The SD Modeling/Learning Process

From SystemsWiki

Revision as of 03:14, 22 September 2017 by Phoenix (Talk | contribs)
(diff) ← Older revision | Current revision (diff) | Newer revision → (diff)
Jump to: navigation, search


The SD Modeling Process


SD Modelling is a process of learning and persuading, which produces a useful model. The purpose of the model might be to be more confident we are focusing on the right problem, to persuade ourselves and others to implement the right solution in the right way, or to check that our intended actions or policies do not make things worse in the future. The quality of the model produced is judged by its adequacy and suitability for a particular purpose. It is a tool for helping us think more clearly, and involves abductive, deductive and inductive inference. During the process we represent and test theoretical concepts relevant to a problem, play out the consequences of mental model assumptions and search for mechanisms to explain puzzling general regularities. The modeling process can be divided into a spiral of phases with different functions, activities and methods.

The system dynamics methodology has been well described by many people, including Forrester at wikiSD

Applying the Systems Approach

This is a more general wider view of the method, including more detail on Qualitative Conceptual Modeling including Boundary Critique.

Steps in the Modeling Process

    • Figure from a presentation by Hasmiller Lich based on a slide by Jack Homer
  • Formulate Problem & Set Objectives
    • Problem or issue formulation as a puzzling behaviour over time (reference mode)
  • Build and Test Conceptual Model of Explicit System Theories and Assumptions
    • Developing a dynamic hypothesis or structure to explain the puzzle, based on participants mental models and knowledge of system structure and behaviour.
    • Representing these constructs as explicit maps that can be converted into computational models using appropriate tools
  • Translate into Computer Model
  • Perform Virtual experiments and Test against real System Results
    • Calibrating and testing the model hypotheses against data and observations
    • Extensive policy analysis and design to experiment with changing structures and policies and interventions.
    • A series of experiments and presentations to persuade people to act effectively.

Evaluating the model

  • dimensional consistency
  • correspondence
  • general pattern
  • historical fit
  • extreme conditions
  • sensitivity

When the model is fit for purpose, we stop. This may need several iterations around the steps described, including backtracking, new data collection, new hypotheses and new representations to convince different audiences, using metaphors they are familiar with (See Richmond, Zagonel (iterative nature) Lich from Homer for health) Further, in real world applications, models should be developed so as to achieve client confidence and acceptance (using various approaches including "group model building").

SD Competencies

Competency is defined as “knowing-to-act by successfully combining internal and external resources in a family of situations" Based on Bloom's taxonomy, to successful learn SD we need to progress through the stages of knowing, understanding, applying, analysing, creating and evaluating SD modelling skills.

  • Fundamental modes of dynamic behaviour linked to generic cause effect feedback interactions (balancing and re-inforcing loops)
  • Leverage Point and Video

System Dynamics and the Philosophy of Science

Structure Agency Theory

David Lane places System Dynamics in the general structure agency approach of the social sciences and summarises it in the following diagram:

Casting SD within the structure agency approach goes something like this:

  • People as agents make sense of the external world using mental models that guide their actions.
  • These mental models also result in social structures or institutions that then shape our mental models.
  • There is a range of viewpoints that people take, based on their experience and social interactions.
  • These mental models are the beliefs that affect the way we perceive the world and act to produce intended effects.
  • However we are bounded in our rationality, with limited information, limited mental models and limited options for action.
  • The complex interactions among agents and structures comprise the social systems that determine behaviour over time.
  • See also

Philosophical Thoughts

  • Its engineering origins often lead to SD being criticized as overly hard, impersonal and deterministic, like LaPlace’s clockwock universe.
  • Forrester initially emphasized the pragmatic nature of the approach, the search for real world empirical solutions and the need for the stock flow representations to correspond to real counterparts in the external world as well as an explicit symbol of theories of causation in our mental models, our beliefs to explain patterns of behaviour.
  • He also emphasized that confidence in the models stems from the ability of the structure to explain puzzling behaviour over time, rather than the accuracy of estimating the parameters in the model.
  • A good SD model can perhaps be described to adapt Aristotle’s definition of knowledge as a justified true belief to a socially justified improvable belief.
  • It provides a way to explicitly arrange and present knowledge, or a working theory of multiple complex causality, used to explain patterns of behaviour in a way that convinces people to act effectively.
  • This explanation of a pattern of behaviour is referred to as a dynamic hypothesis.
  • Using virtual experiments we can test our beliefs, iteratively modify and improve the way we present them, so we can learn to take more effective actions in similar situations.
  • This leads to a collection of generic structures and behaviours than can be used to guide our search for explanations and our corrective actions when faced with puzzling situations.
  • This explicit representation of our mental models in SD has been described as
    • presentationalism,
    • idealism, a search for positive knowledge within an idealist epistemology,
    • instrumentalism "not claims about the world but instruments for systematizing observations and for boosting learning processes using experimentation via simulation", and
    • relativism rather than positivism.

To summarise

  • Focus on better policies and explanations of problems
    • explains causal structure of a problem or situation
    • explains how the problem is created by this structure
    • explains why one policy has high impact while others do not
    • explains how established and defended policies are the underlying cause of the problem behaviour
    • argues in favour of better policies
  • Formal computer models are constructed following the scientific method (reference mode of problem behaviour, dynamic hypothesis, formal model, testing of the hypothesis against data, extensive analysis, and policy design


  • Camilo Olaya System Dynamics: Philosophical Background and Underpinnings Encyclopedia of Complexity and Systems Science Springer 2009 Volume 9 p9057-9078
  • David C. Lane and Elke Husemann Steering without Circe: attending to reinforcing loops in social systems Syst. Dyn. Rev. 24, 37–61, (2008)
Questions & Comments to Geoff McDonnell
Personal tools