From: Eliezer S. Yudkowsky (sentience@pobox.com)
Date: Fri Apr 18 2003 - 16:54:44 MDT
Harvey Newstrom wrote:
>
>>This leads to the approach to nutritional science
>>that Eliezer seems to be suggesting and which I also
>>recommend (in which any non-paleolithic nutritional
>>hypothesis is the competing hypothesis which
>>must disprove the default paleo hypothesis). Do you follow?
>
> Yes, but I disagree with this method of debate. I think it is invalid to
> choose a default position which does not have to be proven. Established
> theories must become established with evidence and proof. They do not
> become "default" first, with the burden of disproving them falling to
> others. Scientists must falsify their own theories, test them either way,
> and provide the results as evidence. There is no "default" position or
> "burden-of-proof" in science. Such arguments are more often used by
> religions and cults.
What I said was that if you were going to perform an outcome study, you
should test the modern setting of a metabolically salient variable (for
example, sodium-potassium balance) against its ancestral setting. This
says nothing about what the default hypothesis for *best diet* should be,
only that this is the way of testing one change at a time. This becomes
clearer if you look at the chaos of present dietary studies - people
testing, e.g., lots of sodium against almost no sodium with no attempt to
control for potassium, rather than testing a modern sodium-potassium
balance against an ancestral sodium-potassium balance. In scientific
terms this is simply an obvious, straightforward method for testing the
effects of incremental changes in order to determine the workings of the
human metabolism. One may or may not take the ancestral diet as the
"default optimum" but it is certainly the setting that you would test any
specific diet *against* if you wanted to see its effect on the human
metabolism. It's not a question of burden-of-proof tennis; I'm not
telling you what the "null hypothesis" is. I'm saying what the control
group in an outcome study should look like.
Or rather, it doesn't matter what the null hypothesis or Bayesian prior or
whatever *was*, because there are now enough specific cases of modern
diets being detrimental because of violating ancestral invariants that I
would, indeed, tend to take as the *new* working assumption that the
ancestral diet is better until proven otherwise. That's not playing
burden-of-proof tennis, that's a generalization from what's been learned
so far.
I again advise everyone interesting in arguing this to start reading
through the papers at:
http://www.beyondveg.com/cat/paleodiet/index.shtml
http://www.thepaleodiet.com/articles.htm
The first rule of scientific argument on the Internet is to get your
information by Googling the science papers, not by reading other people's
impressions.
-- Eliezer S. Yudkowsky http://singinst.org/ Research Fellow, Singularity Institute for Artificial Intelligence
This archive was generated by hypermail 2.1.5 : Fri Apr 18 2003 - 17:05:45 MDT