Friam

Bonabeau article

Classic

List

Threaded

6 messages Options

Friam mailing list

Bonabeau article

This is a multi-part message in MIME format.

------=_NextPart_000_0002_01C311AE.E8194DE0
Content-Type: text/plain;
charset="us-ascii"
Content-Transfer-Encoding: 7bit

The May 2003 Harvard Business Review has an article by Eric Bonabeau "Don't
Trust your Gut". It describes how the latest decision-support tools
(agent-based model, genetic programs, the usual suspects) are best used to
complement (rather than replace) executives' intuitions.

I've got a PDF version of this article so if anyone wants a copy, email me.

Robert

------=_NextPart_000_0002_01C311AE.E8194DE0
Content-Type: text/html;
charset="us-ascii"
Content-Transfer-Encoding: quoted-printable

<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN">
<HTML>
<HEAD>
<META HTTP-EQUIV=3D"Content-Type" CONTENT=3D"text/html; =
charset=3Dus-ascii">
<META NAME=3D"Generator" CONTENT=3D"MS Exchange Server version =
6.0.4630.0">
<TITLE>Bonabeau article</TITLE>
</HEAD>
<BODY>


<P><FONT SIZE=3D2 FACE=3D"Trebuchet MS">The May 2003 Harvard Business =
Review has an article by Eric Bonabeau "Don't Trust your Gut". =
It describes how the latest decision-support tools (agent-based model, =
genetic programs, the usual suspects) are best used to complement =
(rather than replace) executives' intuitions.</FONT></P>

<P><FONT SIZE=3D2 FACE=3D"Trebuchet MS">I've got a PDF version of this =
article so if anyone wants a copy, email me.</FONT>
</P>

<P><FONT SIZE=3D2 FACE=3D"Trebuchet MS">Robert</FONT>
</P>

</BODY>
</HTML>
------=_NextPart_000_0002_01C311AE.E8194DE0--

Friam mailing list

Bonabeau article

Yes, please!

On Saturday, May 3, 2003, at 08:02 PM, Robert Holmes wrote:

> The May 2003 Harvard Business Review has an article by Eric Bonabeau
> "Don't Trust your Gut". It describes how the latest decision-support
> tools (agent-based model, genetic programs, the usual suspects) are
> best used to complement (rather than replace) executives' intuitions.
>
> I've got a PDF version of this article so if anyone wants a copy,
> email me.
>
> Robert
>
>

Owen Densmore 451 Camino Don Miguel Santa Fe, NM 87505
Work: 505-983-6305 Cell: 505-570-0168 Home: 505-988-3787
[hidden email] http://backspaces.net
http://complexityworkshop.com

Friam mailing list

Bonabeau article

Me too..

----- Original Message -----
From: "Owen Densmore" <[hidden email]>
To: <[hidden email]>
Sent: Sunday, May 04, 2003 7:54 AM
Subject: Re: [FRIAM] Bonabeau article

> Yes, please!
>
> On Saturday, May 3, 2003, at 08:02 PM, Robert Holmes wrote:
>
> > The May 2003 Harvard Business Review has an article by Eric Bonabeau
> > "Don't Trust your Gut". It describes how the latest decision-support
> > tools (agent-based model, genetic programs, the usual suspects) are
> > best used to complement (rather than replace) executives' intuitions.
> >
> > I've got a PDF version of this article so if anyone wants a copy,
> > email me.
> >
> > Robert
> >
> >
> Owen Densmore 451 Camino Don Miguel Santa Fe, NM 87505
> Work: 505-983-6305 Cell: 505-570-0168 Home: 505-988-3787
> [hidden email] http://backspaces.net
> http://complexityworkshop.com
>
>
> =========================================================
> FRIAM Complexity Coffee listserv
> Meets Fridays 9AM @ Museum Hill Cafe
> Archives, unsubscribe, etc.:
> http://www.redfish.com/mailman/listinfo/friam_redfish.com
>
>

Friam mailing list

Bonabeau article

In reply to this post by Friam mailing list

Me too....
----- Original Message -----
From: "Owen Densmore" <[hidden email]>
To: <[hidden email]>
Sent: Sunday, May 04, 2003 7:54 AM
Subject: Re: [FRIAM] Bonabeau article

Friam mailing list

Bonabeau article

In reply to this post by Friam mailing list

This is a multi-part message in MIME format.

------=_NextPart_000_0125_01C312F9.0D60F300
Content-Type: text/plain;
charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable

Bonabeau articleIntuition is inevitable!!!

=20

Even as I agree with Bonabeau that more informed analysis makes for =
better understanding of the problem (hence better decisions and better =
"intuition"), the problem of modeling associated intuition remains. =20

=20

Any model is not a full representation of reality on a computer, but =
some approximation of that reality in way that captures essential =
features of the phenomenon in question. Model building is an art where =
one exercises judgment and intuition in selecting which problem features =
are relevant. When I am building a model of swinging pendulum, I apply =
Newton's first and second law, but not his third, Maxwell's laws, the =
laws of heat transfer, relativity or quantum mechanics even though they =
due play a part (albeit an insignificant one to the questions I am =
trying to answer). I can make well justified choices about which =
physical laws are relevant to my model. I may also be able to estimate =
the errors incurred by ignoring those other laws. =20

=20

Agent bases models with their tightly coupled nonlinear interaction =
amplify those modeling errors and choices. Furthermore, in the absence =
of a finite set of possible choices of model features (like the physical =
laws in the above example), the modeler relies to a much greater degree =
on his intuition. He has to make many decisions in the absence of =
essential information and an understanding of what he maybe ignoring. I =
am not even aware if somebody has tried to even suggest a methodology =
for building ABMs for businesses. Maybe Ethnographers and =
Anthropologists have something to say here. Often in a consulting =
situation were ABMs are built, very little model verification is carried =
out. A disaster might be in the making if important strategic decisions =
are made based on unverified ABMs.=20

=20

Optimizers also often exploit modeling errors. When I was starting my =
PhD optimizing wing shapes for BAE systems, my GA would often produce =
wing shapes seemed optimal yet they were right of the boundary of where =
the CFD code is known to produce unreliable results (I was using VSAero =
panel code at the time). I have seen this often phenomena repeated in =
other engineering design problems. Often, if you had a new simulation =
model that you want to test to breaking point, run it through an =
optimizer.=20

=20

In the absence of sound methodologies for modeling business process and =
of verification tools, intuition will continue to play a very important =
part in model building. Criticisms about putting too much faith in =
intuition will also hold for such sophisticated computer models. One can =
even argue that intuitively built computer models can be far more =
dangerous when combined with false sense of security and a lack of =
understanding of the model limitations.=20

=20

Mohammed=20

__________________________________________________
[hidden email] http://www.optomatica.com =20
mobile: +20 (10) 191 6650

----- Original Message -----=20
From: Robert Holmes=20
To: [hidden email]=20
Sent: Sunday, May 04, 2003 5:02 AM
Subject: [FRIAM] Bonabeau article

The May 2003 Harvard Business Review has an article by Eric Bonabeau =
"Don't Trust your Gut". It describes how the latest decision-support =
tools (agent-based model, genetic programs, the usual suspects) are best =
used to complement (rather than replace) executives' intuitions.

I've got a PDF version of this article so if anyone wants a copy, =
email me.=20

Robert=20

------=_NextPart_000_0125_01C312F9.0D60F300
Content-Type: text/html;
charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable

<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN">
<HTML><HEAD><TITLE>Bonabeau article</TITLE>
<META http-equiv=3DContent-Type content=3D"text/html; =
charset=3Diso-8859-1">
<META content=3D"MSHTML 6.00.2726.2500" name=3DGENERATOR>
<STYLE></STYLE>
</HEAD>
<BODY bgColor=3D#ffffff>
<DIV><FONT face=3DArial size=3D2>
<P class=3DMsoNormal style=3D"MARGIN: 0in 0in 0pt"><SPAN=20
style=3D"FONT-SIZE: 10pt; FONT-FAMILY: Arial">Intuition is=20
inevitable!!!<?xml:namespace prefix =3D o ns =3D=20
"urn:schemas-microsoft-com:office:office" /><o:p></o:p></SPAN></P>
<P class=3DMsoNormal style=3D"MARGIN: 0in 0in 0pt"><SPAN=20
style=3D"FONT-SIZE: 10pt; FONT-FAMILY: =
Arial"><o:p> </o:p></SPAN></P>
<P class=3DMsoNormal style=3D"MARGIN: 0in 0in 0pt"><SPAN=20
style=3D"FONT-SIZE: 10pt; FONT-FAMILY: Arial">Even as I agree with =
Bonabeau that=20
more informed analysis makes for better understanding of the problem =
(hence=20
better decisions and better =93intuition=94), the problem of modeling =
associated=20
intuition remains. <SPAN=20
style=3D"mso-spacerun: yes"> </SPAN><o:p></o:p></SPAN></P>
<P class=3DMsoNormal style=3D"MARGIN: 0in 0in 0pt"><SPAN=20
style=3D"FONT-SIZE: 10pt; FONT-FAMILY: =
Arial"><o:p> </o:p></SPAN></P>
<P class=3DMsoNormal style=3D"MARGIN: 0in 0in 0pt"><SPAN=20
style=3D"FONT-SIZE: 10pt; FONT-FAMILY: Arial">Any model is not a full=20
representation of reality on a computer, but some approximation of that =
reality=20
in way that captures essential features of the phenomenon in question. =
Model=20
building is an art where one exercises judgment and intuition in =
selecting which=20
problem features are relevant. When I am building a model of swinging =
pendulum,=20
I apply Newton=92s first and second law, but not his third, Maxwell=92s =
laws, the=20
laws of heat transfer, relativity or quantum mechanics even though they =
due play=20
a part (albeit an insignificant one to the questions I am trying to =
answer). I=20
can make well justified choices about which physical laws are relevant =
to my=20
model. I may also be able to estimate the errors incurred by ignoring =
those=20
other laws. <SPAN style=3D"mso-spacerun: yes"> </SPAN><SPAN=20
style=3D"mso-spacerun: yes"> </SPAN><o:p></o:p></SPAN></P>
<P class=3DMsoNormal style=3D"MARGIN: 0in 0in 0pt"><SPAN=20
style=3D"FONT-SIZE: 10pt; FONT-FAMILY: =
Arial"><o:p> </o:p></SPAN></P>
<P class=3DMsoNormal style=3D"MARGIN: 0in 0in 0pt"><SPAN=20
style=3D"FONT-SIZE: 10pt; FONT-FAMILY: Arial">Agent bases models with =
their=20
tightly coupled nonlinear interaction amplify those modeling errors and =
choices.=20
Furthermore, in the absence of a finite set of possible choices of model =

features (like the physical laws in the above example), the modeler =
relies to a=20
much greater degree on his intuition. He has to make many decisions in =
the=20
absence of essential information and an understanding of what he maybe =
ignoring.=20
I am not even aware if somebody has tried to even suggest a methodology =
for=20
building ABMs for businesses. Maybe Ethnographers and Anthropologists =
have=20
something to say here. Often in a consulting situation were ABMs are =
built, very=20
little model verification is carried out. A disaster might be in the =
making if=20
important strategic decisions are made based on unverified ABMs.=20
<o:p></o:p></SPAN></P>
<P class=3DMsoNormal style=3D"MARGIN: 0in 0in 0pt"><SPAN=20
style=3D"FONT-SIZE: 10pt; FONT-FAMILY: =
Arial"><o:p> </o:p></SPAN></P>
<P class=3DMsoNormal style=3D"MARGIN: 0in 0in 0pt"><SPAN=20
style=3D"FONT-SIZE: 10pt; FONT-FAMILY: Arial">Optimizers also often =
exploit=20
modeling errors. When I was starting my PhD optimizing wing shapes for =
BAE=20
systems, my GA would often produce wing shapes seemed optimal yet they =
were=20
right of the boundary of where the CFD code is known to produce =
unreliable=20
results (I was using VSAero panel code at the time). <SPAN=20
style=3D"mso-spacerun: yes"> </SPAN>I have seen this often =
phenomena repeated=20
in other engineering design problems. Often, if you had a new simulation =
model=20
that you want to test to breaking point, run it through an optimizer.=20
<o:p></o:p></SPAN></P>
<P class=3DMsoNormal style=3D"MARGIN: 0in 0in 0pt"><SPAN=20
style=3D"FONT-SIZE: 10pt; FONT-FAMILY: =
Arial"><o:p> </o:p></SPAN></P>
<P class=3DMsoNormal style=3D"MARGIN: 0in 0in 0pt"><SPAN=20
style=3D"FONT-SIZE: 10pt; FONT-FAMILY: Arial">In the absence of sound=20
methodologies for modeling business process and of verification tools, =
intuition=20
will continue to play a very important part in model building. =
Criticisms about=20
putting too much faith in intuition will also hold for such =
sophisticated=20
computer models. One can even argue that intuitively built computer =
models can=20
be far more dangerous when combined with false sense of security and a =
lack of=20
understanding of the model limitations. <o:p></o:p></SPAN></P>
<P class=3DMsoNormal style=3D"MARGIN: 0in 0in 0pt"><SPAN=20
style=3D"FONT-SIZE: 10pt; FONT-FAMILY: =
Arial"><o:p> </o:p></SPAN></P>
<P class=3DMsoNormal style=3D"MARGIN: 0in 0in 0pt"><SPAN=20
style=3D"FONT-SIZE: 10pt; FONT-FAMILY: Arial">Mohammed </SPAN></P>
<P class=3DMsoNormal style=3D"MARGIN: 0in 0in 0pt"><SPAN=20
style=3D"FONT-SIZE: 10pt; FONT-FAMILY: =
Arial">__________________________________________________<BR><A=20
href=3D"mailto:[hidden email]">[hidden email]</A> =
<A=20
href=3D"http://www.optomatica.com">http://www.optomatica.com</A>  =
<BR>mobile:=20
+20 (10) 191 6650</SPAN></P></FONT></DIV>
<BLOCKQUOTE dir=3Dltr=20
style=3D"PADDING-RIGHT: 0px; PADDING-LEFT: 5px; MARGIN-LEFT: 5px; =
BORDER-LEFT: #000000 2px solid; MARGIN-RIGHT: 0px">
<DIV style=3D"FONT: 10pt arial">----- Original Message ----- </DIV>
<DIV=20
style=3D"BACKGROUND: #e4e4e4; FONT: 10pt arial; font-color: =
black"><B>From:</B>=20
<A title=[hidden email] =
href=3D"mailto:[hidden email]">Robert=20
Holmes</A> </DIV>
<DIV style=3D"FONT: 10pt arial"><B>To:</B> <A =
title=[hidden email]=20
href=3D"mailto:[hidden email]">[hidden email]</A> </DIV>
<DIV style=3D"FONT: 10pt arial"><B>Sent:</B> Sunday, May 04, 2003 5:02 =
AM</DIV>
<DIV style=3D"FONT: 10pt arial"><B>Subject:</B> [FRIAM] Bonabeau =
article</DIV>
<DIV><BR></DIV>
<P><FONT face=3D"Trebuchet MS" size=3D2>The May 2003 Harvard Business =
Review has=20
an article by Eric Bonabeau "Don't Trust your Gut". It describes how =
the=20
latest decision-support tools (agent-based model, genetic programs, =
the usual=20
suspects) are best used to complement (rather than replace) =
executives'=20
intuitions.</FONT></P>
<P><FONT face=3D"Trebuchet MS" size=3D2>I've got a PDF version of this =
article so=20
if anyone wants a copy, email me.</FONT> </P>
<P><FONT face=3D"Trebuchet MS" size=3D2>Robert</FONT>=20
</P></BLOCKQUOTE></BODY></HTML>

------=_NextPart_000_0125_01C312F9.0D60F300--

Friam mailing list

Bonabeau article

This is a multi-part message in MIME format.

------=_NextPart_000_0008_01C3130D.D1F17230
Content-Type: text/plain;
charset="iso-8859-1"
Content-Transfer-Encoding: 8bit

Bonabeau articleMohammed writes:
>>I am not even aware if somebody has tried to even suggest a methodology
for building ABMs for businesses.
>>Often in a consulting situation were ABMs are built, very little model
verification is carried out.

I agree, and would add these thoughts:

If we do not have model creation methodologies and standards by which to
apply them, then there is little upon which to base a verification effort.
All we can ask is, is the model internally consistent and is the
implementation correct. We would like to ask questions about the coupling
of the model to the domain environment and to other models, but cannot. If
we can't do that, then validating the client's data bcomes problematic.

What to do? Here is a very incomplete list of suggestions. They don't get
us there, but they might help set
the stage. Feel free to shoot holes in them or come up with your own list.

1. Be more explicit about ontologies used in the model. Don't rely
exclusively on implementation languages
to do the ontology for you. Some suggestions for ontology languages: DAML,
OWL.

2. Bootstrap instances explicitly from experimental scenarios wherever
possible to increase environmental coupling
and reduce tight internal coupling and reliance on random variables..
Growing parts of the model at runtime makes
assumptions more explicit and helps develop alternatives.

3. We need an accounting package that allows us to track runs, input data
and output results. People need to
actually use it. We did try this several times in the last few years, but
ran into client resistance.

4. We could probably all stand periodic refreshers in Experimental Design
101.

5. Plan for ongoing turbulence in client domain ontology and data sources.
Much time is eaten up pretending that
these things don't change after the initial knowledge acquisition phase and
that comes at the expense of model
and experiment rigor downstream.

6. Judicious client choice. This is of course difficult in this economic
environment, but some clients make it
easier to do good science (and thus increase the reputation of the methods)
by knowing what they want
and being realistic about your expertise and abilities to give it to them.
Others....not so much.
After reading the NYT article this last weekend, I would tend to think major
corporation IT departments are
not on the A list. In any case, there is an ongoing responsibility to
educate clients on what's needed. Without
this, they may read some book on object-oriented programming or UML on the
plane and expect magic.

If we don't start moving down the road to greater professionalism and rigor
in constructing these tools, and selecting
and educating our clients, it will all start seeming like snake oil to
potential customers. We will have no one
to blame but ourselves. We approaching (if we haven't already passed it) a
place where the Quick is seldom
worth the Dirty.

Carl

-----Original Message-----
From: [hidden email] [mailto:[hidden email]]On Behalf Of
Mohammed El-Beltagy
Sent: Monday, May 05, 2003 2:26 AM
To: [hidden email]
Subject: Re: [FRIAM] Bonabeau article

Intuition is inevitable!!!

Even as I agree with Bonabeau that more informed analysis makes for better
understanding of the problem (hence better decisions and better
intuition), the problem of modeling associated intuition remains.

Any model is not a full representation of reality on a computer, but some
approximation of that reality in way that captures essential features of the
phenomenon in question. Model building is an art where one exercises
judgment and intuition in selecting which problem features are relevant.
When I am building a model of swinging pendulum, I apply Newtons first and
second law, but not his third, Maxwells laws, the laws of heat transfer,
relativity or quantum mechanics even though they due play a part (albeit an
insignificant one to the questions I am trying to answer). I can make well
justified choices about which physical laws are relevant to my model. I may
also be able to estimate the errors incurred by ignoring those other laws.

Agent bases models with their tightly coupled nonlinear interaction
amplify those modeling errors and choices. Furthermore, in the absence of a
finite set of possible choices of model features (like the physical laws in
the above example), the modeler relies to a much greater degree on his
intuition. He has to make many decisions in the absence of essential
information and an understanding of what he maybe ignoring. I am not even
aware if somebody has tried to even suggest a methodology for building ABMs
for businesses. Maybe Ethnographers and Anthropologists have something to
say here. Often in a consulting situation were ABMs are built, very little
model verification is carried out. A disaster might be in the making if
important strategic decisions are made based on unverified ABMs.

Optimizers also often exploit modeling errors. When I was starting my PhD
optimizing wing shapes for BAE systems, my GA would often produce wing
shapes seemed optimal yet they were right of the boundary of where the CFD
code is known to produce unreliable results (I was using VSAero panel code
at the time). I have seen this often phenomena repeated in other
engineering design problems. Often, if you had a new simulation model that
you want to test to breaking point, run it through an optimizer.

In the absence of sound methodologies for modeling business process and of
verification tools, intuition will continue to play a very important part in
model building. Criticisms about putting too much faith in intuition will
also hold for such sophisticated computer models. One can even argue that
intuitively built computer models can be far more dangerous when combined
with false sense of security and a lack of understanding of the model
limitations.

Mohammed

__________________________________________________
[hidden email] http://www.optomatica.com
mobile: +20 (10) 191 6650

----- Original Message -----
From: Robert Holmes
To: [hidden email]
Sent: Sunday, May 04, 2003 5:02 AM
Subject: [FRIAM] Bonabeau article

The May 2003 Harvard Business Review has an article by Eric Bonabeau
"Don't Trust your Gut". It describes how the latest decision-support tools
(agent-based model, genetic programs, the usual suspects) are best used to
complement (rather than replace) executives' intuitions.

I've got a PDF version of this article so if anyone wants a copy, email
me.

Robert

------=_NextPart_000_0008_01C3130D.D1F17230
Content-Type: text/html;
charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable

<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN">
<HTML xmlns:o =3D =
"urn:schemas-microsoft-com:office:office"><HEAD><TITLE>Bonabeau =
article</TITLE>
<META http-equiv=3DContent-Type content=3D"text/html; =
charset=3Diso-8859-1">
<META content=3D"MSHTML 6.00.2715.400" name=3DGENERATOR>
<STYLE></STYLE>
</HEAD>
<BODY bgColor=3D#ffffff>
<DIV><SPAN class=3D791001818-05052003>Mohammed writes:</SPAN></DIV>
<DIV><SPAN class=3D791001818-05052003></SPAN><SPAN =
class=3D791001818-05052003> =20
>></SPAN>I am not even aware if somebody has tried to even suggest =
a=20
methodology for building ABMs for businesses. </DIV>
<DIV><FONT face=3DArial color=3D#0000ff size=3D2><SPAN =
class=3D791001818-05052003> =20
>><FONT face=3D"Times New Roman" color=3D#000000 size=3D3>Often in =
a consulting=20
situation were ABMs are built, very little model verification is carried =
out.=20
</FONT></SPAN></FONT></DIV>
<DIV><FONT face=3DArial color=3D#0000ff size=3D2><SPAN=20
class=3D791001818-05052003></SPAN></FONT> </DIV>
<DIV><FONT face=3DArial color=3D#0000ff size=3D2><SPAN =
class=3D791001818-05052003>I=20
agree, and would add these thoughts:</SPAN></FONT></DIV>
<DIV><FONT face=3DArial color=3D#0000ff size=3D2><SPAN=20
class=3D791001818-05052003></SPAN></FONT> </DIV>
<DIV><SPAN class=3D791001818-05052003><FONT face=3DArial color=3D#0000ff =
size=3D2>If we=20
do not have model creation methodologies and standards by which to apply =
them,=20
then there is little upon which to base a verification effort.  All =
we can=20
ask is, is the model internally consistent and is the implementation=20
correct.  We would like to ask questions about the coupling of the =
model to=20
the domain environment and to other models, but cannot.   If =
we can't=20
do that, then validating the client's data bcomes=20
problematic.</FONT></SPAN></DIV>
<DIV><SPAN class=3D791001818-05052003><FONT face=3DArial color=3D#0000ff =

size=3D2></FONT></SPAN> </DIV>
<DIV><SPAN class=3D791001818-05052003><FONT face=3DArial color=3D#0000ff =
size=3D2>What=20
to do?  Here is a very incomplete list of suggestions.  They =
don't get=20
us there, but they might help set</FONT></SPAN></DIV>
<DIV><SPAN class=3D791001818-05052003><FONT face=3DArial color=3D#0000ff =
size=3D2>the=20
stage.  Feel free to shoot holes in them or come up with your own=20
list.</FONT></SPAN></DIV>
<DIV><SPAN class=3D791001818-05052003><FONT face=3DArial color=3D#0000ff =

size=3D2></FONT></SPAN> </DIV>
<DIV><SPAN class=3D791001818-05052003><FONT face=3DArial color=3D#0000ff =
size=3D2>1. Be=20
more explicit about ontologies used in the model.  Don't rely =
exclusively=20
on implementation languages</FONT></SPAN></DIV>
<DIV><SPAN class=3D791001818-05052003><FONT face=3DArial color=3D#0000ff =
size=3D2>to do=20
the ontology for you.  Some suggestions for ontology languages: =
DAML,=20
OWL.</FONT></SPAN></DIV>
<DIV><SPAN class=3D791001818-05052003><FONT face=3DArial color=3D#0000ff =

size=3D2></FONT></SPAN> </DIV>
<DIV><SPAN class=3D791001818-05052003><FONT face=3DArial color=3D#0000ff =
size=3D2>2.=20
Bootstrap instances explicitly from experimental scenarios wherever =
possible to=20
increase environmental coupling</FONT></SPAN></DIV>
<DIV><SPAN class=3D791001818-05052003><FONT face=3DArial color=3D#0000ff =
size=3D2>and=20
reduce tight internal coupling </FONT></SPAN><SPAN=20
class=3D791001818-05052003><FONT face=3DArial color=3D#0000ff =
size=3D2>and reliance on=20
random variables..  </FONT></SPAN><SPAN =
class=3D791001818-05052003><FONT=20
face=3DArial color=3D#0000ff size=3D2>Growing parts of the model at =
runtime makes=20
</FONT></SPAN></DIV>
<DIV><SPAN class=3D791001818-05052003><FONT face=3DArial color=3D#0000ff =

size=3D2>assumptions more explicit and helps develop=20
alternatives.</FONT></SPAN></DIV>
<DIV><SPAN class=3D791001818-05052003><FONT face=3DArial color=3D#0000ff =

size=3D2></FONT></SPAN> </DIV>
<DIV><SPAN class=3D791001818-05052003><FONT face=3DArial color=3D#0000ff =
size=3D2>3. We=20
need an accounting package that allows us to track runs, input data and=20
output results.  People need to</FONT></SPAN></DIV>
<DIV><SPAN class=3D791001818-05052003><FONT face=3DArial color=3D#0000ff =

size=3D2>actually use it.</FONT> <FONT face=3DArial color=3D#0000ff =
size=3D2> We=20
did try this several times in the last few years, but ran into client=20
resistance.</FONT></SPAN></DIV>
<DIV><SPAN class=3D791001818-05052003><FONT face=3DArial color=3D#0000ff =

size=3D2></FONT></SPAN> </DIV>
<DIV><SPAN class=3D791001818-05052003><FONT face=3DArial color=3D#0000ff =
size=3D2>4. We=20
could probably all stand periodic refreshers in Experimental Design=20
101.</FONT></SPAN></DIV>
<DIV><SPAN class=3D791001818-05052003><FONT face=3DArial color=3D#0000ff =

size=3D2></FONT></SPAN> </DIV>
<DIV><SPAN class=3D791001818-05052003><FONT face=3DArial color=3D#0000ff =
size=3D2><SPAN=20
class=3D791001818-05052003><FONT face=3DArial color=3D#0000ff =
size=3D2>5. Plan for=20
ongoing turbulence in client domain ontology and data sources.  =
Much time=20
is eaten up pretending that</FONT></SPAN></FONT></SPAN></DIV>
<DIV><SPAN class=3D791001818-05052003><FONT face=3DArial color=3D#0000ff =
size=3D2><SPAN=20
class=3D791001818-05052003>these things don't change after the initial =
knowledge=20
acquisition phase and that comes at the expense of=20
model</SPAN></FONT></SPAN></DIV>
<DIV><SPAN class=3D791001818-05052003><FONT face=3DArial color=3D#0000ff =
size=3D2><SPAN=20
class=3D791001818-05052003>and experiment rigor=20
downstream.</SPAN></FONT></SPAN></DIV>
<DIV><SPAN class=3D791001818-05052003><FONT face=3DArial color=3D#0000ff =
size=3D2><SPAN=20
class=3D791001818-05052003></SPAN></FONT></SPAN> </DIV>
<DIV><SPAN class=3D791001818-05052003><FONT face=3DArial color=3D#0000ff =
size=3D2>6.=20
Judicious client choice.  This is of course difficult in this =
economic=20
environment, but some clients make it</FONT></SPAN></DIV>
<DIV><SPAN class=3D791001818-05052003><FONT face=3DArial color=3D#0000ff =
size=3D2>easier=20
to do good science (and thus increase the reputation of the methods) by =
knowing=20
what they want</FONT></SPAN></DIV>
<DIV><SPAN class=3D791001818-05052003><FONT face=3DArial color=3D#0000ff =
size=3D2>and=20
being realistic about your expertise and abilities to give it to =
them. =20
Others....not so much.  </FONT></SPAN></DIV>
<DIV><SPAN class=3D791001818-05052003><FONT face=3DArial color=3D#0000ff =
size=3D2>After=20
reading the NYT article this last weekend, I would tend to think major=20
corporation IT departments are</FONT></SPAN></DIV>
<DIV><SPAN class=3D791001818-05052003><FONT face=3DArial color=3D#0000ff =
size=3D2>not on=20
the A list.  In any case, there is an ongoing responsibility to =
educate=20
clients on what's needed.  Without</FONT></SPAN></DIV>
<DIV><SPAN class=3D791001818-05052003><FONT face=3DArial color=3D#0000ff =
size=3D2>this,=20
they may read some book on object-oriented programming or UML on the =
plane and=20
expect magic.</FONT></SPAN></DIV>
<DIV><SPAN class=3D791001818-05052003><FONT face=3DArial color=3D#0000ff =

size=3D2></FONT></SPAN> </DIV>
<DIV><SPAN class=3D791001818-05052003><FONT face=3DArial color=3D#0000ff =
size=3D2>If we=20
don't start moving down the road to greater professionalism and rigor in =

constructing these tools, and selecting</FONT></SPAN></DIV>
<DIV><FONT face=3DArial><FONT size=3D2><FONT color=3D#0000ff><SPAN=20
class=3D791001818-05052003>and educating our clients</SPAN><SPAN=20
class=3D791001818-05052003>,  </SPAN></FONT></FONT></FONT><SPAN=20
class=3D791001818-05052003><FONT face=3DArial color=3D#0000ff =
size=3D2>it will all start=20
seeming like snake oil to potential customers.  We will have no=20
one</FONT></SPAN></DIV>
<DIV><SPAN class=3D791001818-05052003><FONT face=3DArial color=3D#0000ff =
size=3D2>to=20
blame but ourselves.  We approaching (if we haven't already passed =
it) a=20
place where the Quick is seldom</FONT></SPAN></DIV>
<DIV><SPAN class=3D791001818-05052003><FONT face=3DArial color=3D#0000ff =
size=3D2>worth=20
the Dirty.</FONT></SPAN></DIV>
<DIV><SPAN class=3D791001818-05052003><FONT face=3DArial color=3D#0000ff =

size=3D2></FONT></SPAN> </DIV>
<DIV><SPAN class=3D791001818-05052003><FONT face=3DArial color=3D#0000ff =

size=3D2>Carl</FONT></SPAN></DIV>
<DIV><SPAN class=3D791001818-05052003><FONT face=3DArial color=3D#0000ff =

size=3D2></FONT></SPAN> </DIV>
<BLOCKQUOTE dir=3Dltr style=3D"MARGIN-RIGHT: 0px">
<DIV class=3DOutlookMessageHeader dir=3Dltr align=3Dleft><FONT =
face=3DTahoma=20
size=3D2>-----Original Message-----<BR><B>From:</B> =
[hidden email]=20
[mailto:[hidden email]]<B>On Behalf Of </B>Mohammed=20
El-Beltagy<BR><B>Sent:</B> Monday, May 05, 2003 2:26 AM<BR><B>To:</B>=20
[hidden email]<BR><B>Subject:</B> Re: [FRIAM] Bonabeau=20
article<BR><BR></FONT></DIV>
<DIV><FONT face=3DArial size=3D2>
<P class=3DMsoNormal style=3D"MARGIN: 0in 0in 0pt"><SPAN=20
style=3D"FONT-SIZE: 10pt; FONT-FAMILY: Arial">Intuition is=20
inevitable!!!<o:p></o:p></SPAN></P>
<P class=3DMsoNormal style=3D"MARGIN: 0in 0in 0pt"><SPAN=20
style=3D"FONT-SIZE: 10pt; FONT-FAMILY: =
Arial"><o:p> </o:p></SPAN></P>
<P class=3DMsoNormal style=3D"MARGIN: 0in 0in 0pt"><SPAN=20
style=3D"FONT-SIZE: 10pt; FONT-FAMILY: Arial">Even as I agree with =
Bonabeau that=20
more informed analysis makes for better understanding of the problem =
(hence=20
better decisions and better =93intuition=94), the problem of modeling =
associated=20
intuition remains. <SPAN=20
style=3D"mso-spacerun: yes"> </SPAN><o:p></o:p></SPAN></P>
<P class=3DMsoNormal style=3D"MARGIN: 0in 0in 0pt"><SPAN=20
style=3D"FONT-SIZE: 10pt; FONT-FAMILY: =
Arial"><o:p> </o:p></SPAN></P>
<P class=3DMsoNormal style=3D"MARGIN: 0in 0in 0pt"><SPAN=20
style=3D"FONT-SIZE: 10pt; FONT-FAMILY: Arial">Any model is not a full=20
representation of reality on a computer, but some approximation of =
that=20
reality in way that captures essential features of the phenomenon in =
question.=20
Model building is an art where one exercises judgment and intuition in =

selecting which problem features are relevant. When I am building a =
model of=20
swinging pendulum, I apply Newton=92s first and second law, but not =
his third,=20
Maxwell=92s laws, the laws of heat transfer, relativity or quantum =
mechanics=20
even though they due play a part (albeit an insignificant one to the =
questions=20
I am trying to answer). I can make well justified choices about which =
physical=20
laws are relevant to my model. I may also be able to estimate the =
errors=20
incurred by ignoring those other laws. <SPAN=20
style=3D"mso-spacerun: yes"> </SPAN><SPAN=20
style=3D"mso-spacerun: yes"> </SPAN><o:p></o:p></SPAN></P>
<P class=3DMsoNormal style=3D"MARGIN: 0in 0in 0pt"><SPAN=20
style=3D"FONT-SIZE: 10pt; FONT-FAMILY: =
Arial"><o:p> </o:p></SPAN></P>
<P class=3DMsoNormal style=3D"MARGIN: 0in 0in 0pt"><SPAN=20
style=3D"FONT-SIZE: 10pt; FONT-FAMILY: Arial">Agent bases models with =
their=20
tightly coupled nonlinear interaction amplify those modeling errors =
and=20
choices. Furthermore, in the absence of a finite set of possible =
choices of=20
model features (like the physical laws in the above example), the =
modeler=20
relies to a much greater degree on his intuition. He has to make many=20
decisions in the absence of essential information and an understanding =
of what=20
he maybe ignoring. I am not even aware if somebody has tried to even =
suggest a=20
methodology for building ABMs for businesses. Maybe Ethnographers and=20
Anthropologists have something to say here. Often in a consulting =
situation=20
were ABMs are built, very little model verification is carried out. A =
disaster=20
might be in the making if important strategic decisions are made based =
on=20
unverified ABMs. <o:p></o:p></SPAN></P>
<P class=3DMsoNormal style=3D"MARGIN: 0in 0in 0pt"><SPAN=20
style=3D"FONT-SIZE: 10pt; FONT-FAMILY: =
Arial"><o:p> </o:p></SPAN></P>
<P class=3DMsoNormal style=3D"MARGIN: 0in 0in 0pt"><SPAN=20
style=3D"FONT-SIZE: 10pt; FONT-FAMILY: Arial">Optimizers also often =
exploit=20
modeling errors. When I was starting my PhD optimizing wing shapes for =
BAE=20
systems, my GA would often produce wing shapes seemed optimal yet they =
were=20
right of the boundary of where the CFD code is known to produce =
unreliable=20
results (I was using VSAero panel code at the time). <SPAN=20
style=3D"mso-spacerun: yes"> </SPAN>I have seen this often =
phenomena=20
repeated in other engineering design problems. Often, if you had a new =

simulation model that you want to test to breaking point, run it =
through an=20
optimizer. <o:p></o:p></SPAN></P>
<P class=3DMsoNormal style=3D"MARGIN: 0in 0in 0pt"><SPAN=20
style=3D"FONT-SIZE: 10pt; FONT-FAMILY: =
Arial"><o:p> </o:p></SPAN></P>
<P class=3DMsoNormal style=3D"MARGIN: 0in 0in 0pt"><SPAN=20
style=3D"FONT-SIZE: 10pt; FONT-FAMILY: Arial">In the absence of sound=20
methodologies for modeling business process and of verification tools, =

intuition will continue to play a very important part in model =
building.=20
Criticisms about putting too much faith in intuition will also hold =
for such=20
sophisticated computer models. One can even argue that intuitively =
built=20
computer models can be far more dangerous when combined with false =
sense of=20
security and a lack of understanding of the model limitations.=20
<o:p></o:p></SPAN></P>
<P class=3DMsoNormal style=3D"MARGIN: 0in 0in 0pt"><SPAN=20
style=3D"FONT-SIZE: 10pt; FONT-FAMILY: =
Arial"><o:p> </o:p></SPAN></P>
<P class=3DMsoNormal style=3D"MARGIN: 0in 0in 0pt"><SPAN=20
style=3D"FONT-SIZE: 10pt; FONT-FAMILY: Arial">Mohammed </SPAN></P>
<P class=3DMsoNormal style=3D"MARGIN: 0in 0in 0pt"><SPAN=20
style=3D"FONT-SIZE: 10pt; FONT-FAMILY: =
Arial">__________________________________________________<BR><A=20
=
href=3D"mailto:[hidden email]">[hidden email]</A> =
<A=20
href=3D"http://www.optomatica.com">http://www.optomatica.com</A>  =

<BR>mobile: +20 (10) 191 6650</SPAN></P></FONT></DIV>
<BLOCKQUOTE dir=3Dltr=20
style=3D"PADDING-RIGHT: 0px; PADDING-LEFT: 5px; MARGIN-LEFT: 5px; =
BORDER-LEFT: #000000 2px solid; MARGIN-RIGHT: 0px">
<DIV style=3D"FONT: 10pt arial">----- Original Message ----- </DIV>
<DIV=20
style=3D"BACKGROUND: #e4e4e4; FONT: 10pt arial; font-color: =
black"><B>From:</B>=20
<A title=[hidden email]=20
href=3D"mailto:[hidden email]">Robert Holmes</A> </DIV>
<DIV style=3D"FONT: 10pt arial"><B>To:</B> <A =
title=[hidden email]=20
href=3D"mailto:[hidden email]">[hidden email]</A> </DIV>
<DIV style=3D"FONT: 10pt arial"><B>Sent:</B> Sunday, May 04, 2003 =
5:02=20
AM</DIV>
<DIV style=3D"FONT: 10pt arial"><B>Subject:</B> [FRIAM] Bonabeau =
article</DIV>
<DIV><BR></DIV>
<P><FONT face=3D"Trebuchet MS" size=3D2>The May 2003 Harvard =
Business Review has=20
an article by Eric Bonabeau "Don't Trust your Gut". It describes how =
the=20
latest decision-support tools (agent-based model, genetic programs, =
the=20
usual suspects) are best used to complement (rather than replace)=20
executives' intuitions.</FONT></P>
<P><FONT face=3D"Trebuchet MS" size=3D2>I've got a PDF version of =
this article=20
so if anyone wants a copy, email me.</FONT> </P>
<P><FONT face=3D"Trebuchet MS" size=3D2>Robert</FONT>=20
</P></BLOCKQUOTE></BLOCKQUOTE></BODY></HTML>

------=_NextPart_000_0008_01C3130D.D1F17230--