From: the Campaign for Real Economics (
Date: Wed Jul 07 1999 - 13:48:33 BST

Date: Wed, 7 Jul 1999 12:48:33 +0000
Message-Id: <>
To: bogus
From: (the Campaign for Real Economics)
Subject: CAMREC: The ETT
In-Reply-To: <CAMREC: Could we propose an 'Economic Turing Test'?>

Message-Id: <>
To: CAMREC list members <>
From: the Campaign for Real Economics <>
Date: Wed, 7 Jul 1999 12:48:33 +0000
Subject: CAMREC: The ETT

Dear All,

I think Bruce put his finger on the target area for the ETT. Some domain
where agents have an incentive to communicate, but not to do so without
reservation, such that "computer generated" messages would not seem
implausible and chat messages would tend not to arise. (Coalitions for
trading perhaps? Something like Diplomacy? Would it help if there was a
danger of messages being intercepted? Maybe making them costly would do?)

One possibility is just to make it a rule of the game that all messages are
subject to "censorship" by the experiment controllers, but that seems very
interventionist. Another way of achieving this would not be to allow free
text, but a standard menu of signals where richness could be controlled.

Two random thoughts that may or may not be related:

1) In talking to people who buy records from me on the web, the purpose of
the talk is only indirectly to deal with price, terms of payment and so on.
We also, perhaps implicitly, reason from general details to expectations of
"honesty" (since there is no enforcement): rapid response time, failure to
respond, message bounces, inconsistency of comments, "type" (Would a ripoff
merchant bother or be able to represent as a music lover?) This sort of
talk probably also takes place in other economic interactions - to
establish bargaining credibility, likely honesty in delivery dates and so
on. In fact, if it didn't bargaining could never take place - how would
bargainers establish one position rather than another if their talk had no
relevant content?

2) One reason why impersonation on the web is rare is that context is very
important. The incongruity of a message not generated by the same person -
in the context of previous discussion - is typically so high that we spot
it almost immediately. Computers are very bad at context. Thus even within
a strictly economic domain, if you could find one, you could fox a computer
by referring to previous events in the domain.



Edmund Chattoe: Research Fellow, Department of Sociology, University of
Surrey, Guildford, GU2 5XH, tel: 01483-876988 fax: 01483-259551, Associate
Director, Centre for Research on Simulation in the Social Sciences (CRESS,, Review Editor, J. Artificial
Societies & Social Simulation

This archive was generated by hypermail 2b29 : Thu Jan 06 2000 - 17:00:31 GMT