Re: poly: How to buy shared secrets cheap

From: Hal Finney <>
Date: Fri Feb 27 1998 - 10:48:07 PST

Robin Hanson, <>, writes:
> Imagine two people share a secret which would hurt
> them each $1000 worth if it got out. You offer to pay them each
> $1 to (verifiably) tell you their secret. If this is a one-shot simultaneous
> game, there are two pure-strategy equilibria: one where they both tell
> and another where neither of them tell. But since the no-tell
> equilibria makes them both better off, your chances aren't good.
> What if, instead, you commit to paying only the first person who
> tells you, and to paying more the longer you have to wait? You'll pay
> $2000 a year from now, but you won't pay anything after that. Now
> they both know they will each tell a year from now, if no one has told
> before then. But then they each would tell you the day before, to be
> first, for only
> $1990. Following this reasoning back day by day, they would each
> tell you the very first day, even if you only offered them $1!
> Shared secrets can be bought very cheaply.

This is an astonishing result! It would seem to be a way to get a
confession in any crime where more than one person was involved, which
is a lot of them.

Two people rob a bank. To get them to confess, the police apply Robin's
strategy. How much does confessing hurt the robbers? Suppose they
lose the $10,000 they stole and have to go to prison for 10 years.
$1 million ought to cover it. The police offer increasing amounts as
Robin describes, culminating at $1 million for a verifiable confession
(to verify it, the robbers could reveal information which only the true
perpetrators would have known). A bit of a strain on the police budget,
but actually once Robin's reasoning is explained to the robbers, they
both confess on the first day for a paltry sum.

Apply the same reasoning to any corporate or political conspiracy as well.
Want to know what really happened between Bill and Monica? Sounds like
it ought to be cheap.

Received on Fri Feb 27 19:09:08 1998

This archive was generated by hypermail 2.1.8 : Tue Mar 07 2006 - 14:45:30 PST