From: Wei Dai <weidai@eskimo.com>
Date: Tue, 28 Nov 1995 20:02:03 +0800
To: Michael Froomkin <froomkin@law.miami.edu>
Subject: Re: towards a theory of reputation
In-Reply-To: <Pine.SUN.3.91.951122095930.22711D-100000@viper.law.miami.edu>
Message-ID: <Pine.SUN.3.91.951128012308.9094C-100000@eskimo.com>
MIME-Version: 1.0
Content-Type: text/plain


On Wed, 22 Nov 1995, Michael Froomkin wrote:

> This discussion puzzles me.  I thought we were bombarded with 
> reputational goods all the time:  brand names, stocks (what is a purchase 
> in the 2ndary market but a purchase of reputation most of the time?), 
> degrees from famous universities.  Anonymity compliates matters only if 
> no systems of unique ID is used.  Throw in digital signatures and we are 
> back at brand names, aren't we?

It's true that we deal routinely with reputations now.
However there is very little formal analysis of reputation as a concept.  
Although our common sense knowledge of reputations seem to serve 
us fairly well(*), there is no guarantee that it will scale well to an 
anonymous market where both the number of participants and the importance 
of reputation are much higher.

* However, the government apparently doesn't think so.  Witness the FDA and 
the SEC.

We need to have formal algorithms to deal with reputations, and we need 
to be able to show that they have desirable properties.  This will reduce 
transaction costs and help bring anonymous markets into the mainstream.  
Perhaps more importantly, good reputation algorithms will make 
agorics computing possible.  There is a very interesting proposal for a 
network routing system based on microcurrency and positive reputations 
(see http://www.webcom.com/agorics/dsr.html).  However it does not say 
what algorithms will be used to handle reputations.  If the system is 
actually implemented, its proper functioning will depend as much on the 
properties of the reputation algorithms used as on the correctness of its 
protocols.

Wei Dai