From: jim bell <jimbell@pacifier.com>
Date: Sun, 29 Sep 1996 17:15:14 +0800
To: Bill Stewart <cypherpunks@toad.com
Subject: Re: Technical difficulties with AP [AP NOISE]
Message-ID: <199609290715.AAA12227@mail.pacifier.com>
MIME-Version: 1.0
Content-Type: text/plain


At 09:07 PM 9/28/96 -0700, Bill Stewart wrote:
>Aside from details like dead bodies, vendettas, and government suppression,
>there are technical complications with Assassination Politics
>that make it more than the simple job Jim Bell is imagining.
>Some of them provide ways to defend against AP, and turn it into more like
>Extortion Politics.
>
>        [Do I _really_ want to bring up more discussion of AP?
>        Not sure, but if it's clear that it won't work very well,
>        it'll be less likely for the government media to freak about :-)]

OTOH, if it's going to work, the earlier you're satisfied about that the 
better, huh?


>Let's look at the critical part of the problem - paying the assassin.
>The model we've generally been using is that the players are the
>victim, the escrow agent who manages the system, an enthusiastic public,
>and the assassin or assassins who are competing for the jackpot.

All of whom could play multiple roles in difference circumstances, 
incidentally, at different times.  The public should be particularly 
interested in this process, BTW, because they realize that someday they may 
want to participate as donors, bettors, or they may have no choice but to 
participate as...victims.


>I can see three approaches to identifying the correct payee:
>1) Payee provides physical evidence to the escrow agent - 
>        the traditional approach doesn't work here: since the
>        escrow agent is anonymous, so you can't mail him the
>        victim's wallet or finger with well-known fingerprint or whatever.
>2) Assassin leaves physical evidence at the scene which the news media
>        would be likely to report, which payee confirms with escrow
>        agent, presumably committing in advance (e.g. after the event,
>        the payee sends a key which allows the escrow agent to decode
>        the encrypted message that said "I'm leaving a note on the body
>        saying 'Jacques De Molay is Avenged - Assassin's Guild Member 
#32767'.")
>        Works fine for the first couple of assassinations, but after a while
>        the police will catch on and stop revealing kinky details to the 
media.
>3) The main solution has been the gambling deal - 
>        it's just a lottery on the date of death of the victim, which 
presumably
>        the payee will win because he knows when the assassin will strike.
>        For a lottery to be effective, prospective assassins need to
>        be able to determine how much the jackpot is and who the victim is,
>        so they can place their bets and be the closest winner.

There are a number of advantages to the gambling scenario.  The first, 
obviously, is that it is, technically, gambling, and gambling is legal in 
many locales and "accepted" (morally) in most.  The second is that being 
gambling, no participant other than an assassin knows, for sure, that the 
person collecting the payoff is actually guilty of any crime.  A third is 
that to many people, gambling is fun, which means that they are already 
primed to partipate.


>        But the prospective victim can also play, individually or as part
>        of an insurance pool (which is especially valuable for victims like
>        "the first IRS agent to be assassinated".)  Obviously you don't want
>        to just bid up the price on your own head, so it needs to be 
accompanied
>        by publicity that the IRS Agents' Benevolent Association is placing
>        a large number of small bets every day to maximize the chances that
>        _they_ will collect the money rather than the assassins.
>        If the times that the bets are for are published, you can beat this,
>        but you also invite speculators to be small bets just before and after
>        your bets, so it becomes a mishmash and perhaps a race condition.
>        If the times aren't published, the assassin can make lots of bets
>        surrounding the planned date of the hit, which is also a warning
>        to the prospective victim to be careful if the bets on his demise
>        start increasing rapidly.

The way I've envisioned it, the bettor must commit to his prediction, but he 
then encrypts it so that nobody else knows what he's predicted, either the 
date or the target, and he must "buy" his prediction by including digital 
cash.  One advantage of this system is that the victim isn't warned 
(although the victim is already aware of how much money has been donated to 
fund the prize), but a potential disadvantage is that there might be a 
collision between the predictions of two or more people.  This probably 
can't be entirely avoided, but it can be minimized by forcing the bettors to 
compete with their bets.  It is impractical for a person who just guesses to 
make such a bet, against somebody who knows the date.

The bettor doesn't reveal the decrypt key to the particular prediction 
involved unless he chooses to do so; when he does presumably, it's because 
the prediction has come true.  A failed prediction isn't disclosed, so it 
doesn't alert anyone else, including the  "escrow agent" AP organization.   
The amount of digital cash included with the prediction will be revealed 
publicly, the moment it is received by the organization, as well as the 
(still encrypted) prediction.  Saving these predictions allows anyone to 
verify (after the decrypt key has been released) that the prediction 
actually identified a particular person.  This allows the public and the 
other players to verify that the game is being played honestly.

       
>        This does make the AP lottery somewhat of an extortion deal - 
>        by advertising that someone is a target, you're forcing them to
>        continually make lots of bets. 

This, of course, is only true if the specifics of the predictions are 
revealed, which is not the scenario I envision.  Obviously, revealing the 
predictions allows the victim to lay low whever he's likely to be hit, which 
is why I think that's a bad idea.


>        But if they've got any way of tracing
>        money, even partially, it'll help them find the escrow agent,
>        who can then be targeted for justice of one sort or another.
>        You're also forcing the assassin to make lots of bets, though in
>        a jackpot system the successful payee will recover most of it.

It is at least conceivable that an assassin won't know the exact date he'll 
be successful.  If he sends a letter-bomb, for instance, he won't know for 
sure when that letter will be delivered, or the day it will be opened.  
However, he need merely make two or three predictions, and he won't have to 
make the last one or two unless the prediction didn't come true on the first 
day.  Multiple predictions does raise his cost, but it will still be 
economical to do given the probabilities.


>        To some extent the defense can be fought if the escrow agent
>        wants to establish a minimum bet, say $100, which an assassin
>        can afford to make a few of for the targeted day, but the
>        victim can't keep paying too much.  This also reduces the
>        since of the potential better pool, and therefore reduces
>        the jackpot and the attractiveness of the job to the assassin.
>        Lots of people might be willing to spend $5 to contribute;
>        $100 bets are much fewer, especially if there are enough
>        targets to successfully overthrow a government.

Originally, I had anticipated that the AP organization would require 
specific payments included with a prediction based on the value of the 
reward and the estimated probability of the death on any particular day.  I 
was never comfortable with this system, because it would involve a great 
deal of calculation based on numbers that no particular people know.  I 
don't know whether you saw my idea a couple weeks ago in which I pointed out 
that it should merely be necessary to allow all predictors to include 
whatever amount of money they want with their prediction, with the reward 
split up and paid to all successful predictors, pro-rated based on the size 
of their contribution.  This shifts the burden to the public to estimate the 
amount they should include, which should be okay since a predictor is 
probably a fairly good judge of the probabilities associated with his target.

At least hypothetically, it means that an assassin might have a portion of 
his reward "stolen" by another lucky  predictor, but this is unlikely and in 
any case, the fact that this loss was genuine can be verified by all 
participates after the fact.  And a predictor need merely increase the value 
included with the prediction to increase his pro-rata share of the reward.


Jim Bell
jimbell@pacifier.com