[Maspace] Agent based modeling/LUCC misc points

Mon, 2 Dec 2002 12:01:23 -0600

This list is having a great conversation about agent-based modeling and
land-use and land-cover change, how fantastic!  For background to the
discussions, check out the report at
http://www.csiss.org/maslucc/ABM-LUCC.pdf (CSISS is the host of this
discussion list as well).

Stepping back a second, keep in mind that the MAS/LUCC report exists
within a larger land-use and land-cover change (LUCC) context.  The
International Geosphere-Biosphere Program (IGBP) and the International
Human Dimensions of Global Environmental Change Programme (IHDP) guide
and complement research and policy initiatives on global environmental
change.  Among these initiatives is the LUCC Science initiative
(IGBP-IHDP 1995, 1999).  The MAS/LUCC report and the activities it
describes are associated with this effort.  Larger questions about the
goals of LUCC research, the role of social science, models of decision
making, and policy considerations are addressed by the larger framework
and others beside - the report doesn't claim ownership over these
issues.  In particular, agent-based modeling as such does not support
any one decision making model over another, although it is a flexible
means of accommodating multiple theories and some of these theories may
be better than others.  It is up to individual practitioners and
disciplines to wrestle over what they consider to be appropriate.

In terms of the balance between prediction and policy, most MAS/LUCC
researchers are interested in general characteristics of LUCC processes
and how these apply to a particular time and place.  Importantly, theory
development, policy work, and prediction are complementary, so I am not
sure there is a real disagreement here.  It is possible (necessary?) to
use general theories of human behavior (institutional, economic,
sociological, etc.) to project which outcomes are more likely than
others while fully acknowledging the contingent nature of projections.
Gil Pontius is dead on - as he points out, there are a number of serious
issues related to validation in terms of model outcome,  which is in
turn related to issues of  structural integrity, theoretical framework,
policy relevance, and intended use of the model (the MAS/LUCC report
touches on these issues).

In some cases, a projection is sufficiently bounded in space, time, and
statistical stationarity (essential facets of validation) to make
'predictions' but this modeling is also often about exploring possible
scenarios.  As such, a projection can affect the events it is meant to
project, sure, but any modeling effort is necessarily reflexive - that
is one of things we should be looking for as social scientists.  We
should have some sort of impact on reality while being aware of the
pitfalls (e.g., the sometimes rocky history of the modernist endeavor
and international development).  Finally, in terms of external
researchers vs. local knowledge, many of the projects described in the
report are driven by local stakeholders and exemplify the 'tools to
think with' concept.

IGBP-IHDP (1995). Land-Use and Land-Cover Change: Science/Research Plan.
Stockholm and Geneva, IGBP/HDP.

IGBP-IHDP (1999). Land-Use and Land-Cover Change Implementation
Strategy. IGBP Report 46/IHDP Report 10. Stockholm, International
Geosphere-Biosphere Programme and International & Human Dimensions
Programme.

--

Steven M. Manson
Assistant Professor
Department of Geography
University of Minnesota
414 Social Sciences
267 19th Avenue South
Minneapolis, MN 55455 
USA

612 625 4577
612 624 1044 (fax)

www.tc.umn.edu/~manson

-----Original Message-----
From: maspace-admin@'www.csiss.org' [mailto:maspace-admin@'www.csiss.org'] On
Behalf Of maspace-request@'www.csiss.org'
Sent: Friday, November 29, 2002 2:00 PM
To: 
Subject: Maspace digest, Vol 1 #20 - 1 msg

Send Maspace mailing list submissions to

To subscribe or unsubscribe via the World Wide Web, visit
        http://csiss.ncgia.ucsb.edu/mailman/listinfo/maspace
or, via email, send a message with subject or body 'help' to

You can reach the person managing the list at

When replying, please edit your Subject line so it is more specific
than "Re: Contents of Maspace digest..."

Today's Topics:

   1. RE: Agent-Based Models Workshop Report (Robert Gilmore Pontius
Jr.)

--__--__--

Message: 1
From: "Robert Gilmore Pontius  Jr." <>
To: "'McIntosh, Brian '" <>,
   "' '" <>
Subject: RE: [Maspace] Agent-Based Models Workshop Report
Date: Fri, 29 Nov 2002 06:44:39 -0500

Sure, I'd agree to have that as a friendly amendment.

The importance of doing validation is to estimate the maximum confidence
that we should have in an extrapolation into an undetermined future.
During
the validation, we have better information than we have during the
extrapolation, therefore the we should expect the accuracy of the
extrapolation to be at most equal to the accuracy of the validation.
Just this week, I have derived the equations necessary to estimate the
precision for an extrapolation into the future, based on the validation
from
a previous time. The precision diminishes the farther away in time the
extrapolation is made.

I have observed that some modelers have much greater confidence in their
model results than could ever be justified by rigorous validation. I
think
rigorous validation can offer an estimate of the maximum confidence we
should have in an extrapolation from a model. If the model performs
poorly
for the conditions for which we have good information, then we would
think
that the model would perform even worse for conditions that are
completely
new. Therefore, it is important to have an estimate of the maximum
confidence we should have in a model.

On the other note, if the model's extrapolation predicts with precision
an
undesirable future, then humans should intervene to prevent the
extrapolation from coming true. The reason why we do modeling is to
learn to
improve management of the future. I do not believe that the future is
predetermined; it is not simply waiting there for us to attempt to
predict.

Gil

-----Original Message-----
From: McIntosh, Brian
To: 
Sent: 11/26/2002 3:09 AM
Subject: RE: [Maspace] Agent-Based Models Workshop Report

My two (euro) cents worth ...

I wonder if the sentence "our model predicts with an accuracy of X% at a
resolution of Y over a period of Z" might be more accurately stated
(without being inflammatory or denigrating):

"our model predicted state B from state A with an accuracy of X% at a
resolution of Y over the period Z but we are not sure that the model
will maintain the same predictive capability because the system we are
studying is not time invariant, and anyway if we intervene with policy
based on the results of the model we may invalidate it".

So, some fundamental questions might be what constitutes effective
agent-based LUCC policy support, for who and for what reason? 

It is my belief that traditional predictive modelling just won't do at
all, partly because of the properties of socio-natural systems and
partly because stakeholders and policy makers have their own agendas,
methods, language etc. and simply won't accept the notion of an external
researcher coming along with a model stating 'here this is what will
happen under conditions X' (an observation borne of experience).
Facilitating the exploration of possible futures, possible explanations
and possible problems may be much more helpful. Constructionist but
fruitful?  

Facilitating understanding is very important in decision contexts so
model-based research might be more profitably turned towards the
creation of 'learning tools' or 'tools to think with' (ala MIT media
labs and various other groups around the world). Perhaps it might be
that a useful purpose for ABM-LUCC research is to become embedded in
(local? regional? national?) decision-making structures and
organisations, determine how best to support LUCC stakeholders and
provide suitable learning tools for doing so? The French work on
ABM-based role-playing games that I have seen so far looks to hold
excellent potential for this agenda.

Cheers.
Brian McIntosh, School of Water Sciences, Cranfield University.

--__--__--

_______________________________________________
Maspace mailing list

http://csiss.ncgia.ucsb.edu/mailman/listinfo/maspace

End of Maspace Digest