• From: "Len Bullard" <cbullard@h...>
• To: "'bryan rasmussen'" <rasmussen.bryan@g...>
• Date: Fri, 29 Sep 2006 10:39:24 -0500

First, take a look at this article and take note of the 'network effect' and
'locking'.  

http://www.ft.com/cms/s/cd46a446-4edf-11db-b600-0000779e2340.html

To understand locking, it helps if you understand second order systems.
Locking is an effect of the selectors feeding back to the index of scalars.
You think of that as a 'democratic decision' and in a naïve way, it is and
that is why there are few if any democratic governments but lots of
republics.   It helps if you understand why that is the case but it won't
sit well with anyone who thinks a flat democracy is a good form of
government.  You are probably trolling so I don't expect you to give a good
answer, but actually, once the selections are made, what you have are
derivative results, so you don't end up with a democracy anyway.  You end up
with the choices of the choosers of the choices increasing in value and
producing the so-called 'long tail' power law.  It doesn't necessarily
produce sales of low volume ideas.  It can effectively extinguish them.

http://blogs.zdnet.com/Ratcliffe/?p=171

If someone is basing their evaluation on an eigen-value index, they are not
going to see new ideas, products, and particularly, language.  This is a
model based on previous selections, not a topical model (see topical vector
model) so, you have to be willing to go deep into the search returns to find
weak signals (the edge of the network in that model and in a topical vector
index, the outliers.  If you don't you won't understand or even recognize
emergent events.

If you like, we can take up the ecosystem extinction effects arising from
catastrophe, competition, speciation, bottlenecks, functional failures, and
coextinction, or why it is better to use immunological techniques that
affect the ability of a bacteria to communicate (quorum sensing) as gaming
strategies but the eigen-index model is sufficient to make the point.

Now addressing your points:

>> Force is the feedback effect over time of the naïve indices in use.

>Do You mean: due to the simplicity of the indexing, no metadata etc.,
>the primary mode of determining importance of any specific resource is
>a democratic decision.

No.  See above.

>Indirectly you are indicating that Google's not pointing at enough
>resources for "false eigen-index locking" suggests that their index is
>inadequate to an expert's requirements in a particular subject matter?

That depends on the user and the subject matter.  What it does mean is that
simply slicing off the top Google returns is a weak assertion of authority
for any decision.   In a modeling world, you would only be querying the
environment, not the situation.  In orbital mechanics, you may have a topic
in a lagrange point.

>Do you mean: A reference to the point above - because of the massive
>scale of Google's index can be gamed by sending in more information.
>That it needs to be filtered to make sure there is no gaming?

Essentially yes, but really, by careful or naive construction of the
information sent, and by ignoring those parts of the information returned
that are inconvenient to the claims.  That is why patent review narrows
claims while patent submission attempts to expand them.  This is a game in
and of itself but the game has clarity and is evolutionarily stable UNLESS
the domain in which the claims are made is itself murky.  That is why
becoming familiar with means to improve clarity is in everyone's interest
except the cheaters.  See references to IBM's work with the USPTO.

>> One of those is vetting assertions against other assertions with time-
>> variant properties or restricting the domain of the citation (eg, 
>> inverted indices restrict the domain to the book; library cards restrict
>>it to the book title, author, date, etc; cross-domain indices make no
>>assertions beyond location, and so on).

>Do you mean: Google Scholar, Froogle and other specialized access
>points to the Google Index attempt to combat the above named problem?

Yes.  Although the problem is if all they are indexing is URI-resources,
they are missing a large portion of the prior art.  I suspect that is why
scanning and publishing is a big part of their plan.  It is analogous to the
problem of network surveillance in a packet-based system:  certainty
requires that all of the information be scanned.  Otherwise, citations
require a bounded set to make logical assertions (another reason for the BOS
in Hytime) and why Sterling's suggestion is impractical in his formulation
but workable where common authoritative domains can be referenced
singularly.  Think of the IP keiretsus of the standards organizations where
a lot of energy is expending vetting the claims before they are accepted.

>If someone had asked me what the false locking of an eigen-index meant
>I would have guessed something similar to "that an operation on a set
>of eigenvalues does not produce non-eigenvalues as the output?" (which
>I would have tried to say with a particularly stupid and humble look
>upon my face) But this is evidently not what it means. What does it
>mean?

No.  It means the eigen-values being selected feedback into the system of
selection and can introduce false relationships which are amplified and
create chaos or distortion unless filtered because those values are scalars.
Again, it is a second order system and I conjecture, tensor being
time-variant over a topical space/manifold.

len

• References:
  • Re: XML-DEV list
    • From: "bryan rasmussen" <rasmussen.bryan@g...>