MENTAL MODEL · M13141

Aleatory vs Epistemic Uncertainty

SystemsHigh supportRisk Epistemology

Included

Updated 2026-08-17

Loading revision record…

INTRODUCTION

English translation pending.

CORE DEFINITION

The distinction separates uncertainty about a genuinely random process, called aleatory, from uncertainty caused by incomplete knowledge, called epistemic. Aleatory uncertainty is irreducible: no amount of study removes the spread in the outcome of a fair die or a radioactive decay, and it is handled with probabilities, redundancy, and buffers. Epistemic uncertainty is reducible: it reflects ignorance about a fixed but unknown quantity, such as a model parameter or a reservoir's true capacity, and it shrinks with measurement, experiment, or better modelling. The two are often conflated with Frank Knight's 1921 contrast between measurable risk and unmeasurable uncertainty, which is related but not identical, and in practice most real problems mix both.

SCAFFOLDING EFFECT

psychology

Reduce cognitive load

- Uncertainty sorting: Label each unknown as reducible by study or irreducible, before deciding how to treat it. - Research budget: Spend measurement effort only on the epistemic part you can actually shrink. - Buffer design: Handle the irreducible part with redundancy, staged exposure, and explicit safety margins.

anchor

Anchor fast decisions

The two kinds respond to different actions, so treating them alike wastes effort and creates blind exposure. Measurement reduces epistemic uncertainty because the unknown quantity is fixed and the data carry information about it. Nothing reduces aleatory uncertainty, because the variation is generated by the process itself, so the only available responses are to absorb it through redundancy, to limit exposure, or to price it. Misclassifying a reducible unknown as random buys expensive insurance against a question that study would have answered.

MINIMUM ACTION

In progress 0/1

Practice this model in one real situation:

List the uncertainties that matter to the decision you are about to make. 2. For each one, ask whether more data could in principle narrow it. 3. Spend research effort only on the uncertainties that pass that test. 4. Handle the irreducible ones with redundancy, staged exposure, or explicit margins. 5. Revisit the classification later, since some unknowns turn out to be learnable.

Check to track your progress (stored locally)

Learning progress0%

account_treeGenealogyexpand_more

menu_bookReferencesexpand_more

Source support: Explicit

link
github.comhttps://github.com/kcchien/model-thinkingZH · Explicit
verified

RELATED MODELS