[The Electronic Oracle] ④ Three Yardsticks That Keep Computer Models from Becoming Oracles

[The Electronic Oracle] ④ Three Yardsticks That Keep Computer Models from Becoming Oracles


Before introducing the types of models. Instead, this article organizes three critical distinctions—Structure/Parameter, Exogenous/Endogenous, and Accuracy/Precision—that we must verify to ensure any model does not become an "oracle."

Questioning the Essence of Modeling: Three Core Standards from The Electronic Oracle

In The Electronic Oracle, D. H. Meadows and J. M. Robinson place these ideas early because they separate not only methods, but mindsets. Authors define what we should look for in a model. These concepts are not mere definitions; they are criteria that distinguish the modeler’s attitude and philosophy toward the world. (p.12)

1. Structure vs. Parameter

Structure + Endogenous = 'Why.'(Explanation)
Parameters + Exogenous = 'Given.' (Settings)

  • Structure:

    • Definition: the system’s working mechanism—the pattern of causality that generates behavior over time. Structure includes more than “A affects B.” It includes stocks and flows, feedback loops, delays, and nonlinear decision rules.

    • The SD View: ๏ผก system's long-term behavior emerges from its internal structure rather than its parameters.

  • Parameter:

    • Definition: Specific numbers that quantify the assumptions of the structure (e.g., birth rate 0.02, price elasticity -0.5). Therefore, parameters include not only coefficients but also 'numerical settings' such as initial values, thresholds, and adjustment time.

Comparison & Evaluation:

  • Many statistical models excel at estimating parameters based on past data. In contrast, SD places structural hypotheses (feedback, delay, endogeneity) at the forefront.

  • A model can look impressive while hiding a weak structure behind elegant calibration.

๐Ÿ’ก Key Insight:

  • Policy often changes structure (rules and feedback). Parameters mostly change magnitudes.

2. Exogenous vs. Endogenous

"The boundary of the model becomes the boundary of responsibility."

  • Exogenous:

    • Definition: Variables set from outside the system. The model takes it as “given.” It can affect the system, but the model does not let the system affect it. This does not mean the influence is one-way in the real world; rather, it reflects the modeler's choice to exclude feedback within the model.

    • Examples: "Global oil price fluctuations," "Unpredictable natural disasters," "External legal/institutional changes."

  • Endogenous:

    • Definition: Variables generated by the model’s internal interactions—by feedback, delays.

    • Examples: "Production volume determined by price-demand-investment feedback," "Number of teachers determined by the interaction between workload and turnover rate."

Comparison & Evaluation:

  • SD attempts to explain the causes of problems as endogenously as possible through feedback loops. Conversely, econometric models tend to treat many variables as exogenous, viewing the system as 'Open'.
  • Case Study: From an SD perspective, "a disaster itself may be an exogenous variable, but vulnerability and the scale of damage are endogenous variables." regarding the underpass flooding tragedy, while heavy rain due to climate change was an exogenous factor, SD diagnoses that endogenous factors (e.g., failure of the traffic control system) were the mechanisms that amplified vulnerability and damage.

๐Ÿ’ก Key Insight:

To endogenize a variable is to pull responsibility into the system. To exogenize it is to push the explanation outside the boundary.

3. Accuracy vs. Precision

"Precision creates authority, but it does not guarantee accuracy." "This is the boundary between 'Plausibility' and 'Validity'."

  • Accuracy:

    • Definition: Closeness to reality—how small the error is. The book sometimes frames it as the “absence of error,” but in social modeling, it often means: does the model reproduce the real situation in a way that supports sound decisions? "About 2" is an accurate expression if the true value is 2.00563. In summary, accuracy is about how close to reality something is (the size of the error). Improvement in accuracy means the error relative to reality decreases.

  • Precision:

    • Definition: The exactness of expression or calculation. It refers to the number of significant digits. "2.97351" is very precise, but if the true value is 2.0, it is inaccurate. To re-emphasize: Precision concerns only the level of detail in expression/calculation.

  • Comparison & Evaluation:

    • Because computer models inherently output many digits, they risk producing 'False Precision'—the illusion that the model is more certain than reality. SD often values accurate behavior modes (the shape of change over time) over precise point forecasts.

  • ๐Ÿ’ก Key Insight:

    • A precise number can win a meeting. An accurate pattern can prevent a policy mistake.

Comments

Post a Comment

Popular posts from this blog

Thinking in Systems Guide (1) – Preface

Image
   The Beginning of Systems Thinking: A New Lens on the World Taught by a Slinky "If a factory is torn down but the rationality which produced it is left standing, then that rationality will simply produce another factory." - Robert Pirsig (Here, “rationality” can be understood simply as the reason or purpose behind it.)   Why do we always face the same problems? Have you ever had this experience? You worked hard to solve a problem at work, but before long, a similar issue appeared again in a different form. Or you may have tried to break a bad habit, succeeded for a while, and then suddenly found yourself back at square one. These kinds of repetitions happen because we focus only on the symptoms of the problem. The true cause—the underlying structure —remains untouched. The Slinky Experiment: The First Insight of Systems Thinking Donella Meadows, who taught system dynamics at MIT, often conducted a fascinating classroom experiment. She used nothing more than a Slinky—t...
Read more

The + and - Symbols: Simple Enough for Elementary Students, Yet the World's Most Confusing Concept

Image
Let me talk about the most challenging concept in the world—at least for me. It’s so simple that even elementary school students understand it. But believe me, it gets you when you least expect it. I’m talking about... + and -. And yes, there’s a twist. Actually, there’s a twist to the twist—so buckle up.  In my systems thinking lectures, I typically explain that when cause and effect move in the same direction, we use the + symbol and refer to it as a positive causal relationship. When they move in opposite directions, we use the - symbol and call it a negative causal relationship. I always emphasize firmly that "positive" and "negative" here are purely mathematical concepts—they have nothing to do with good or bad. Elementary School Level   For example, when the birth rate increases, the total population also increases. Since both birth rates and population changes move in the same direction, we write:  Births → Population (+)  Conversely, when the death rate incr...
Read more

[The Electronic Oracle] ③ Assumptions Aren’t "Guesses": A Proposal for Transparent Democracy

Image
  [The Electronic Oracle] ③ Assumptions Aren’t “Guesses”: A Proposal for Transparent Democracy When I teach System Dynamics or computer modeling, I often encounter a word that instinctively annoys both students and practitioners: "Assumption." The dictionary reveals the source of this discomfort. Assumption :  “a thing that is accepted as true or as certain to happen, without proof .” (Oxford Languages) For scientists, the phrase "without proof"  can sound like an insult. So when someone says, “That’s just your assumption,” scientists rush to defend themselves: "No! I have evidence!" In this post, I want to restore the honor of this word. Modeling assumptions are not mere tricks to patch missing data; they are influential political acts that transform secret dictatorships into transparent democracies. This article distinguishes two layers of assumptions: The Assumption as Belief (Type A)   hides in the mind and operates on authority The Model Assumption (T...
Read more