From: Alexander Chislenko <sasha@cs.umb.edu>
Date: Tue, 22 Feb 94 20:44:38 PST
To: hughes@ah.com
Subject: Re:  RATINGS: Subject tags
Message-ID: <199402230444.AA29186@eris.cs.umb.edu>
MIME-Version: 1.0
Content-Type: text/plain


  For quite a while now, I have been advocating a rating system based
on *calculating correlations* of people's interests in each topic domain.
 
   This allows to:
 
   - give personalized ratings to each user;
   - filter out any number of "evil tentacles" recommending you junk;
   - suggest incentives to each user to rate articles;
   - introduce digicash payments for all messages and ratings;
   - implicitly use quality criteria that are difficult to formalize;
   - find like-minded people, even if you never post anything
 
  In short, it works like this:
 
you read stuff, if you like it, you pay for it as much as you want;
what you pay is considered your rating of the article; the System
calculates correlations of your ratings with everybody else's;
you can select articles that people with your interests in the given
area liked most; so it makes sense for you to pay and be sincere -
otherwise the system may not work for you, and you will read too
much junk.  Also, people whose ratings helped you choose it, can get
some share of your payment - another incentive to read early and
carefully, and be fair (you can actually earn credits as a critic).
 
  Authors can establish minimal payments for their texts, but it will
be not their advertizing headers, but the opinions of like-minded
with *you* people that will help you choose texts.
 
   I have a more or less detailed development proposal for such a system 
and would be happy to share it if anybody is interested.

------------------------------------------------------------------------------
|  Alexander Chislenko | sasha@cs.umb.edu | Cambridge, MA  |  (617) 864-3382 |
------------------------------------------------------------------------------