The Delft Method for System Dynamics

The Delft Method for System Dynamics develops modeling skills through repeated practice, feedback, and progressively deeper theory. Willem Auping, Floortje d'Hont, Merla Kubli, and coauthors connect nine chapters on stocks, flows, feedback, formulation, evaluation, and policy use. This guide follows the November 2024 edition's actual chapter order and emphasizes explaining mechanisms and testing assumptions alongside software work.

Chapter-wise summaries and study tasks

Chapter 1 Introduction to System Dynamics

System dynamics relates behavior over time to an underlying structure of accumulations, rates, and feedback. Stocks store quantities, flows change them, and auxiliary relationships connect decisions and conditions. Reinforcing and balancing feedback can generate patterns that differ from a simple one-way chain of causes. Delays make the timing of a response as important as its direction. The chapter introduces simulation as a way to examine these relationships, not as a machine that guarantees a forecast.

The modeling cycle links a problem with conceptualization, equations, evaluation, and use. Learning by doing means constructing and inspecting small working models while using theory to understand what happens. Define units early and distinguish a model's representation of a system from the system itself. An attractive curve needs an explanation in terms of structure and assumptions.

Study task. Describe a fictional accumulation problem, label stocks and flows with units, and sketch a feedback loop. Predict a behavior pattern before running a simple model.

Chapter 2 The first modelling cycle

The first modeling cycle develops practical familiarity with moving from a description to a diagram, equations, and output. Theory exercises test whether the learner can recognize the role of variables and read causal or stock-flow representations. Modeling tasks provide constrained situations where a proposed structure can be compared with expected behavior. The purpose is not merely to reproduce a graph, but to learn the relationship between the description, formulation, and simulation.

Build incrementally so that each new relationship can be checked. A model that runs may still contain a conceptual mistake, while a mismatch with an example may arise from units, initial values, signs, or simulation settings. Document the cause of a correction rather than adjusting several parameters until the output looks familiar.

Study task. Complete a source exercise in small increments, recording each variable's role and units. Diagnose a deliberately introduced error through the structure and equation rather than fitting the final curve.

Chapter 3 Model formulation 1

The first formulation exercise chapter strengthens the translation between verbal relationships and quantitative models. Stocks, flows, and auxiliary variables need different kinds of equations, and a rate must be consistent with the stock it changes. Qualitative reasoning remains important: inspect whether the direction of an effect matches the problem description before trusting numerical output.

Theory and modeling exercises develop the habit of explaining a formulation choice. Algebraically valid equations can still represent the wrong mechanism, particularly when a fraction, a time constant, and an absolute rate are confused. Use a dimension check and a simple limiting case to test each relationship. Keep a working version so that the effect of a new formulation is visible and reversible.

Study task. For a selected exercise, explain every equation in words and test it at a zero or extreme input where meaningful. Distinguish a fractional change rate from a flow measured in quantity per time.

Chapter 4 Model formulation 2

The second formulation chapter extends practice with more involved structures and temporal relationships. Delays, information processes, and material transfers can have different meanings even when their outputs look similar. Translating a causal description into a stock-flow model requires identifying what accumulates and how information about a condition affects action.

A model should not acquire extra stocks merely to make a diagram look complex, nor omit an accumulation because it is inconvenient. Compare the order and type of a delay with the process being represented. Inspect how initial conditions and timing assumptions shape transient behavior. The exercises support a progression from identifying structure to predicting and explaining its consequences, with theory used to diagnose why two plausible implementations disagree.

Study task. Compare two delay representations for a fictional process, stating what is stored or perceived in each. Explain their initial conditions and identify an observation that could distinguish the resulting behavior.

Chapter 5 Model evaluation

The exercise-based evaluation chapter asks learners to inspect models as constructed explanations. Errors can arise in structure, equations, data, numerical implementation, or interpretation, and a useful test targets a specific possible failure. A model that reproduces one expected trajectory may still behave implausibly under another input or initial state.

Use dimensional checks, boundary cases, alternative settings, and comparison with the stated description to evaluate the formulation. Separate correcting an implementation error from changing a modeling assumption. Preserve test results so that a later revision can be evaluated against earlier behavior. This practical chapter prepares the more systematic discussion of verification and validation that follows; evaluation should accompany construction rather than become a final cosmetic step.

Study task. Create a test log for an exercise model with units, initial-state, extreme-input, and time-step checks. Record what each test could detect and what a passing result would not establish.

Chapter 6 Problem Formulation and Conceptualisation

Problem formulation identifies an undesirable behavior over time and a reason to use system dynamics. A reference pattern, purpose, scope, and boundary help determine which mechanisms belong in the model. Conceptualization then develops a dynamic hypothesis: an explicit account of how feedback and accumulation could generate the pattern.

Diagrammatic conventions make that hypothesis inspectable. Link polarity concerns a relationship under specified conditions, not a moral judgment or a guarantee about the entire system's trajectory. Loop polarity, delays, and interactions need careful reasoning. Archetypes can suggest questions but should not replace investigation with a familiar template. Explain what evidence supports the proposed structure and what omitted process might matter. A diagram is a hypothesis to test, not proof that its arrows are causal.

Study task. Write a reference-mode description for a fictional recurring problem, draw a dynamic hypothesis, and justify its boundary. Check link meanings and identify evidence that would challenge one proposed feedback mechanism.

Chapter 7 Formulation

Formulation turns the conceptual model into equations and numerical integration. Stocks require initial values and net rates of change; flows must use compatible units and represent physically or behaviorally meaningful mechanisms. Auxiliary variables organize intermediate relationships. Standard structures and functions are useful only when their assumptions fit the process.

Numerical methods approximate change over finite steps, so the integration method and step size can influence the result. Table functions and nonlinear relationships need justified domains and behavior outside observed values. A stock becoming negative may reveal a missing constraint, a poor numerical setting, or a genuinely incorrect mechanism. Formulate one relationship at a time, explain it, and test it before adding more interactions. Software syntax should not be allowed to conceal the intended system logic.

Study task. Implement a small stock-flow model and compare time steps. Document initial values, equations, units, constraints, and a nonlinear relationship, explaining which features come from evidence and which are assumptions.

Chapter 8 Evaluation

Verification asks whether the implemented model behaves according to its intended formulation. Validation examines whether the structure and behavior are sufficiently credible for a defined purpose. Neither is established by a single close fit to historical data. Tests can examine dimensions, integration settings, extreme conditions, structural plausibility, behavior patterns, and sensitivity.

Interpretation should explain why a result occurs and which assumption drives it. Scenario development explores plausible external conditions rather than silently treating one future as known. A model can be useful for comparing policies while remaining unsuitable for precise point prediction. Record failed tests and revisions, and assess whether the level of evidence matches the consequence of the decision. Confidence is purpose-dependent and should remain open to new observations.

Study task. Design a verification and validation plan for a fictional model, separating implementation tests from evidence about realism. Compare two external scenarios and explain how they alter the interpretation of the same policy.

Chapter 9 Model use and policy testing

Model use connects tested understanding with policy design and robustness. A policy changes a rule, information flow, capacity, or other mechanism rather than simply commanding the desired output. Feedback can compensate for an intervention or shift a problem elsewhere, making comparison over time essential. Evaluate several outcomes and the transition path, not only a final target.

Robustness concerns performance across plausible conditions and assumptions. A policy optimized for one forecast may fail when uncertain factors differ. Compare alternatives using explicit objectives, constraints, and failure conditions, and explain who experiences benefits or costs. System-dynamics analysis can support discussion of mechanisms and trade-offs; it does not remove value judgments or institutional constraints. Report the domain in which the model supports an inference and the circumstances requiring reassessment.

Study task. Compare two fictional policies across several plausible assumptions. State success and failure criteria, identify compensating feedback, and recommend a monitoring trigger for revisiting the model or policy.

How to study this book

1. Attempt the exercises early, keeping theory available to explain difficulties rather than waiting to master every concept before modeling.
2. Define units and initial values immediately, build one relationship at a time, and preserve working versions.
3. Predict behavior before running a simulation, then explain differences between prediction and output.
4. Maintain an equation record and test log separating formulation choices, implementation errors, and uncertain assumptions.
5. Use the actual numbered contents when navigating: the source preface contains some placeholder chapter references that do not match the final sequence.

Questions for reflection

• Which accumulation and feedback generate the behavior?
• Does an equation represent a mechanism or merely reproduce a desired curve?
• What does a successful test establish, and what remains uncertain?
• Would the policy still work under another plausible future?

Source edition and coverage

The guide follows the nine numbered chapters in the November 2024 TU Delft source. Its exercise-first chapters are distinct from the later systematic theory chapters. Source-specific software examples support learning, while the tasks here use fictional systems and do not claim a validated real-world policy model. These are original study explanations and learning tasks. Read the complete book at the official source.