Abstract:

This paper contributes to teaching and learning conceptual model development for complex dynamic problems. The ability to conceptualize is important for critical thinking at large and for computational modeling in approaches like system dynamics (SD).
SD modeling helps to understand and solve complex dynamic problems by developing a simulation model of the variables that interact to generate the problem. Model development is an iterative sequence of steps, starting with an initial problem statement, followed by conceptualization, quantification, validation, and exploitation. I focus on conceptualization: the ability to grasp the relevant factors of a problem and their interdependencies. A conceptual model represents the problematic behavior patterns and the causal mechanisms generating them in natural language. Conceptualization flaws compromise the downstream phases of modeling.

Textbooks and formal courses for prospective presuppose a series of analytical and synthetic reasoning skills and systems knowledge. The skills enable an individual to mobilize and orchestrate the systems knowledge needed to translate a problematic and initially unknown situation into a conceptual model. However, learners at the undergraduate level or in high school have not learned these skills yet. Nor do many self-directed learners.

Model conceptualization is a competence that novices develop over stages, starting by following context-free rules to compensate for personal skills and systems knowledge. As learners’ personal experience grows, they gradually become independent from the abstract rules. The current system dynamics competence framework names the implied skills, but there is a lack of abstract rules describing how to carry out the specific activities.

The paper describes a first version of such abstract rules. It starts with the commonly followed model conceptualization tasks, which are decomposed into sub-tasks and then procedures over several hierarchical levels. The lowest level are decision rules for defining the problem, recognizing variables and elementary behavior patterns, and organizing them into episodes and transitions, detecting causal links and classifying their polarity, detecting and classifying feedback loops, and developing a tentative explanation of how the interplay of these variables generates the behavior patterns in text and diagram. The procedures imitate the reasoning of an experienced modeler, and the decision rules represent the systems knowledge needed to take the modeling decisions.

At the current stage of development, two dynamic decision problems accompany the set of procedures and rules. Each problem has (a) a written introduction and (b) a black-box simulator, which allows learners to get pseudo-historical data about the problematic behaviors. A set of pre-formatted spreadsheets is used to structure the execution of the procedures. A group of 34 second-year students at my university is applying the procedures and rules to both problems. Currently, the visible effects of the stepwise procedures have been: (a) improved understanding of the dynamic behavior of the variables, (b) the unearthing and resolution of systemic misconceptions in the hypothetical causal links, and (c) fewer errors in the equation formulation.

At the conference, I will present the outline of the procedures and rules together with these effects and key elements of the students’ experience.

Keywords:

Feedback-rich systems, Dynamic problems, System dynamics modeling, Conceptualization.