Session 1: Stocks: Horse Market, Oil Market, and Transportation System
📚 Main topics
- Course purpose and formatHazhir Rahmandad introduces a hands-on seminar designed to bridge basic system dynamics training and the skills needed for research and consulting projects. Students will build models weekly, review and critique selected submissions, and discuss architectural choices for upcoming problems. 0:00
- Different modeling purposesSystem dynamics models can support stakeholder conversations, build general insights or test theories, make predictions, inform major or repeated decisions, and train decision-makers. Their required detail, empirical grounding, and complexity depend on their intended use and audience. 12:09
- Modeling process and problem definitionThe course emphasizes simulation modeling and analysis, using supplied problem definitions and qualitative structures as starting points. Rahmandad describes reference-mode, mechanism-driven, and policy-driven approaches to defining problems, and stresses making the model boundary and other architectural choices explicit. 23:49
- Architectural choices in practiceThe class applies questions about audience, time horizon, model boundary, aggregation, mechanisms, and sources of inertia to the thoroughbred horse market and urban transportation examples. The appropriate choices differ depending on whether a model serves theoretical insight or operational decision-making. 1:24:10
- Horse market and transportation examplesThe horse-market discussion considers how price expectations, speculative demand, status, breeding, and horse maturation can shape boom-and-bust dynamics. The transportation discussion considers congestion, road and transit capacity, travel preferences, car ownership, and urban growth. 1:25:41
- Next week’s SIR extensionThe next modeling task adds behavioral responses to epidemic risk, waning immunity, and vaccination to the SIR model. Students are encouraged to keep the model as aggregate as possible while meeting the exercise requirements. 2:22:17
✨ Key takeaways
- Match the model to its purposeA model for general theoretical insight can often be simple and highly aggregated, while a model supporting a consequential investment or operational decision may require more detail, data, and calibration. 14:16
- Make the model boundary deliberateA clear problem definition helps determine what belongs in the model and prevents unnecessary detail from making it unwieldy or difficult to explain. 33:26
- Choose aggregation consciouslyMore aggregation can simplify analysis and communication, while disaggregation may better use detailed data, capture important mechanisms, or represent decision-makers’ actual policy levers. 44:40
- Identify stocks from sources of inertiaStart by considering where delays and accumulations come from, then select clear, relevant stock variables at a consistent level of aggregation. 1:08:26
- Keep concepts clear and distinctDefine variables in relation to a concrete case where possible, and avoid overlapping constructs that represent the same underlying idea more than once. 1:06:50
- Reuse established formulationsLearn and apply existing system dynamics formulations before inventing new ones; keep equations understandable, limit their inputs where practical, and check unit consistency. 1:13:15
🧠 Lessons learned
- Build and analyze in small chunksStart with a manageable model component, such as a stock and its flows, and understand its behavior before adding more structure. Putting a component in equilibrium can make its response to shocks easier to interpret. 53:42
- Test explanations of behaviorUse causal tracing to identify a candidate feedback loop or external driver, then disrupt the hypothesized cause and check whether the behavior changes. 55:50
- Develop intuition before changing parametersPredict how parameter changes will affect outcomes and compare those predictions with the model’s behavior. Avoid blind parameter searches or adding structure just to force a desired result. 59:56
- Treat stock selection as purpose-dependentFor the horse-market example, a high-level theoretical model may need only a few horse stocks and a price-trend formulation; a decision-support model could require more detailed data and market structure. 1:49:43
- Trace inertia to its underlying causeIn transportation, perceptions of travel time and transit quality may capture longer-term shifts in travel choices better than a short-lived count of people currently using transit. 2:15:58
🏁 Conclusion/next steps
- Use iteration as the core working methodStudents should begin with the most important manageable structures, analyze them, and add complexity only when it is justified by the problem definition and the model’s behavior. 28:01
- Prepare the epidemic modelExtend the SIR structure in stages by adding behavioral responses to risk, waning immunity, and vaccination, while considering the relevant stocks and maintaining an appropriately simple model. 2:22:17
Transcript excerpt
0:00 [SQUEAKING] [RUSTLING] [CLICKING] HAZHIR RAHMANDAD: Welcome. This is 15.879, System Dynamics Research Seminar. And this is a seminar series that each semester we have a variant of it. So we have about four different versions of this course that we iterate through over a two-year period.
0:31 So this version is one focused on building classical modeling skills, covering the gap between what you get in 15.873 or any variant of a full semester course on system dynamics and what you need to be able to do serious modeling in an actual consulting project, in research projects. This is a gap that typically students don't have much opportunities, in terms of resources,
1:04 to do self-study on. We also don't have that many courses around-- across the world to cover the skill sets here. So I hope we have a fun seminar series throughout the semester together. But also, I hope that we can create some material that is useful, more broadly, for anybody who wants to learn system dynamics beyond the basics. So that's why we are actually recording these sessions. And we will share that through the OpenCourseWare Project
🔒 The full, searchable transcript is available with Pro.
Questions & Answers
Common questions about this video
¿Qué tipos de usos de los modelos de dinámica de sistemas distingue el instructor?
Menciona, entre otros, la consulta de procesos para facilitar conversaciones entre grupos, la creación de ideas teóricas generales, la prueba de teorías con datos, la predicción, el diseño de políticas y las herramientas de apoyo a decisiones repetidas. Cada uso requiere distintos niveles de detalle, calibración y participación de las partes interesadas. 13:11
¿Cuáles son tres formas de definir un problema de modelado?
Se puede partir de un modo de referencia, es decir, un patrón de comportamiento observado; de mecanismos que se quieren estudiar, como la competencia o el boca a boca; o de una pregunta de política concreta que el modelo debe ayudar a responder. 35:02
¿Por qué es importante analizar cada fragmento del modelo antes de seguir añadiendo estructuras?
Analizar cada fragmento ayuda a entender por qué se produce su comportamiento y evita que las dinámicas nuevas queden confundidas con transitorios previos. El instructor recomienda formular una hipótesis mediante trazado causal y probarla interrumpiendo el mecanismo propuesto para ver si cambia el comportamiento. 54:16
¿Qué significa el principio MECE aplicado a la selección de stocks?
Significa que los stocks deben ser mutuamente excluyentes y colectivamente exhaustivos: no deberían contener simultáneamente la misma persona, objeto o recurso, y juntos deberían cubrir el conjunto que se quiere representar. 1:10:00
¿Por qué recomienda el instructor usar formulaciones existentes en vez de inventar otras nuevas?
La comunidad ya ha desarrollado formulaciones útiles, recopiladas, entre otros recursos, en las moléculas de dinámica de sistemas. Aprenderlas y probar cómo responden a entradas típicas ayuda a construir modelos comprensibles; las formulaciones realmente nuevas son poco comunes y deben resolver un problema significativo. 1:13:15
¿Cómo cambia el modelo del mercado de caballos pura sangre según la audiencia?
Para un artículo teórico dirigido a una revista de sociología, el modelo puede ser más simple, agregado y centrado en comunicar mecanismos. Para ayudar a una persona de alto patrimonio a decidir sobre inversiones, se necesitarían más datos, mejor calibración y más detalle, además de considerar factores externos como las tasas de interés y el mercado bursátil. 1:34:16
¿Qué stocks propone considerar en un modelo agregado de transporte urbano a largo plazo?
Entre los candidatos están la capacidad de las autopistas y del transporte público, los residentes urbanos y suburbanos, los automóviles y las percepciones de congestión y de calidad del transporte público. La elección depende del horizonte temporal: las preferencias de transporte, por ejemplo, pueden ser relevantes en un modelo de décadas, pero no en uno de planificación de pocos años. 2:12:18
¿Qué extensiones del modelo SIR se asignan para la semana siguiente?
Se propone añadir respuestas de comportamiento ante el riesgo, como reducir las interacciones durante una epidemia; la pérdida gradual de inmunidad; y la vacunación. Se sugiere incorporarlas de manera iterativa y mantener el modelo tan agregado como sea posible para un ejercicio conceptual. 2:22:17
Related summaries
Wikipedia & MIT Open Learning: A New Strategic Collaboration
Lecture 24: Entanglement: QComputing, EPR, and Bell's Theorem
1. Introduction and Supply & Demand
Stanford CME295 Transformers & LLMs | Autumn 2026 | Lecture 4 - Reinforcement Learning with LLMs
ARCANO 52: EL ERROR DE ACTUAR SIN PENSAR | Tarot y Kábala
