Analysis of Causal Loop Diagrams

["David N. Ford"]

The paper "A Behavioral Approach to Feedback Loop Dominance Analysis"
will be published in the next issue of the System Dynamics Review. An
important distinction must made and kept clear between feedback loop
analysis, which the paper addresses, and causal loop diagrams, which the
original inquiry to this list addresses. Researchers as notable of Doerner
have suggested that topographic features of systems such as those mentioned
in the inquiry can suggest structural dominance. However I am not aware of
any rigorous or reliable method of analyzing systems for loop dominance
based only on causal loop diagrams. Causal loop diagrams do not describe
structure specificlly enough for quantitative dominance analysis. As an
example take any of the causal loop diagram descriptions of the system
archetypes and ask "When and under what conditions do each of these loops
dominate?". We typically answer "Well, it depends..." and then begin to
describe the structure and its relation to the behavior more specificlly.

My comments should not be interpreted as degrading the qualitative analysis
of loop dominance (in our program I give a lecture introducing the concept
of loop dominance using qualitative analysis before formal modeling skills
are developed), only that qualitative analysis is severly limited. As for
how we can use system behavior to analyze feedback loops more
rigorously...youll have to read the paper.

David Ford
System Dynamics Program
University of Bergen
From: "David N. Ford" <David.Ford@ifi.uib.no>

[Alexandre J G P Rodrigues]

Dear Guenter,

On a practical perspective, qualitative understanding of a systems =
feedback structure is important in various ways. This understanding can =
and should be developed from a qualitative analysis exercise, prior to =
quantification.

In practice, it is important because, if properly done, it provides a =
number of advantages over just quantitative simulation:
(1) interpreting results: it provides an excellent high-level view of =
the underlying dynamics of the system, to which one can later relate =
when trying to interpret the quantitative results;
(2) identifying solutions: it is an excellent means to devise stategic =
solutions to problems, which can then be tested in the simulation model. =
For example, the presence of a vicious circle known to be dominant =
calls for an action to eleminate, attenuate or invert its influence over =
the system;
(3) driving as opposed to being driven by the simulation: it helps the =
modeller to keep in charge of the analysis, as opposed to the simulation =
model driving the modeller. For this, one needs to know when to stop =
moving in a certain direction and what new direction should be taken. =
And for this, the modeller needs to have in mind at all times the =
essence of the systems feedback structure, which can be developed with =
qualitative analysis.

Of course, qualitative analysis can be done in a more or less structure =
way. But formal qualitative analysis is not the same as quantitative =
analysis through simulation modelling. A formal process of qualitative =
analysis should be able to guide the modeller/analyst through a series =
of steps and procedures that ultimately lead to valid conclusions about =
the "whats"and "whys" of the system behaviour. The process may imply =
some type of calculations (e.g. counting loops) but not quantitative =
simulation. Studies or theories about loop dominance which use =
calculations can be seen as "tools" that support this qualitative =
process; but the ones that use simulation will fall in the field of =
quantitative analysis and address the critical issue of formally =
relating model structure to model behaviour.

In the future, one could see both (1) formal methods of qualitative =
analysis and (2) formal methods of relating model structure to model =
behaviour, as making the formal bridge between qualitative and =
quantitative SD within a single, complete and fully formalised =
SD-process. Until then, both are areas of on-going research.

Regards,

Alexandre Rodrigues

_________________________________
Alexandre J G P Rodrigues

Departamento de Sistemas de Informa=E7=E3o
Escola de Engenharia
Universidade do Minho
4800 Guimar=E3es
Portugal, EU

Department of Information Systems
The School of Engineering
University of Minho
4800 Guimar=E3es
Portugal, EU

Tel.: +351 (0)53 510 149
Fax.: +351 (0)53 510 250
Email: Alex.Rodrigues@dsi.uminho.pt
Web: http://www.dsi.uminho.pt/~alex/