3 ms·
"In the mid-1970s, we learned that dietary fat raises LDL. In the late 1970s, we learned that high LDL correlates to cardiovascular disease. So, the conclusion
by blahedo 17y ago
"In the mid-1970s, we learned that dietary fat raises LDL. In the late 1970s, we learned that high LDL correlates to cardiovascular disease. So, the conclusion was that if A leads to B, and B leads to C, then A leads to C. But the premise is incorrect. Why? It suggests that it’s transitive. That isn’t the case."
Specifically, there's no problem with transitivity, but (just based off you've written) a correlation/causation problem: A leads to B but B correlates with C. A conclusion that A leads to C would therefore be invalid. (Not sure if this is an error in the summary or the video, and in any case it doesn't undermine the larger point; I'm just being pedantic. :)
- jerf 17y agoActually, in the video he points out the following is invalid: 1. A -> B 2. B -> C 3. A -> C 4. ~A -> ~C Step 3 is valid (skipping over the correlation issue); the problem is actually step 4. ~C -> ~A is valid, but ~A -> ~C is a fallacy. The correlation issue doesn't come up, possibly because it isn't even necessary; this logic fallacy dominates it. I used standard Aristotelian logic notation here; it really ought to be fuzzy logic, but step 4 isn't valid there either, so it would just muck things up, and I don't know the numbers for the fuzzy logic anyhow. Doesn't matter.
- ced 17y agoYou should use probabilities, actually. And while step 4 is not a logical consequence, it may have substantial probability, per Bayes' theorem: P(C | A) = P(A | C) * P(C) / P(A)