LAXCORP RESEARCH · DEEPFIELD PROGRAM

Glossary

Agent-based modeling

Simulating a system from the bottom up, as many interacting agents following simple rules, so that collective outcomes emerge from behavior rather than being assumed in advance.

Base rate

How often something happens in general, before the specifics of the case at hand are considered. Forecasts that ignore base rates drift toward story and away from statistics.

Bayesian updating

The discipline of revising a probability as evidence arrives, in proportion to how strongly the evidence favors one explanation over another.

Black swan

Taleb’s term for an event so far outside prior experience that models trained on the past cannot see it coming. Respecting the possibility is part of honest forecasting.

Brier score

A way of scoring probabilistic predictions that rewards being both correct and well calibrated, and penalizes confident wrong answers most of all.

Calibration

Whether stated confidence matches observed frequency. If a set of forecasts is made at 80 percent confidence, roughly 80 percent should come true. Calibration, not raw accuracy, is the honest measure of a forecaster.

Causal inference

The study of cause and effect, as distinct from correlation. It asks what would happen if something were changed, not merely what tends to occur alongside what.

Change-point detection

Methods for identifying the moment a process shifts from one regime to another, separating a real change from ordinary noise as data arrives.

Complex adaptive system

A system of many interacting parts that learn and adapt, where the behavior of the whole cannot be read off from any single part. Economies, ecosystems, and markets are the canonical examples.

Conformal prediction

A technique that turns a model into an interval with a guaranteed coverage rate, so a prediction carries an honest error bar rather than false precision.

Contagion

How stress propagates through a network of connected institutions or markets. Studied with models first developed in epidemiology, where the mathematics of spread was worked out.

Counterfactual

What would have happened under a different action or condition. The test of causal understanding is the ability to answer counterfactual questions, not merely to fit the data.

Critical slowing down

A signature that often precedes a tipping point in a complex system. As the system nears a transition, it recovers from disturbances more slowly and its fluctuations grow wider.

Early-warning signal

A measurable statistical change, such as rising variance, slowing recovery, or tightening correlation, that tends to precede a transition in a complex system.

Epistemic and aleatoric uncertainty

The two kinds of not knowing: what could be learned with more information, and what is irreducibly random. They look alike in a forecast and warrant different responses.

Extremizing

A finding from forecasting research: aggregated forecasts are often too timid, and pushing them modestly away from fifty percent improves accuracy. Crowds hedge; the truth rarely does.

Fermi estimation

Breaking an intractable question into smaller questions that can each be estimated, then recombining them. Named for Enrico Fermi, who estimated almost anything this way.

Goodhart’s law

When a measure becomes a target, it ceases to be a good measure. The reason a single headline metric corrupts, and a full distribution of evidence is preferred.

Information cascade

When people infer from the choices of others rather than from their own information, so a belief spreads because it is spreading, not because it is true.

Nowcasting

Estimating the present before official figures arrive, using high-frequency signals such as shipping, energy use, or search activity as proxies for slower statistics.

Prospect theory

A model of how people actually weigh risk and loss, showing that human decisions depart from pure rationality in systematic, predictable ways.

Reflexivity

The feedback by which observing or describing a system changes it. In markets, a widely held belief can help bring about the outcome it predicts.

Regime

A period during which a system behaves according to one stable set of relationships. Detecting when one regime gives way to another is central to anticipating change.

Superforecasting

The finding, from large forecasting tournaments, that certain habits of thought produce consistently better predictions than expertise or intuition alone.

Tipping point

The threshold beyond which a system shifts to a qualitatively different state, often abruptly, and from which return is difficult or impossible.

Transfer entropy

An information-theoretic measure of how much knowing one variable’s past reduces uncertainty about another’s future, used to infer the direction of influence.