[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: A problem with anonymity


At 05:15 PM 9/1/95 -0700, you wrote:
>At 10:45 AM 9/2/95, David Murray wrote:
>>[Akerlof, if I remember my economics right (and I am confident that I will
>>be corrected if I don't) analysed a market for used cars.
>>... So the only cars for sale would be lemons :-)]
>I haven't encountered this example, but it clearly misses some important
>real-world issues.
>Like a lot of simple game-theoretic models, the application to the real
>world is quite different.

True. But it does make the point that, in a perhaps surprising way, the
bad can drive out the good. [I certainly don't want to get into any kind
of normative v. postive methodological debate.] The lesson is not that
"You can never find a decent used car", but that, in a market with particular
characterisics (particularly with regard to what information is available
to whom), lemons drive out the good. This just means that owners of good
used cars offer them for sale in a market where pre-purchase checks are
possible, etc..

But this is to concentrate on the example, not on the substance...

>But I certainly agree that crypto will reignite interest in analyses of
>such game theory questions. Another way of viewing anonymity vs.
>non-anonymity is that knowing the True Name of a party with whom one trades
>is just _one element_ of a transaction. By no means is it essential.

I guess I was using True Name somewhat unusually. I didn't mean "True Name=
state approved unique identifier of a human being" so much as "True Name<>
easily discardable/transferrable/sellable digital pseudonym". 

I certainly did not mean to imply that net.commerce is impossible, or that
it would only take place on the basis of True Names, however defined. Rather
I was commenting that one pervasive feature of non-net.commerce is the
ability to track down someone that owes you something and sue them/beat it
out of them. You can't do this to a digital pseudonym. As you rightly
pointed out, you can't do this to someone who escapes to South America,
or to the guy who runs the market stall that won't be there tomorrow. So 
you take a bond, or you take your chances. I still think that, because of
the (perfect) ease with which net.rep's are transferrable/cash-in-able,
the chances you take in the digital domain are so much higher as to be
(almost?) qualatatively different.

>>If you can't rely on the unsecured promise of a digital pseudonym, and you
>>can't accept reputation as 'security', how do you extend credit?
>I am willing to extend some amound of credit to PrOduct Cypher, Black
>Unicorn, etc., based on their past reputation and on the fact that I can
>show to others the transactions into which their pseudonyms entered and
>thus expose them if they default. Now _how much_ I'm willing to extend is
>of course a more complicated issue, but the principal is still there: a
>purely digital pseudonym, with no possibility of being tied to a True Name,
>can still be extended credit....I just said I would do so.

This would, of course, allow PrOduct Cypher (for example) to cash in on hir
rep. [Hir - never thought I'd see myself use it: ughhhhhhh <shiver>]. But,
unlike the unscrupulous stall owner, who would have to sell the gold watch
before he decamped, PC could sell hir rep before the rip-off had been done.
This would be a sort of division of labour - rep-builders and rep-exploiters.
In cybersapce, you could never tell whether you were dealing with the rep
builder (buying a good car) or a rep exploiter (buying a lemon).

Yes the rep-exploitation would be a one-shot thing. Within seconds of the
sting PC's name would be mud. [And, yes, you could probably tell whether
it was one of those two you were dealing with - if it asked for credit, it
wouldn't be 'corn or 'pher :-)]

But it is the _possibility_ of the scam that would shape the market.

By the way, my final question was not rhetorical. I _do_ think methods of
extending credit to pseudonyms will be developed. I just think they will
be based on (possibly new) types of security interest. An old type of 
security interest (perhaps the oldest) could be used right away - the pledge.
If Bob Pseudonymous pledges $100 worth of digital certificates, I would be
glad to lend him $100 (well, perhaps $80 - gotta secure the interest:-)

Cryptography might develop other ways of protecting the interests of
creditors, while preserving the anonymity of debtors. In fact, I'm certain
it will. And I'm just as sure the law/mercantile practice will assist in
the process.

And like you, I believe it is not only possible, but preferable, that this
is done without the interevention/'assistance' of the state.



Version: 2.6.2