[NEC] 2.3: Power Laws, Weblogs, and Inequality

S. Alexander Jacobson alex@i2x.com
Sat, 1 Mar 2003 14:07:55 -0500 (Eastern Standard Time)

On Fri, 28 Feb 2003, Dave Long wrote:
> That is to say, I understand your story for
> (A implies B), and we started this thread
> by being given B, but as (B->(A->B)) is not
> a wonderful inference, it would help to at
> least make an attempt at showing A.

I am not sure what you mean by showing A.  To me
it is obvious that consumers have correlated
preferences.  If they didn't, all products and all
media would be entirely random.  The existence of
reviews, recommendations, marketing, editing,
and quality control jobs as well as my belief that
I exist and make judgements that are not random
all constitute fairly solid evidence in favor of
the notion that consumers have correlated
preferences.  If you want data specifically about
blogs, that is not unreasonable.  But, given the
pervasiveness of correlated preferences in almost
every other aspect of human choice, I think the
burden is really on those who would argue that
blog preferences are NOT correlated.

Given that consumers have correlated preferences
and my prior explanation of why correlated
preferences should result in a power distribution
of e.g. blog links, it should not be surprising
that blog links do in fact have a power

Therefore any theory about path dependence
influencing blog links has to take into account
the force of consumer preference and is more
complex.  Unless it can define a different outcome
than that produced by consumer preferences, we can
exclude it using Occam's razor, except to the
extent that path dependence might play a role in
consumer preference*


* To be mroe formal:

(1) A:=consumers have correlated preferences.

(2) B:=blog links exhibit a powerlaw distribution

(3) A (Observation of marketing, editing,
	reviewing, quality control behavior in the world)

(4) B (see NEC 2.3 from Clay Shirky)

(5) A -> B (proof earlier in thread)

(6) C:= Blog links are entirely random

(7) D:= Blog links are normally distributed

(8) C -> D (Central limit theorem)

(9) D -> ~B (normal ~= power distribution)

(10) ~C (C->~B ^ B)

(11) E:= All other theories of B require more
entities than A->B or C->D and data that we have
not yet observed

(12) E (Observation)

(13) F:= we have no other observations that
supplyu entities for a better theory

(14) F

(15) ~A -> ~B (F ^ Occam's Razor)

(16) F:= Path dependency causes power

(17) (F->B)->(F->A) (A -> B ^ ~A -> ~B)

In english, (17) means that if we want to talk
about path dependency, we can only talk about its
effect on consumer preferences.  It doesn't make
sense to talk about it w/r/t blogs directly.

Unless you disagrees on 1-17, I assume we are in
agreement that we have no reason at this point to
believe that the distribution of blog links
originates from any cause other than consumer


S. Alexander Jacobson                   i2x Media
1-212-787-1914 voice                    1-603-288-1280 fax