Article 8: The Bedrock of Science – Axiom A8 Explained

Why Real Science Must Be Able to Be Proven Wrong

Imagine two people arguing:

  • Person A: “I have a magic rock that keeps tigers away.”
  • Person B: “That’s silly. There are no tigers around here anyway.”
  • Person A: “See? It works perfectly!”

This is a non-falsifiable claim. No matter what happens, Person A can say their magic rock worked. There’s no possible test to prove it’s not magic. It might be a fun belief, but it’s not science.

Now imagine an engineer says: “Based on my bridge design calculations, this bridge will hold at least 10,000 cars.” We can test that. We can load it with weight until it breaks. If it breaks with 9,000 cars, the engineer was wrong. If it holds 15,000, they were right. This is falsifiable. It’s real engineering.

Axiom A8 – The Computational Falsifiability Axiom applies this exact standard to our science of history. It’s the rule that keeps CMH from becoming just another storytelling or fortune-telling game. It’s what makes it a real science.

It states:

Every CMH model must generate predictive distributions that can be quantitatively compared with observed data.

In simple terms: Our models must make specific, numerical predictions about the future that we can actually check against what really happens. And we must be willing to throw our models away if they fail the test.

Diving Deeper: The Mechanics of Testing History

1. The Predictive Distribution – Our Scientific Offering

The formal statement gives us the core output of any CMH model:

PM(X(t+Δt)|X(t))P_M(X(t+\Delta t) \mid X(t))

Translation: “The probability distribution of possible future states X at time t+Δt, given current state X(t), according to model M.”

This is not a single prediction like “Rome will fall in 476 AD.” That’s a guess. This is a landscape of possibilities with probabilities attached, like a weather forecast that says:

  • 60% chance of rain
  • 30% chance of clouds
  • 10% chance of sun

For history, it might be:

  • 40% probability of civil war within 10 years
  • 50% probability of stable stagnation
  • 10% probability of rapid expansion

This probabilistic forecast is what we test.

2. The Testing Arsenal: Scoring Rules

How do we test a probability? We use proper scoring rules – mathematical tools that judge how good a probabilistic prediction was after we see what actually happened.

Think of it like this: If your weather app said “100% chance of sun” and it rains all day, that prediction gets a terrible score. If it said “60% chance of rain” and it rains, that gets a good score – even though it didn’t say “100%.”

For CMH, we might use metrics like:

  • Brier Score: Measures how close our probabilities were to reality (0% or 100%)
  • Logarithmic Score: Heavily penalizes being confidently wrong
  • Calibration Tests: Checks if when we say “40% chance,” the event really happens about 40% of the time across many predictions

3. Computational Falsification – The “Try and Break It” Rule

The word “computational” here is crucial. It means:

  1. Our models must be precise enough to be implemented as computer code
  2. Their predictions must be quantitative enough to be compared numerically
  3. The comparison must be algorithmic – not just “it feels right” but measurable

This creates a beautiful, self-correcting system:

  1. Build model M based on axioms 1-7
  2. Use M to generate probability distributions for future states
  3. Wait for reality to happen (or look at historical data we didn’t use to build the model)
  4. Score the predictions computationally
  5. If scores are consistently bad: Reject or revise model M
  6. Repeat

This is the scientific method applied to history.

Powerful Consequences: The Discipline of Accountability

What It ALLOWS Us To Do:

  • Evolve and Improve: We can have competing models of history and actually determine which is better through evidence, not rhetoric. Better models survive; worse ones get discarded.
  • Build Cumulative Knowledge: Like physics or medicine, CMH can build on tested models. We don’t start from scratch each generation.
  • Separate Science from Speculation: Any historical theory that cannot generate testable probabilistic forecasts is, by definition, not CMH. It might be interesting philosophy or narrative history, but it’s not this science.
  • Use Simulation as Evidence: Running thousands of simulations isn’t just “playing with computers” – it’s generating the probability distributions that we then test against reality. The simulation is the theory.

What It FORBIDS Us From Doing:

  • Making Vague, Untestable Claims: Saying “empires tend to fall when they get corrupt” is not CMH. Saying “when the elite overproduction index exceeds 0.7 and the state fiscal capacity falls below 0.3, the probability of state collapse within 30 years exceeds 75%” is CMH – and it’s testable.
  • Explaining Everything After the Fact: The “just-so story” problem – where you can explain anything that already happened but can’t predict what will happen – is banned. CMH models must predict before we see the outcome.
  • Hiding Behind Complexity: We cannot say “my model is too complex to test” or “history is too special to be tested.” If it can’t be tested, it’s not CMH.
  • Cherry-Picking Evidence: We must test against all relevant data, not just the examples where our model happens to work.

Real-World Example: Demographic Predictions vs. “Cycles of History”

Consider two approaches to predicting population:

Non-CMH Approach (Untestable):
“Civilizations have natural life cycles of growth and decline, like organisms.” This is poetic, maybe insightful, but how do you test it? What exactly predicts when the “decline” phase starts? How long do cycles last? There’s no precise, testable prediction here.

CMH Approach (Testable):
The Demographic Transition Model makes specific, quantitative predictions:

  1. As GDP per capita rises above $X, birth rates will begin falling with coefficient Y
  2. As female education reaches level Z, fertility rate will decline by W
  3. Given current trends, Country A’s population will peak in 2045 at between 1.2M and 1.4M with 90% confidence

These predictions are falsifiable. We can wait and see if birth rates actually fall when GDP reaches $X. We can check if Country A’s population actually peaks around 2045. When demographers get these predictions wrong (and they sometimes do), they go back and improve their models. That’s CMH in action.

Another example: Peter Turchin’s structural-demographic theory makes specific predictions about periods of instability based on measures of elite overproduction and popular welfare. He predicted increased instability in Western societies around 2020 based on 2010 data. Whether he was right or wrong, the prediction was specific and testable – that’s Axiom A8 in practice.

The Philosophical Core: The Courage to Be Wrong

Axiom A8 is what separates CMH from prophecy, from fortune-telling, from just-so stories about history. It embodies Karl Popper’s great insight: What makes something scientific isn’t that it can be proven right, but that it can be proven wrong.

This requires immense courage. It means putting your carefully crafted model of history – perhaps built over years – on the line every time it makes a prediction. It means saying: “Here’s what my model says will probably happen. If reality consistently contradicts these probabilities, then my model is wrong and should be discarded.”

This is why the Foundation’s Plan in Asimov’s stories had to be secret. If everyone knew the predictions, they might change their behavior (Axiom A6), making the predictions untestable. But the psychohistorians themselves had to be able to test their models against reality, even if the public couldn’t.

In the end, Axiom A8 grounds our entire enterprise. It ensures that Computational Macrohistory remains connected to reality, accountable to evidence, and worthy of the name “science.” It transforms history from a collection of interesting stories into a rigorous, evolving, self-correcting understanding of human civilization’s dynamics.

With all eight axioms now in place, we have the complete foundation. We know:

  • What we study (large groups, not individuals)
  • That we can learn from the past
  • That events have structural causes
  • That change is continuous
  • That the future is inherently uncertain
  • That predictions change reality
  • That our vision fades with time
  • And finally, that everything we claim must be testable

This is the bedrock. Now we can build.

Similar Posts