[SplawnDarts]

[ QUOTE ]
[ QUOTE ]
Is that really what you've termed "Luckyme syndrome" or is it simply a case of 2 conflicting sets of background probabilities? In her experience, P(dead|symptoms) = 1. In your experience P(!dead|wet) = 1.

In my technical experience, because I know a thing or two about wet electronics, I would expect the phone to recover. But I don't see an obvious way, by reasoning from probabilities, that one could conclude you were right and she was wrong.

[/ QUOTE ]


I can't speak for DS, but let me give this a try:

Yes, he is talking about 2 conflicting background probabilities. The difference is, one is much more specific than the other: his own personal experience with that particular phone under his particular usage patterns, versus all phones (of that same make and model) under all usage patterns by all people. I think he is saying that in this case, #1 takes precedence over #2.

[/ QUOTE ]
Her reasoning is specific as well - specific to the symptom.

What we're trying to compute is P(dead|symptom, wet) and Dave has P(dead|wet) and she has P(dead|symptom). We can't apply a naive version of Bayes' Theorem and combine the two pieces of information, because it's very possible that wet is not conditionally independent from symptom. From a probability POV, I believe we're stuck.

I do believe there's a logical flaw in her reasoning, but it's different from what Dave's suggesting. I believe the flaw is a failure to apply Occam's razor.

Dave's explanation of the phone's failure has one obvious cause, and one obvious effect, and it's plausible that they're connected. That's the simplest possible kind of theory in a case like this.

Her explanation has one real (but unnamed) cause, an extra "fake" cause of being wet, an effect, and we'll give her the benefit of the doubt and assume whatever cause she suggested really is plausible for the result.

Now, her theory is more complicated because of the extra fake cause, so by Occam's razor it's less preferable than Dave's.