Re: CAMREC: Could we propose an 'Economic Turing Test'?

From: the Campaign for Real Economics (camrec@mmu.ac.uk)
Date: Wed Jul 07 1999 - 09:50:53 BST


Date: Wed, 07 Jul 1999 09:50:53 +0100
Message-Id: <199907070850.JAA21097@alpheratz.cpm.aca.mmu.ac.uk>
To: bogus
From: camrec@mmu.ac.uk (the Campaign for Real Economics)
Subject: Re: CAMREC: Could we propose an 'Economic Turing Test'?
In-Reply-To: <CAMREC: Could we propose an 'Economic Turing Test'?>


Message-Id: <199907070850.JAA21097@alpheratz.cpm.aca.mmu.ac.uk>
To: CAMREC list members <camrec@mmu.ac.uk>
From: the Campaign for Real Economics <camrec@mmu.ac.uk>
Date: Wed, 07 Jul 1999 09:50:53 +0100
Subject: Re: CAMREC: Could we propose an 'Economic Turing Test'?

Edmund Chattoe wrote (in a private correspondance):
> PS like your idea for economic Turing test, but it reveals an oddity in
> economics. What, another one? Strictly speaking, in some trading game or
> whatever, if the only communication is through trades, no meaningful social
> interaction can take place and it would be impossible to distinguish humans
> from computers unless the strategies were exceptionally smart (favoured
> interpretation: human) or dumb (favoured interpretation: human?) I'm
> reminded of the Not The Nine O'Clock News sketch where the Russian chess
> master is playing, for no clear reason, a complete idiot who doesn't even
> know the rules of chess. Each time the idiot makes a move, the master
> thinks: "Ah, the great Ashenazy gambit". or such like. If all you get to
> work with is a string of CDCDCDDC or BUY SELL BUY etc, the data is not
> "thick enough" to support credible arguments for intelligence. (Interesting
> to know if things like adaptation and responsiveness tended to favour
> attribution of humanity.) If, on the other hand, you let players "chat",
> the computer then has to pass not just the economic Turing Test (ETT) but
> the general one. (Best question for failing the ETT: What's your favourite
> food?) I suppose your question could be recast as: Is there an economic
> task of sufficient richness that agents could "converse" about strategies,
> but at the same time of sufficient formality that a computer could be
> programmed to generate plausible responses in that domain. My guess is that
> the humans would cheat and try to "generalise" the domain, knowing that
> humans could cope with this and computers couldn't. (Like foxing ELIZA by
> using a more complex sentence part:
>
> I am scared of mice
> Why are you scared of mice?
>
> I am scared that I am going mad
> Why are you scared that I am going mad?)

I think that this is right, the task needs to be sufficiently formal so
that it does not just become the normal Turing test, but sufficiently
rich to be really testing (e.g. include negotiation or similar).
Prehaps the ETT could be composed of a series of tests? Maybe it should
be something like the game of poker where it is not necessarily an
advantage to communicate anyway.

Of course, the fact that the ETT, necessarily comes close to the TT, and
that pure price-based communication is utterly inadequate for most real
economic activities is part of the point of proposing such a test - it
neatly illustrates the ridiculousness of unreal economics.

Regards.

--------------------------------------------------
Bruce Edmonds,
Centre for Policy Modelling,
Manchester Metropolitan University, Aytoun Bldg.,
Aytoun St., Manchester, M1 3GH. UK.
Tel: +44 161 247 6479 Fax: +44 161 247 6802
http://www.cpm.mmu.ac.uk/~bruce



This archive was generated by hypermail 2b29 : Thu Jan 06 2000 - 17:00:30 GMT