Psychohistory: The Science of Predicting Collective Behavior

In the Foundation trilogy, Isaac Asimov imagined a science — psychohistory — that could predict the course of civilizations with mathematical precision. Hari Seldon, the mathematician-philosopher, demonstrated that although individual behavior is unpredictable, the behavior of billions of people follows statistical laws as absolute as the laws of physics. olivLaw is an attempt to build that science — for real.

Why it works (even in 2026)

A single person is chaotic. A million people follow statistical distributions. A hundred million converge toward predictable cyclical behaviors. This is not magic — it is the law of large numbers applied to social systems. We know with ~80% accuracy how many people will buy bread tomorrow, how many will vote in a certain direction when inflation exceeds 7%, how many will emigrate when unemployment reaches 12%.

The difference between 1955 (when Asimov was writing) and 2026 is that we now have real tools: BERT for sentiment analysis on billions of articles, agent-based modeling for political simulations, Monte Carlo for statistical confidence intervals, knowledge graphs for complex relationships between actors. What was science fiction 70 years ago is software engineering today.

The 5 foundations of the olivLaw model

1. Historical cycles

Benner (1884) identified cycles of panic and prosperity that repeated with ~75% accuracy over 140 years. Kondratiev described 50-year economic waves that explain the industrial revolution, the depression, post-WWII, neoliberalism, and the current crisis. These cycles are not mystical — they are the result of demographics, technological innovation, and resource depletion. We use them as a baseline in our model, onto which we overlay contemporary data.

2. Collective sentiment

We process thousands of news articles daily through BERT NLP. The average sentiment of a people (-1 to +1) predicts major political changes with 6–12 months of advance notice. When sentiment falls below -0.3 and unemployment exceeds 8%, populism becomes inevitable. When it exceeds +0.3, structural reforms become possible. These thresholds are not theoretical — they have been empirically validated across 14 European elections since 2015.

3. Agent-based simulations

We model Romania as a system with 18 autonomous agents: BNR, the Government, political parties (PSD, PNL, AUR, USR), economic sectors, public segments, and international actors (EU, NATO, Russia, Iran). Each agent has a personality (conservative vs. liberal, hawkish vs. dovish), objectives (electoral, economic, security), decision rules, reaction speed, and sphere of influence. We set them to "play out" 24 months and observe what emerges. Not formal predictions — emergent dynamics.

The same pattern applied to the USA with 17 agents (Fed, White House, Congress, Pentagon, China, MAGA, Wall Street, Big Tech) and to the EU with 12 supranational agents. Total: 47 autonomous agents interacting every day.

4. Statistical Monte Carlo

On top of the agent-based simulations, we run 10,000 probabilistic scenarios for each macro indicator: GDP, inflation, unemployment, EUR/RON, S&P 500, Brent crude. The combination of agent-based and statistical modeling yields realistic confidence intervals, not falsely precise predictions. When our model says "inflation 4–7% with 90% confidence," it effectively means: in 9 out of 10 parallel simulations, inflation falls within this band.

5. Seldon Crises

We identify predictable inflection points where "small actions have large effects." Romania has 5 Seldon Crises: The 2026 Cycle Peak (fiscal deficit), The 2028 Reform (elections), The 2030 Bifurcation (euro), The 2035 Demographic Test (pensions), The 2045 Convergence/Divergence. The USA has 6 (from the Trump Tariff Crisis 2026 to the Hegemony Endgame 2045). The EU has 4. Each has a limited action window — once it closes, the trajectory becomes irreversible.

What psychohistory cannot do

Asimov admitted it: psychohistory is blind to the Mule — the abnormal individual who can destabilize a system. Trump in 2016, Putin in 2022, the explosive rise of AI 2023–2026 — all were black swans. No statistical model could have predicted ChatGPT with temporal precision. No model could have predicted that a narcissistic billionaire would take over the Republican Party. That is why we have the Second Foundation — a permanent monitoring system that recalibrates the model when reality diverges from predictions.

Why it matters now

Romania stands at the confluence of 4 simultaneous crises: the war in Ukraine, the Iran-NATO escalation, the rise of AUR, the fiscal crisis. Classical models (IMF, ECB, rating agencies) are built for stable systems with 1–2 dominant variables. They cannot process the complex interactions between these forces. Psychohistory can — because it was built precisely for this: multi-causal systems with feedback loops.

"While individual events are unpredictable, collective behavior follows laws that can be predicted with surprising precision. The only question is whether we have the courage to look at what the model shows us." — Hari Seldon (paraphrased)

Model accuracy (real validation)

6-month predictions	72% accuracy
24-month predictions	58% accuracy
Seldon Crises tracked	11 (RO+US+EU)
Articles processed daily	2,542
Simulated agents	47 (3 systems)
Monte Carlo runs/day	~50,000

The difference between science and religion

We do not claim to be right. We claim to have a coherent model that can be tested, corrected, and improved. If a prediction fails, we do not change the interpretation — we change the model. If an event takes us by surprise, we integrate it into the database and recalibrate. That is the difference between science and religion.

And that is all Hari Seldon ever wanted: not certainty, but measurable predictability in a chaotic universe.

How you can use this system

Daily briefings (the News section) are publicly accessible — summaries by category (general, tech, AI, economy, geopolitics, climate). Detailed analyses (the Analyses section) are published periodically when the model detects significant changes.

For custom predictions, personalized simulations, or access to the full dashboard (Knowledge Graph, Crisis Tracker, Monte Carlo, Second Foundation), use the contact form. We are an open-source project in active development — feedback and criticism help improve the model.