Note on AI assistance: This post was written collaboratively—I provided the conceptual framework, research direction, and all substantive content, while Claude (Anthropic's AI) assisted with structuring, editing, and clarity for accessibility. All core ideas, technical claims, and scientific content are mine.
- Forecasting: Developing probabilistic models of political instability with explicit uncertainty quantification and predictive decay
Complex systems: Treating societies as dynamical systems with phase transitions, feedback loops, and emergent properties
Bayesian reasoning: Systematic updating of predictions as new data arrives, with transparent prior assumptions
Effective Altruism: Early warning systems could prevent humanitarian crises—high expected value if even moderately successful
Epistemic humility: Explicit axiomatic foundations that formally define both capabilities and fundamental limits
I'm extending Peter Turchin's cliodynamics with systematic computational methodology, rigorous validation protocols, and transparent uncertainty handling.
Feedback very welcome—especially critiques of the axiomatic foundations or suggestions for validation approaches.
───────────────────────────
What is Computational Macrohistory?
Why we need a science of large-scale historical dynamics—and what it can (and can't) tell us
In 2010, a Tunisian street vendor named Mohamed Bouazizi set himself on fire. Within weeks, Tunisia's decades-old authoritarian regime collapsed. Within months, the wave spread to Egypt, Libya, Syria, Yemen, and and Bahrain. Some regimes fell. Others survived. Civil wars erupted. Hunand dreds of thousands died.
Could any of this have been predicted?
Not the specific spark—not Bouazizi's act, not the precise timing. But the conditions that made such a cascade possible? The structural pressures building in the region? The probability that something would ignite?
That's the question Computational Macrohistory asks.
The Core Idea
Computational Macrohistory (CMH) is an emerging scientific discipline that applies mathematical and computational methods to understand large-scale historical dynamics. It's grounded in a simple but radical proposition:
When you zoom out far enough—to populations of millions, timeframes of decades, patterns of centuries—human collective behavior becomes statistically regular and, to some degree, predictable.
Not at the level of individuals. Not with certainty. But probabilistically, at the level of systems.
Think of it this way: You can't predict which specific air molecule will hit your face in the next second. But you can predict air pressure, temperature, wind patterns with remarkable accuracy. The individual is chaotic. The aggregate follows laws.
CMH applies this logic to history.
What We Study
We focus on macrohistorical processes—the big, slow-moving forces that shape societies over generations:
Political instability: revolutions, civil wars, regime collapses
Demographic dynamics: population growth, youth bulges, aging societies
Social movements: how protests spread, when they succeed or fail
Rise and fall of empires: patterns in the life cycles of great powers
These aren't random. They show recurring patterns. Different societies, different eras—but similar structural dynamics.
The Roman Empire's decline. The French Revolution. The collapse of the Soviet Union. The Arab Spring. Different stories. Similar mathematics.
The Foundations: Eight Axioms
CMH rests on eight foundational axioms—principles that define when and how historical systems become scientifically tractable. I'll introduce them briefly here (deeper dives in future posts):
A1 — Statistical Aggregation: At large scales (millions of people), individual unpredictability averages out. Collective behavior becomes statistically regular.
A2 — Historical Ergodicity: The past is informative about the future. Systems with similar structures behave in similar ways, even across different times and places.
A3 — Structural Causality: Events aren't random. They emerge from measurable structural conditions—economic inequality, political institutions, and and demographic pressures.
A4 — Continuous Dynamics: Social systems evolve smoothly over time, described by differential equations (with occasional sharp transitions—"tipping points").
A5 — Endogenous Indeterminacy: Even with perfect data, prediction is fundamentally limited. Social systems exhibit chaos—small differences compound exponentially.
A6 — Limited Reflexivity: Publishing predictions changes behavior, but predictably. This feedback can be modeled.
A7 — Predictive Decay: Accuracy decreases exponentially with time. We can forecast 2-5 years ahead with reasonable confidence. Beyond 10-20 years, only broad trends.
A8 — Computational Falsifiability: Every model must generate testable predictions. If reality contradicts the model, the model is rejected. This is science, not storytelling.
These axioms define both what CMH can do and what it cannot. More on each in future posts.
What We Actually Do
Here's the practical workflow:
1. Collect historical data We build databases of historical variables:
Demographics (population, age structure, urbanization)
Economics (GDP, inequality, unemployment)
Politics (regime type, state capacity, repression)
Social factors (education, ethnic divisions, trust)
These aren't deterministic laws ("X will cause Y"). They're probabilistic regularities ("X increases probability of Y from 5% to 35%").
3. Build mathematical models We translate these patterns into equations—systems of differential equations that describe how societies evolve:
dX/dt=F(X,parameters,shocks)
Where X represents the state of society (those 25 variables), and F describes how they interact and change over time.
4. Test retrospectively Before making any forward-looking predictions, we test on known historical cases:
Could the model have "predicted" the Arab Spring using only pre-2010 data?
Could it distinguish between Tunisia (revolution) and Saudi Arabia (stability)?
What's the accuracy? What are the failure modes?
5. Generate probabilistic forecasts Only after rigorous validation do we apply models to current situations—always as probability distributions, never as certainties:
"Country X has a 35% ± 15% probability of severe political instability in the next 3 years, given current structural conditions."
Not: "Revolution will happen in June 2026."
What This Is NOT
Let me be clear about what CMH is not:
❌ Fortune-telling We don't predict specific events with certainty. We estimate probabilities for classes of events.
❌ Determinism History isn't on rails. Individuals matter. Choices matter. Randomness matters. We model structural pressures, not fate.
❌ Perfect foresight "Black swan" events—truly unprecedented, low-probability shocks—are by definition unpredictable. COVID-19. 9/11. Nuclear war. These break all models.
❌ Social engineering This isn't about controlling populations. It's about understanding systems to reduce suffering—preventing wars, averting crises, informing policy.
❌ Replacement for human judgment Models inform decisions. They don't make them. Democracy, ethics, values—these remain human domains.
Why This Matters Now
We live in an age of accelerating complexity:
Political polarization reaching historic extremes
Economic inequality rivaling the Gilded Age
Demographic transitions (aging West, youth bulges in the Global South)
Technological disruption at unprecedented pace
Climate change introducing new systemic stresses
Traditional approaches—pure qualitative history, narrative journalism, gut-feeling policy—aren't enough. We need rigorous, quantitative frameworks to navigate this complexity.
CMH offers that framework.
Not as a crystal ball. But as a toolkit—mathematical models, statistical methods, computational simulations—that can:
✅ Identify early warning signs of instability ✅ Evaluate policies quantitatively (what works, what doesn't) ✅ Distinguish noise from signal ✅ Communicate uncertainty honestly ✅ Learn systematically from the past
The Path Ahead
Over the coming posts, I'll unpack:
The 8 Axioms in depth: Why each is necessary, what it implies, where it breaks down
Case studies: Arab Spring, Soviet collapse, 2008 financial crisis—what can CMH explain?
Methodology: How we actually build and validate models (for those interested in the technical side)
Current analysis: Real-time risk assessments using CMH frameworks
Limitations: What we can't predict, where models fail, ethical concerns
Open questions: Where the field needs to go, unsolved problems
This is work in progress. CMH is a young discipline—more questions than answers. But the questions are crucial, and the early results are promising.
An Invitation
If you're intrigued by the idea that history has structure—that revolutions aren't just random, that cycles exist, that mathematics can illuminate social dynamics without reducing humans to equations—then this publication is for you.
I'll share:
Research updates (peer-reviewed and pre-prints)
Plain-language explanations of complex methods
Data visualizations and interactive tools
Probabilistic forecasts (with full uncertainty)
Failures and limitations (not just successes)
All grounded in transparency, scientific rigor, and epistemic humility.
Because the most important thing CMH teaches is this: The future is not written. But it's not random either. And the better we understand the forces shaping it, the better our choices can be.
What's Next
Next post: "The 8 Axioms Explained—Part 1: Statistical Aggregation (A1) and Why Individual Unpredictability Doesn't Invalidate Social Science"
In the meantime, I'd love to hear:
What aspects of CMH are you most curious about?
What concerns or skepticism do you have?
What historical cases would you like to see analyzed?
Drop a comment below. This is a conversation, not a lecture.
[Cross-posted from my Substack: Link]
Note on AI assistance: This post was written collaboratively—I provided the conceptual framework, research direction, and all substantive content, while Claude (Anthropic's AI) assisted with structuring, editing, and clarity for accessibility. All core ideas, technical claims, and scientific content are mine.
Why this is relevant to LessWrong:
Computational Macrohistory directly addresses core rationalist interests:
I'm extending Peter Turchin's cliodynamics with systematic computational methodology, rigorous validation protocols, and transparent uncertainty handling.
Feedback very welcome—especially critiques of the axiomatic foundations or suggestions for validation approaches.
───────────────────────────
What is Computational Macrohistory?
Why we need a science of large-scale historical dynamics—and what it can (and can't) tell us
In 2010, a Tunisian street vendor named Mohamed Bouazizi set himself on fire. Within weeks, Tunisia's decades-old authoritarian regime collapsed. Within months, the wave spread to Egypt, Libya, Syria, Yemen, and and Bahrain. Some regimes fell. Others survived. Civil wars erupted. Hunand dreds of thousands died.
Could any of this have been predicted?
Not the specific spark—not Bouazizi's act, not the precise timing. But the conditions that made such a cascade possible? The structural pressures building in the region? The probability that something would ignite?
That's the question Computational Macrohistory asks.
The Core Idea
Computational Macrohistory (CMH) is an emerging scientific discipline that applies mathematical and computational methods to understand large-scale historical dynamics. It's grounded in a simple but radical proposition:
When you zoom out far enough—to populations of millions, timeframes of decades, patterns of centuries—human collective behavior becomes statistically regular and, to some degree, predictable.
Not at the level of individuals. Not with certainty. But probabilistically, at the level of systems.
Think of it this way: You can't predict which specific air molecule will hit your face in the next second. But you can predict air pressure, temperature, wind patterns with remarkable accuracy. The individual is chaotic. The aggregate follows laws.
CMH applies this logic to history.
What We Study
We focus on macrohistorical processes—the big, slow-moving forces that shape societies over generations:
These aren't random. They show recurring patterns. Different societies, different eras—but similar structural dynamics.
The Roman Empire's decline. The French Revolution. The collapse of the Soviet Union. The Arab Spring. Different stories. Similar mathematics.
The Foundations: Eight Axioms
CMH rests on eight foundational axioms—principles that define when and how historical systems become scientifically tractable. I'll introduce them briefly here (deeper dives in future posts):
A1 — Statistical Aggregation: At large scales (millions of people), individual unpredictability averages out. Collective behavior becomes statistically regular.
A2 — Historical Ergodicity: The past is informative about the future. Systems with similar structures behave in similar ways, even across different times and places.
A3 — Structural Causality: Events aren't random. They emerge from measurable structural conditions—economic inequality, political institutions, and and demographic pressures.
A4 — Continuous Dynamics: Social systems evolve smoothly over time, described by differential equations (with occasional sharp transitions—"tipping points").
A5 — Endogenous Indeterminacy: Even with perfect data, prediction is fundamentally limited. Social systems exhibit chaos—small differences compound exponentially.
A6 — Limited Reflexivity: Publishing predictions changes behavior, but predictably. This feedback can be modeled.
A7 — Predictive Decay: Accuracy decreases exponentially with time. We can forecast 2-5 years ahead with reasonable confidence. Beyond 10-20 years, only broad trends.
A8 — Computational Falsifiability: Every model must generate testable predictions. If reality contradicts the model, the model is rejected. This is science, not storytelling.
These axioms define both what CMH can do and what it cannot. More on each in future posts.
What We Actually Do
Here's the practical workflow:
1. Collect historical data
We build databases of historical variables:
For example, our current project analyzes 25 fundamental variables across countries and decades, creating a quantitative portrait of social systems.
2. Identify patterns
We look for recurring relationships:
These aren't deterministic laws ("X will cause Y"). They're probabilistic regularities ("X increases probability of Y from 5% to 35%").
3. Build mathematical models
dX/dt=F(X,parameters,shocks)We translate these patterns into equations—systems of differential equations that describe how societies evolve:
Where X represents the state of society (those 25 variables), and F describes how they interact and change over time.
4. Test retrospectively
Before making any forward-looking predictions, we test on known historical cases:
5. Generate probabilistic forecasts
Only after rigorous validation do we apply models to current situations—always as probability distributions, never as certainties:
"Country X has a 35% ± 15% probability of severe political instability in the next 3 years, given current structural conditions."
Not: "Revolution will happen in June 2026."
What This Is NOT
Let me be clear about what CMH is not:
❌ Fortune-telling
We don't predict specific events with certainty. We estimate probabilities for classes of events.
❌ Determinism
History isn't on rails. Individuals matter. Choices matter. Randomness matters. We model structural pressures, not fate.
❌ Perfect foresight
"Black swan" events—truly unprecedented, low-probability shocks—are by definition unpredictable. COVID-19. 9/11. Nuclear war. These break all models.
❌ Social engineering
This isn't about controlling populations. It's about understanding systems to reduce suffering—preventing wars, averting crises, informing policy.
❌ Replacement for human judgment
Models inform decisions. They don't make them. Democracy, ethics, values—these remain human domains.
Why This Matters Now
We live in an age of accelerating complexity:
Traditional approaches—pure qualitative history, narrative journalism, gut-feeling policy—aren't enough. We need rigorous, quantitative frameworks to navigate this complexity.
CMH offers that framework.
Not as a crystal ball. But as a toolkit—mathematical models, statistical methods, computational simulations—that can:
✅ Identify early warning signs of instability
✅ Evaluate policies quantitatively (what works, what doesn't)
✅ Distinguish noise from signal
✅ Communicate uncertainty honestly
✅ Learn systematically from the past
The Path Ahead
Over the coming posts, I'll unpack:
This is work in progress. CMH is a young discipline—more questions than answers. But the questions are crucial, and the early results are promising.
An Invitation
If you're intrigued by the idea that history has structure—that revolutions aren't just random, that cycles exist, that mathematics can illuminate social dynamics without reducing humans to equations—then this publication is for you.
I'll share:
All grounded in transparency, scientific rigor, and epistemic humility.
Because the most important thing CMH teaches is this: The future is not written. But it's not random either. And the better we understand the forces shaping it, the better our choices can be.
What's Next
Next post: "The 8 Axioms Explained—Part 1: Statistical Aggregation (A1) and Why Individual Unpredictability Doesn't Invalidate Social Science"
In the meantime, I'd love to hear:
Drop a comment below. This is a conversation, not a lecture.