MENTAL MODEL · M3805

Active Inference

Learn & MetacognitionHigh supportCognitive Science

Included

Updated 2026-08-01

Loading revision record…

INTRODUCTION

English translation pending.

CORE DEFINITION

Active inference is a framework developed by Karl Friston and colleagues within the free energy principle. It proposes that the brain maintains a generative model of the world and continuously predicts sensory input; the discrepancy between prediction and input is prediction error, and the system minimizes it in two ways, by updating the model or by acting on the world so that input matches the prediction. The key qualifier is that perception and action are one process applied to different ends, and that the theory is a normative account of what an agent should do rather than a literal description of neural wiring.

SCAFFOLDING EFFECT

psychology

Reduce cognitive load

- Action as perception: Treat doing as a way of generating the data that will correct your model. - Error diagnosis: When surprised, ask whether to update the model or change the situation. - Learning loop: Start from an explicit expectation so that error becomes measurable rather than merely felt.

anchor

Anchor fast decisions

A generative model that predicts accurately reduces the need for costly correction, so agents are built to suppress surprise. Prediction error can be resolved either by revising beliefs, which is slow and can destabilize the model, or by acting to change the input, which is often cheaper. This explains why people seek confirming environments and why beliefs resist evidence: acting to make the world match the model is a valid error-reduction strategy, though not always the adaptive one.

MINIMUM ACTION

In progress 0/1

Practice this model in one real situation:

Before acting, write down what you expect to happen and how confident you are. 2. Compare the actual outcome with that prediction and measure the gap. 3. Decide explicitly whether to update the model or change the situation. 4. In new skills, act early and deliberately to generate the sensory data you lack. 5. Repeat the loop and watch whether errors shrink over iterations.

Check to track your progress (stored locally)

Learning progress0%

account_treeGenealogyexpand_more

menu_bookReferencesexpand_more

Source support: Explicit

link
en.wikipedia.orghttps://en.wikipedia.org/wiki/Free_energy_principleZH · Explicit
verified

RELATED MODELS