Structure firstPredict better

A good ontology changes every decision an organization makes. Yet most stop at the door, because of four walls.

First, they don’t know the real question. They know the outcome they want, but cannot define what must be asked to get there.
Second, they cannot draw the causal map. They cannot lay out which factors cause what, and how that leads to the outcome.
Third, the data does not exist. Even when they know what is needed, there is no data to fill the gap.
Fourth, they don’t know the weights. Even with data, they don’t know how strongly, and in which direction, each factor acts.

Alea Capta climbs these four walls in order. Diagnosis pins down the real question, Wiring draws the causal map, Sequencing creates and fills the missing data, and Potentiation reveals the weights by checking against the past.

Once the four walls are behind us, the scattered world becomes a single structure. Prediction only begins on top of that structure.

When the world wants to know public opinion, it runs a survey. Draw a sample, make the calls, ask once and stop. It is expensive, slow, and you cannot ask again.

Alea Capta does not survey opinion. We build the population that holds the opinion, and ask it directly.

Using Statistics Korea data, we build a virtual electorate that reproduces the real population distribution. Each person is given hundreds of attributes — age, gender, income, occupation, housing, political leaning and more — reproduced precisely down to the administrative-dong level. Because it is a population shaped from statistics rather than real people, it is free of personal-data issues.

You can ask this population as many times as you like. Changing the conditions, again and again. When turnout shifts, when a new issue breaks, when a late rally happens, we ask again each time. It is not a one-off survey but a living panel that runs all the time.

This is why we don’t stop when internal data is unavailable, or when the data does not exist anywhere. We create the population we need.

Building the structure and filling in the data is only half the job. The map comes alive only when we know how strong each link is — its weight.

Calibration is not a single step. It is the engine that runs across the entire process of building the ontology. It checks inferred data against reality, adjusts the weight of each factor against the past, and even after the world is complete, re-tunes the whole structure to the values flowing in and out.

Here is how it works in practice. How much does each day’s event move the outcome? We compare past events with the real changes that followed and turn their influence into a score. When a new event breaks, we find the most similar past event to rate its strength, and every week we match forecasts against outcomes and re-adjust the weights.

This is why we call our ontology “alive”: it is a structure that keeps checking itself against reality and correcting itself. This level of precise comparison has only become practical on today’s AI.

Once the structure is complete and the weights are tuned, we can experiment with the future.

Monte Carlo — roll a single event countless times. When a specific event happens, we simulate how its impact spreads, tens of thousands to millions of times in the synthetic world. We get not a single answer but a distribution of outcomes. Instead of “this will happen”, we can say “within this range, with this probability”.

MCTS — chain choices over time to find the best path. We stitch together the single-event distributions from Monte Carlo along the timeline and search the tree of strategies. Like AlphaGo reading moves in Go, it calculates which move to make at which moment to reach the goal.

We don’t stop at predicting the future. We find the best path to the goal, and work backwards to propose what to do, and when, to get on that path. You live through the coming phase in advance, and meet it with the response already in hand.

Not predicting the future, but living it in advance. That is Alea Capta’s simulation.