Aggregation and Behavior

This forum contains all archives from the SD Mailing list (go to http://www.systemdynamics.org/forum/ for more information). This is here as a read-only resource, please post any SD related questions to the SD Discussion forum.
Locked
Francisco Perez francisco77pp ya
Junior Member
Posts: 18
Joined: Fri Mar 29, 2002 3:39 am

Aggregation and Behavior

Post by Francisco Perez francisco77pp ya »

Posted by Francisco Perez <francisco77pp@yahoo.com>
HI All,

I was wondering if any of you have ever questioned the risking of SD modeling when aggregating in order to model the macro system and not to dive into much detail. If Chaos theory is correct, than specific events, as minimal and detail-world they might seem, could generate /influence macro behaviors of higher level systems. I wonder if when we model and aggregate subsystems we may be missing something important, so important that that missing thing could lead to a totally different beheviour if encluded/excluded from the model.

Examples of this are triggering events, unstable or potentially unstable parts of subsystems, domino effects, etc.

Sorry if the Query is somewhat vague, but the previous topic on Emergence and some reflections that I've been carrying in my mind about the capability of ST&SD to aggregate and still preserve real system behavior.

Regards
Francisco
Posted by Francisco Perez <francisco77pp@yahoo.com>
posting date Thu, 1 Dec 2005 18:18:41 -0800 (PST)
Geoff McDonnell gmcdonne bigpond
Junior Member
Posts: 10
Joined: Fri Mar 29, 2002 3:39 am

Aggregation and Behavior

Post by Geoff McDonnell gmcdonne bigpond »

Posted by ""Geoff McDonnell"" <gmcdonne@bigpond.net.au> Francisco I have also been grappling with this problem in health systems simulation for many years. My way of handling it is this: Most people consider SD as a highly aggregated view of the system. However, we are really trying to understand a specific ""problem behavior of interest"" and must extract the ""relevant essence"" of the system structure (say stocks, flows of things and information, feedback interactions, decision rules or
policies) to explain this behavior. This requires abstracting the context and detail relevant to the problem. Barry Richmond best described this as ""bifocal vision"", IMHO. Some parts of the model may be highly aggregated, others disaggregated, depending on their relevance to the problem at hand. We can aggregate space, time, stocks and decision rules (the general pattern of a stream of decisions). However if we observe how individuals make decisions, we may not be able to agree on an aggregate pattern of decision rules (ie there are no agreed policies that people adhere to in making individual decisions -- this is common in health systems problems), and so there is a hierarchy of decision rules (thanks to Sarah Metcalfe and Mark Paich for explaining this to me). Decision rules or ""rules of engagement"" among individuals produce aggregated patterns of behaviour at higher levels in the hierarchy. The patterns are represented by difference equations while the individual behaviours may often be better represented by Agent based modeling or complex adaptive systems conceptualisations. I now tend
to call this ""emergent system dynamics"" (sd rather then SD since
accumulations are less obvious)
You may wish to look at some material from a ISDC2005 Boston workshop on Agent based modeling:why bother? at http://www.xjtek.com/files/papers/abmod ... er2005.pdf
(Sorry, the proceedings link to more workshop material is now broken ) HTH Posted by ""Geoff McDonnell"" <gmcdonne@bigpond.net.au> posting date Sat, 3 Dec 2005 03:40:20 +1100
Jean-Jacques Laublé jean-jacques
Senior Member
Posts: 68
Joined: Fri Mar 29, 2002 3:39 am

Aggregation and Behavior

Post by Jean-Jacques Laublé jean-jacques »

Posted by =?iso-8859-1?Q?Jean-Jacques_Laubl=E9?= <jean-jacques.lauble@wanadoo.fr> Hi Francisco There has already been in the past discussions on aggregation. See the forum http://www.ventanasystems.co.uk/forum/
and make a research on 'aggregat' on the SD list archives.

About your specific question, if there is a sub-system having been aggregated and developing a particular behaviour that is not taken into account by the aggregated system, it may be because it should not have been integrated at first, being probably very different from the other aggregated sub-systems.

Normally when integrating values one should be aware that these aggregated values apart from having a different value, have the same kind of action in the system.

Otherwise you take the risk of just neglecting a more of less important factor in the system.

There is another problem with aggregation: using average data you get average results. And if what you are looking for is an average result, it's okay, but otherwise you cannot get detailed results from average data.

Example: If you want to optimize the pricing of two products A and B that sell towards 20 and 40 each, and aggregate the model you may find that the optimum price is 30. But it will be completely inappropriate and if used dangerous. If you sell both products 30, you will sell a lot of products B and loose money every time you sell one and sell no products A. Regards. J.J. Laublé Allocar

Strasbourg France
Posted by =?iso-8859-1?Q?Jean-Jacques_Laubl=E9?= <jean-jacques.lauble@wanadoo.fr> posting date Fri, 2 Dec 2005 15:13:20 +0100
Locked