The Raven Paradox
Does seeing a green apple provide evidence that all ravens are black? Surprisingly, logic says yes!
Confirm the Hypothesis
Click objects that confirm "All ravens are black"
The Hypothesis
The Logic
Philosopher Carl Hempel observed in 1945 that these two statements are logically equivalent:
This is contraposition: "If X then Y" is equivalent to "If not-Y then not-X."
Now consider two principles that seem obviously true:
- Nicod's Criterion: Observing a black raven confirms "all ravens are black"
- Equivalence Condition: If evidence confirms a statement, it confirms all logically equivalent statements
But then: seeing a green apple (non-black, non-raven) confirms "all non-black things are non-ravens" which confirms "all ravens are black"!
Why It's Strange
🤔 The Problem
Intuitively, learning about ravens should require actually looking at ravens. How can examining apples, shoes, and sky teach us anything about birds?
If every non-black non-raven confirms "all ravens are black," then you could confirm ornithological hypotheses without leaving your kitchen!
Even stranger implications:
- A green apple confirms "all ravens are black"
- The same green apple confirms "all ravens are white" (via contraposition)
- In fact, it confirms "all ravens are [any color]"!
Something seems wrong with either Nicod's Criterion, the Equivalence Condition, or our intuitions.
The Bayesian Resolution
Many philosophers accept that green apples DO confirm "all ravens are black"—just by an incredibly tiny amount!
📊 The Numbers
Suppose there are 10,000 ravens and 10,000,000,000 non-black objects in the world.
Seeing a black raven: You've checked 1/10,000 of the potentially falsifying cases.
Seeing a green apple: You've checked 1/10,000,000,000 of the potentially falsifying cases.
Both confirm, but the raven provides 1,000,000× more confirmation!
The paradox arises from treating all confirmations as equal. They're not. Context matters:
- How many ravens exist vs. non-black things?
- How likely was finding this evidence?
- What's the prior probability of the hypothesis?
Alternative Resolutions
1. Selection Effects
HOW you found the evidence matters. If you specifically sought non-black things and found a non-raven, that's less informative than random sampling.
2. Reject Equivalence
Some philosophers deny that evidence for one statement transfers to logical equivalents. The statements are true together but confirmed differently.
3. Natural Kinds
"Raven" is a natural kind; "non-black thing" isn't. Confirmation only works for genuine categories, not arbitrary negations.
4. Accept the Paradox
Green apples really do confirm! Our intuition is wrong. We just vastly underestimate the difference in confirmation strength.
Why This Matters
The raven paradox illuminates deep questions about scientific confirmation:
1. What counts as evidence?
Not all observations are equally relevant. Confirming "electrons have negative charge" by not finding positively-charged electrons seems silly, but it's logically valid.
2. The role of background knowledge
We always interpret evidence against what we already know. A green apple confirms "all ravens are black" only if we didn't already know it was non-black and non-raven.
3. Limits of deduction
Logical equivalence doesn't preserve all properties. Two statements can be true together yet differ in meaning, evidence, and explanatory power.
Fun Variations
Indoor Ornithology: A philosopher locked in a windowless room examines only household objects. By checking that none are non-black ravens, they accumulate evidence for "all ravens are black"—without ever seeing a raven!
The Grue Paradox (Goodman): Define "grue" as "green if observed before 2030, blue otherwise." All emeralds observed so far equally confirm "all emeralds are green" AND "all emeralds are grue"—but they predict different things about future emeralds!
Infinite Confirmations: Any observation confirms infinitely many hypotheses. Seeing a black raven confirms "all ravens are black," "all ravens are black or invisible unicorns exist," etc. How do we choose among them?