[extropy-chat] Re: monty hall paradox again: reds and green gorfs

[Eliezer S. Yudkowsky]

Harvey Newstrom wrote:
> Spike writes:
> 
>> If
>> most people say red and green gorfs are equal but a
>> small and perhaps not very credible minority insist
>> reds are better, while no one is actually claiming the
>> greens are better, then I would choose a red gorf,
>> assuming the same price.
> 
> Spike, this would lead you to superstition.  Throwing salt over your 
> shoulder after you spill it won't hurt and doesn't cost anything.  Most 
> people say it doesn't matter.  But some few do claim that it will keep 
> evil away.  So you might as well do it just in case, right?  The same 
> goes for blessing someone after they sneeze, and avoiding the path of a 
> black cat, or any other unfounded superstition.  Might as well do it, 
> just in case, right?

We're talking about a difference in *taste*, right?  There's nothing 
supernatural about a taste difference that some people can detect and some 
people can't.  The prior probability of the hypothesis is much higher. 
Assuming the same price, I'd buy a red gorf.

If I can buy more than one, I'd buy a red gorf, a green gorf, and have 
someone feed it to me blindfolded, before I knew which was which, just to 
see if I could tell the difference, because I would be curious.

There's also a time cost to the other acts you mentioned.  And I would not 
object in the slightest if someone were to test the assertion that 
throwing salt over your shoulder leads to good luck (as measured by a 
series of coinflip gambles), or that blessing people after they sneeze 
(using atheists and theists delivering insincere and sincere blessings) 
leads to less sneezing.  That's good science, so long as you use large 
sample sizes and severe significance tests (p < 0.001, say) to make up for 
the low prior probability, and so long as you drop the hypothesis like a 
wet noodle after the disproving evidence comes in.  My inclination in such 
matters is that the thing to do with a hypothesis is test it.

-- 
Eliezer S. Yudkowsky                          http://singinst.org/
Research Fellow, Singularity Institute for Artificial Intelligence