
   This forces them to think about $P(E\mid \neg H)$ — even without math.
3. Ask how expected the evidence is:
   “Would we see this kind of evidence even if you were wrong?”
   That’s the heart of Bayesian skepticism.
4. Ask about independent confirmation:
   “If this were real, who else would have noticed by now?”
   The more independent confirmation, the stronger the evidence.

---

8. The Practical Mindset

Bayesian thinking teaches you to:

• Calibrate your confidence (not every claim deserves equal weight).
• Change your mind slowly and proportionally.
• Be open but not gullible.
• Separate possibility from probability.

---

In One Sentence:

“Bayesian reasoning is the habit of asking:
How much should this new piece of information really change what I already thought?”
That’s it — no equations needed.
Every time you instinctively pause before reacting, weigh alternative explanations, or adjust your belief only slightly, you’re already thinking like Bayes.

Using Bayes’ Theorem Without Math

Bayesian reasoning isn’t about crunching numbers — it’s about updating your beliefs proportionally when new evidence appears.

Step 1: Start With a Prior

Before reacting, ask yourself: “How likely was this claim before I heard this new evidence?”
That’s your prior belief — your best sense of what’s plausible given what you already know.
Example: Someone says, “My friend’s uncle cured his cancer with herbal tea.” You already know cancer is complex and miracle cures rarely check out, so your prior is very low.

Step 2: How Likely Is the Evidence If the Claim Were True?

Ask: “If this really were true, would this kind of evidence make sense?”
If herbal tea cured cancer, you’d expect many independent studies and reports, not a single anecdote. The evidence is weak because it’s not what truth would normally look like.

Step 3: How Likely Is the Evidence If the Claim Were False?

Ask: “Would I also see this kind of evidence if the claim were false?”
Yes — coincidences, misremembering, and remission stories happen. If the evidence fits both “true” and “false” worlds, it’s weak.

Step 4: Update — But Only a Little

Ask: “Given all that, how much should this change my mind?”
Usually: a little, not a lot. Anecdotes and rumors rarely justify large belief shifts. Wait for multiple, independent confirmations.

Step 5: Keep Updating

Bayesian thinking is incremental. Each reliable source nudges you one step, not ten. Beliefs evolve through many small updates — like science.

Everyday Bayesian Thinking

Situation	Ask Yourself (Bayes-Style)
Rumor	“Who said this, and how likely are they to be right?”
News Story	“Would this story still exist if it later proved false?”
Political Claim	“Could both sides easily produce this same ‘evidence’?”
Conspiracy Theory	“If this were true, wouldn’t we see lots of confirming data?”
Scientific Study	“Was this predicted beforehand and replicated?”
Friend’s Excuse	“Is this exactly what I’d hear if it weren’t true?”

Quick Questions for Debates

•  Has this happened before? → establishes a prior
•  What’s another possible explanation? → considers alternatives
•  Would we see this evidence even if you were wrong? → tests evidential strength
•  Who else confirms it? → checks independence and replication

The Bayesian Mindset

• Calibrate confidence — not every claim deserves equal weight.
• Change your mind slowly and proportionally.
• Stay open, but not gullible.
• Separate possible from probable.

Core Principle

“How much should this new piece of information really change what I already thought?”

Handwritten Notebook Pages

Source

Open Bayes’ theorem notebook pages — start at notebook p. 162 (file not included in this release).