Re: Reputations

[mccoy@communities.com (Jim McCoy)]

gregburk@netcom.com (Greg Burk) writes:
> mccoy@communities.com (Jim McCoy) writes:
> > Greg Burk writes:
[...]
> > Go read Ender's Game by Orson Scott Card (a good book to read anyway :)
> > and examine the nature of the computer network "discussion groups" he talks
> > about: a classic example of reputation markets in many-to-many discussions.
[...]
> I have read it, a long time ago. Frankly, it's a spectacularly bad
> example. He writes of the two child-protagonists gaining reputations as
> great philosophers on a sort of Usenet. (At the time I believe OSC was a
> member of Delphi, UNCLEORSON)
>
> Look around on the real Usenet. OSC could not have been more wrong.

If you look closer you would probably be surprised to see things starting to
move in this direction.  As the number of participants has grown the "noise"
in most newsgroups has grown to an unmanageable level.  Now most newsreaders
allow you to score authors or article threads so that you can keep individual
reputation and interest files.  I am actually an anomoly at my company, a
collection of very net-savvy people, because I actually still participate in
a few newsgroups; most of the interesting net discussions now take place on
mailing lists which allow further reputation filtering (most mail agents have
better and more flexible filters than news agents) and most bleeding-edge
traffic happens in private mailing lists where one cannot even participate
unless they have already established their reputation.  If these lists were
gatewayed to read-only newsgroups you would have what Card was talking about.

> > > As an "asset", it is extremely non-liquid:
> > >
> > > How exactly would you "convert" your reputation into other capital?
> > > Would you accept bribes and tell lies? Seems to me you would only get a
> > > one-shot "conversion" and it couldn't possibly hope to equal your
> > > investment.
> >
> > Tell that to Walter Cronkite, Siskel & Ebert, Moody's and others who have
> > converted reputation capital into large piles of money [...]
>
> You don't seem to realize you are actually including at least one major
> example of a counterfeit reputation here.
>
> I speak of Siskel & Ebert, whom I have caught at least once giving a
> strikingly dishonest review. I had seen the movie (See You In The
> Morning) on opening day, before they reviewed it. It stunk.

No, this is just an example of how reputations are not global values, each
reputation is modified by the perspective of the user.  _You_ disagreed
with the review and have probably used your experience to weight the
values of future reviews by those particular reviewers.  This is why there
are hundreds of different sources of reviews for movies, people weight the
recommendation given by the reviewer with

> Conclusion: They knew it stunk, but for some reason I won't speculate on
> they wanted to say they liked it so they misreported it, and covered
> their butts with a review that would look mixed later but sound like a
> rave now. I see a counterfeit reputation.

Incorrect.  One thought it was a good film and the other disagreed.  You
did not find it to be a good film and have since modified your weighting
of the Siskel & Ebert reputation value to reflect this.  It is highly
improbable that there were any behind-the-scenes machinations between the
movie backers
and the reviewers: such a person has a value which is directly proportional to
being viewed as impartial and once they have established a reputation the
value in maintaining the reputation outweighs the value a potential briber
could gain by trying to influence the review (nothing will drop the
reputation faster than getting caught cheating, and a single reviewer does
not have enough influence on the public to impact a films box office returns
enough to make the bribe worthwhile.)  It is more likely that you just
disagreed with the review and
you have since learned your lesson and now seek multiple review sources or
else switched to a different source for movie review information (dropping
your personal weighting of Siskel & Ebert down below other sources.)

> > No, I think that you just don't understand the mechanics of reputations and
> > how they interact with the most important resource in most people's lives:
> > time.
>
> I'm tempted to tit-for-tat, but I will not refute your points by telling
> you you just don't understand.

I guess I could have been more diplomatic, but it seems that you just do not
understand that reputations are not a global value, rather they are a
weighted value which is modified over time as the user seeks to determine a
balance of raters and reviewers which most closely represents their
particular viewpoints, interests, and experiences.  There is no one single
reputation which a given person has, all reputations are dependant upon the
source of the reputation and the context in which that particular reputation
is used.

> > time.  Instead of thinking of "reputation" look at it from the other end and
> > consider the "attention marketplace."
>
> Fine, but resolving good vs counterfeit reputations takes time too.

*Sigh*  There is no such thing as a "counterfeit" reputation.  When someone
joins a network with a particular set of interests they will start off by
finding a reputation service(s) which they think, though various channels
ranging from advertisement to word of mouth, closely matches their interests
and views.  This is the only time that outright deception can influence a
person and it is also the point at which deception is least profitable
(because the deceiver will be easily revealed once the user compares the
reputations with what they expect to see and because most new users will
choose multiple services to perform comparison shopping.)  There may even be
reputation services which rate other reputation services to let people know
how the service compares to its stated viewpoints and advertisements.  A
reputation service gains income by establishing a long-term replationship
with the customer, so it is in the services interests to maintain credibility
with its users.  If they do not then that reputation service will have a
negative weighting depending on what the user is interested in, so the
problem of correct vs. incorrect reputations will itself be handled by
reputation services.

An individual will have multiple reputations depending on which service is
providing the reputation and the context in which the reputation is being
used.  "Tim May" may have a relatively high reputation in most services on
cryptography and crypto-anarchy issues but this reputation will not apply to
football predictions or articles posted to soc.culture.swedish.  Someone may
try to burn a reputation to pass off a false statement as truth, but this is
as unlikely to work as it is for Peter Jennings to tell all of his viewers
that this afternoon Bill Clinton appointed me his senior domestic policy
advisor; people now have a wide variety of news and information sources to
use for comparing the veracity of the statement, getting caught diminishes
his reputation and this has a monetary value to him, and because his audience
is larger due to his increased reputation there is a greater chance that
others will investigate the matter and so his chance of getting caught is
higher.

> > Right now reputation markets have a
> > limited presence on the internet (mostly through killfiles) because the
> > tools required are not integreated into the tools used to browse the
> > information. In time this will change.
>
> How? I ask for something more specific than In The Future Everything
> Will Be Done Right.

Version 0.1 (coming to a news server near you by the end of the year) will
take the form of a service whereby you can subscribe to a usenet filtering
service which will present your newsreader with a database of articles which
have already been filtered by the reputation service to remove off-topic and
"me too" posts (or perhaps based upon other filtering criteria.)  The agency
making this service possible will also sell to individuals or groups the
ability to start their own service on this news host and perform whatever
filtering they want, this will also include adaptive filters (if I can ever
get the little bastards to use an internal weighting function which does not
converge too quickly) which will attempt to learn the general weighting
criteria are so that the people running reputation filtering need only update
the filters occasionally and not score each and every posting.

The hard part, and the part which is slowly gaining enough momentum to make
this possible, is the integration of cryptographic signatures into messages
so that one can determine the authenticity of a message and thereby assign
a reputation value to a real identity instead of an easily forgeable email
address.

jim