From: tcmay@got.net (Timothy C. May)
Date: Thu, 5 Sep 1996 14:44:42 +0800
To: cypherpunks@toad.com
Subject: Re: Reputations
Message-ID: <ae5392e4070210049904@[207.167.93.63]>
MIME-Version: 1.0
Content-Type: text/plain


At 1:11 AM 9/5/96, Greg Burk wrote:

>That part of the "reputation capital" theory has always seemed
>suspicious to me. "reputation capital" doesn't behave linearly. There's
>too much incentive to bottom-feed and too little incentive to shoot for
>the heights. As an "asset", it is extremely non-liquid. It is hard to
>spend it in a controlled manner.

Sure, it isn't fungible, it isn't transitive, it isn't neat and clean.

But it's the best thing we've got, imperfect as it is (and must be, I believe).

>Too much incentive to bottom-feed:
>
>For example, let's say there's someone well-known who frequently speaks
>nonsense on crypto issues. We'll call her "Norothy Nenning". She makes a
>recommendation on some particular crypto issue, say "The government's
>Nipper chip is a safe and effective form of crupto". Plenty of naive
>people will credit her to some degree. True, fewer people than if she
>had carefully husbanded her reputation, and to a lesser degree, but
>still a lot more than zero.
>
>Notice that that's a zero cost/benefit ratio. She never does anything to
>husband her reputation, she just spends it every chance she gets. And
>while no single expenditure rewards her as much as it would if she made
>the same expenditure with a good reputation, she spends so much more
>freely that it is a good strategy for her on the whole.

To stick with my restaurant example, consider _advertising_. MacDonald's
and Burger King spend hundreds of millions of dollars every year claiming
their "restaurants" are great. Many millions of people obviously are
swayed. So?

Others choose not to trust the advice of the MacDonald's hucksters. Maybe
only a tiny fraction choose Chez Panisse over MacDonald's. This is the way
of the world.

It's still the give and take of reputations. It ain't perfect (in that it
doesn't produce results I believe are empirically valid and optimum :-}).
But it's all we have. It's the market. The agora.

>"Reputation capital" is hard to spend down to absolute 0 because it is
>significant work to distinguish valid "reputation capital" from
>worthless counterfeit, and it is easy to counterfeit... just talk.

I strongly disagree. It's quite possible for Person A to quickly convert
his reputation to Person B to a _negative_ value. Real quick, in fact.

Perhaps my short article did not fully explain a few things. Reputations
are a _tensor_ or _matrix_ quantity. Person A has a reputation R(A,B) to
Person B, a reputation R(A,C) to Person C, and so on. (And the matrix may
be further broken down into reputations for advice on various subjects, in
various fields, etc.)

We may lump a lot of folks together and say, for example, that MacDonald's
has a reputation of R (MacDonald's, lots of people) = 0.7531. And perhaps R
(Chez Panisse, lots of people) = 0.0013 (i.e., they don't know what it is,
and so value the rep of Chez Panisse at near zero).

And so on. Lots of examples could be given.

Now suppose that J. Anonymous Gourmand announces that MacDonald's is shit.
How much will anonymous claim hurt MacDonald's? Obviously, not much. But
what if the American Heart Association publishes a detailed study on the
fat levels of MacDonald's products and declares it to "Dangerous." The
effect will probably be greater, as R (AHA, many people) = high, and by the
kind of Dempster-Shafer belief calculus I discussed a few months ago, the
rep of the AHA propagates semi-transitively to the rep of MacDonald's.

(This all happened recently, with the famous studies of fat levels of movie
theater food...sales dropped almost overnight, and now the fat levels of
popcorn, etc., have been changed for the better.)

This is a real example of how reputations matter, how negative and positive
reps matter, etc. Note especially that the "identity" (in the Dyson sense
of providing True Name accountability) of an opinion-giver is not what it
is important...it is not the essence of why people believe or don't believe
the opinions of others.

(Some years ago on the CBS station in San Francisco, there was an
"anonymous gourmet" who visited restaurants and gave reviews. His reviews
were taken quite seriously, and his anonymity did not matter, provided his
personna was _persistent_. That is, provided that people thought it was
"the guy they had come to trust," and not, say, a guy the station recruited
off the street each day and sent out as the "anonymous gourmet." In the
case of this guy, his face was cloaked in shadows, but his voice was
distinctive. (His voice on the show was probably different from his
food-ordering voice, so restaurants would not know who was ordering and
alter the food or service.) Much could be written abou the role of
anonymity in such reviews, in tests of service, etc.)

>I anticipate the answer "Well, the work pays off". But that misses the
>point. Frequently the work required to tell the good "reputation
>capital" from the worthless is as much as would be required to find the
>straight dope yourself.

Reputations work OK for me in the real world. Given the limits on a lot of
ontological facts, hard to see how it could be better.

I've already spent too much time writing this, so I can't address the
remainder of your points.

--Tim May

We got computers, we're tapping phone lines, I know that that ain't allowed.
---------:---------:---------:---------:---------:---------:---------:----
Timothy C. May              | Crypto Anarchy: encryption, digital money,
tcmay@got.net  408-728-0152 | anonymous networks, digital pseudonyms, zero
W.A.S.T.E.: Corralitos, CA  | knowledge, reputations, information markets,
Higher Power: 2^1,257,787-1 | black markets, collapse of governments.
"National borders aren't even speed bumps on the information superhighway."