Moving beyond "Reputation"--the Market View of Reality

[Wei Dai]

On Sun, Nov 25, 2001 at 03:05:18PM -0800, Tim May wrote:
> For many years some of us have argued strongly for "reputation" as a 
> core concept. Someone, perhaps even one of our own, even coined the 
> phrase "reputation capital."
> 
> Reputation is an easily understandable concept which explains a lot 
> about how imperfect protocols in the real world nevertheless "work." I 
> won't go into what reputation is, even as defined by folks like us.
> 
> But there are many aspects of reputation which lead to problems:
> 
> 1. The assumption that an agent or actor possesses a "reputation." A 
> kind of scalar number attached to a person, a bank, an institution, or 
> even a nym.

But there is a scalar number attached to a person which deserves the name
"reputation capital", namely his own judgement of what his reputation is
worth. The idea of reputation capital solves an important problem: How do
we prevent nyms from doing bad things, disappearing, and coming back under
a different nym? If a nym has a positive reputation capital, then
disappearing is costly, so that provides a disincentive to do bad things. 

> 2. When in fact different people have different assessments of some 
> agent's reputation. Thus suggesting strongly that reputation is not 
> something attached as simply as above.

Yes, that's why we should distinguish between reputation and reputation
capital.

> 3. All of the nonsense about how "Alice's reputation has been harmed," 
> deriving from the faulty notion of this scalar property attached to 
> Alice.

But "Alice's reputation capital has been harmed" is not nonsense. That
just means Alice thinks her reputation is now worth less than before.

> Aren't we stuck with reputation?
> 
> No, a broader ontology of objects and beliefs about them is a better way 
> to go.
> 
> The "reputation of the dollar" is related to my belief, and the belief 
> of billions of others around the planet, that for whatever reason a 
> piece of paper with the right markings on it will in fact be accepted by 
> billions of others, by millions of small banks and moneychangers, and 
> even by the U.S. Government. And the related belief that loans, IOUs, 
> promissory notes, bonds, and numerous other instruments denominated in 
> these "dollars" will very likely be accepted or exchanged, blah blah, by 
> millions or billions of other actors. Such is not the case with Monopoly 
> money or even with E-gold.
> 
> Thus, what is the "reputation of the dollar"? Is it because of foolproof 
> anti-forgery measures? Is it because of the laws of the U.S.? Etc.?
> 
> No, it is a kind of collective hallucination.

There are lots of situations where Alice does X only because she expects
Bob to do Y, and Bob does Y only because he expects Alice to do X. Money
is an example of this, and so are virtually all other social phenomena. I
would call this collective reality, not collective hallucination. 

> Before James Donald freaks out and cites Objectivist arguments that Some 
> Things Are Real, etc., let me point out that "collective hallucination" 
> is mostly a cute phrase. In actuality, our perception of reality is more 
> than just an opium dream. Empiricism, falsifiability, Popper, all that 
> good stuff. But our monetary system is vastly less provably real than 
> the world of atoms and stars is. Because money is fundamentally about 
> bets on the future: will something be exchanged for something else, will 
> governments support what they print, what will the dollar be worth in 5 
> years, etc.

I don't see why our monetary system is less provably real than the world
of atoms and stars. Every proton in an atom can spontaneously decay.
Everyone in the world can spontaneously decide to stop accepting US
dollars as payment. It's all a matter of probabilities.

> All crypto is economics. All money is based on belief. All a matter of 
> "betting," of risk/benefit analysis. Related concepts, of course.

All crypto is economics. Unfortunately the economics doesn't seem to favor
much of the more advanced crypto we're interested in. Just to cite one
example, there are a small number of people who value privacy very highly,
and a larger number of people who value privacy somewhat. But there's no
way to charge them different prices to provide the same untraceable
communcation services to them. Since you need a large number of users to
provide cover traffic, you have to charge a low price for everyone, and
that doesn't seems to be a profitable business.

Anonymous ecash seems to have this problem, plus many more.

I think we may have been mislead by the extremely favorable economics of
basic crypto (i.e. exponential attack/defense cost ratios for encryption
and authentication) into thinking that all crypto have favorable
economics.

> Even slightly flawed protocols still "work," given the right embeddings 
> in other systems. (For example, a common flaw cited with remailers is 
> that if there is not enough cover traffic, traceability still exists. 
> But exactly the same flaw exists with money: try getting untraceability 
> with coins if only a few coins exist. Ditto for bearer bonds. Ditto for 
> lots of things  where the "protocol fails for small N" but works 
> reasonably well--in the "betting" sense--when a lot of actors are 
> trading a lot of coins and currency.
> 
> The value of a monetary token is NOT something that is determined by 
> precise mathematical protocols. It's a value based on _belief_ or 
> _expectation_ about the behaviors of other actors, and about the future. 
> Currency suspected of being counterfeit may sell for 10 cents on the 
> dollar, to a sophisticated buyer, while currency suspected of being 
> legit may or may not sell for at or near face value. (Even perfectly 
> legit currency would sell at a discount in large quanties, probably, 
> because a buyer would be a money launderer. Hence the discount for risk. 
> That is, a market decision based on the obvious tradeoffs.)
> 
> Back to reputations.
> 
> Seen as part of a larger ecology of a "market construction of reality," 
> there are no fixed or absolute values, no fixed or absolute truths. Some 
> assertions are "many nines" likely to be true, and some are even 
> constructed to be true (*)
> 
> (* As in "2 + 2 = 4," though the streetwise person who says "What's the 
> trick?" is realizing that even "known to be true" assertions may not be 
> true, as in base 3. Magicians and con men have known this for a long 
> time.)
> 
> Thus, there is no fixed "reputation" of either a person, an idea, or a 
> unit of value. Everything is a matter of belief, of expectation...
>
> Instead of an ontology of objects and their attached methods and 
> property lists, including "reputations" and "monetary values," we should 
> be thinking in terms of these objects as just other actors, with each 
> actor maintaining his own internal model of "possible worlds" (how he 
> thinks the other actors will behave, what he thinks may be future 
> outcomes, what his own goals and expectations are). Seen this way, there 
> is no "reputation" or "value" that is universal. Everything is relative. 
> Everything is seen through the light of internal states/possible worlds.
> 
> This is the market view of reality. There is no "Reality." Just 
> ensembles of actors, various facets, incomplete knowledge...all 
> lubricated by betting. Every street kid knows this.

Isn't this just standard Baysian probability theory/decision theory/game
theory? Where are you going with this?