Dear All,

I think on one level, it is impossible to get away from organisms tending to
seek "the good" and avoid "the bad". If nothing else, it might be a matter
of definition that what we would avoid, left to ourselves, is "bad" in some
sense.

But I think there is plenty more mileage in identifying _in what sense_ we
seek the good.

1) It cannot be equated with what we do, as some economists have claimed,
because then it really does become an empty truism. (There remains the
possibility that, if utilitarian good is defined in the long term, agents
may act in the short term in ways that they do not prefer, but there don't
seem to be any models like this, possibly because the utility concept is so
formalistic that there isn't much point in speculating about whether
disutility provokes different behaviour patterns than utility does. If one
must labour in the salt mines for 10 years to win the hand of a prospective
spouse, does one reconsider after 5 years of misery? If one is "rational"
and utilities are correctly known at the outset, the decision becomes
indivisble and unshakeable will effectively comes free!)

2) We are happy to accept that, in some cases, expert knowledge about the
"good" can outweigh individual knowledge. This is true both for technical
experts (like doctors) and paternalism (with regard to children and some
others). And yet, mysteriously, the preferences of adult economic agents are
typically assumed to be fixed and incorrigible. (In practice we recognise
that being "right" does not necessarily imply permission to interfere with
someone's autonomy. Economics could be politically wedded to the maintenance
of autonomy without being obliged to maintain incorrigible preference.) If
we learn our preferences, either individually, or socially, it is much
clearer why a) they should exhibit regularities, b) how we can come to be
"wrong" about what we like, either by faulty inference or because our tastes
do change endogenously (pregnancy may result in some foods inducing nausea
when they were always liked before) and c) how people can come to do what
they do not like. (In trying a new food, one can have perfectly good reasons
for expecting to like it and still not.)

3) Economic agents are disembodied so even if we try to make preferences
"meaningful", we cannot. Only in the context of an agent with a physical
body can "hunger" be given a sensible interpretation. (You can try and
specify the patterns of different needs in the structure of the utility
function, but this remains arbitrary and presumes comparability between
needs.) Simulated agents learn to deal with internal signals in the same way
that they learn to deal with external ones and can be said to develop
preferences as they are said to develop behaviours. (But in the same way
that behavioural responses to complex environments can be abstracted to
develop plans and coping strategies, needs and preferences may be "managed"
by progressive induction of relationships between them. Drinking will
assuage hunger, at least briefly, but one must learn this.) The data
generation processes which the agents must learn can be designed by the
simulator (how do we think "hunger" impacts on us and our behaviour?) but we
can triangulate our "introspective" knowledge of wants and needs with
external behaviours in humans and agents.

4) Even if we accept quite "traditional" views of preference, we can still
reject subsidiary assumptions about utility in simulation. Suppose hunger
and thirst are only imperfectly comparable. What happens if agents choose on
the basis of lexicographic preferences? How will agents function if they
have joint preference functions with family? Many of these possibilities
have been rejected "theoretically" or rhetorically, because they make
analytical solution impossible? Computational variants may be able to
sidestep these prob

I suppose the upshot of this message is that there are some ways that social
simulation can help us to deconstruct the rather widespread "cost benefit"
view of social action by implementing agents with different kinds of
motivational architecture. The main difficulty is that economics often mixes
axioms with empirical claims to the benefit of neither. If "cost" and
"benefits" are not to be mere truisms, what sort of processes (simulated or
real) will they have to emerge from?

ATB,

Edmund
==========================================================================
Edmund Chattoe:    Research Fellow, Department of Sociology, University of
Surrey, Guildford, GU2 5XH,  tel: 01483-300-800 extension 3005,  Associate 
Director,  Centre for Research on Simulation in the Social Sciences (CRESS
http://www.soc.surrey.ac.uk/research/cress),  Review Editor, J. Artificial
Societies & Social Simulation http://www.soc.surrey.ac.uk/JASSS/JASSS.html
==========================================================================



%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%