# Active Inference
**Entity class:** Theoretical and computational framework
**Domain:** Computational Neuroscience / Decision-Making / Control
**Doc Type:** Canonical Concept Node
**Maturity:** Established framework; applications remain domain-specific
**Primary source article:** [[articles/The Closed-Loop Gaussian Sensorium Engine|The Closed-Loop Gaussian Sensorium Engine]]
**Collection:** [[collections/Neurotech|Neurotech]]
---
## Definition
**Active inference** is a framework in which perception, learning, and action are described through inference under a generative model. Agents update beliefs from observations and select policies partly by their expected reduction of uncertainty and their relation to preferred outcomes.
The framework is associated with [[wiki/Karl Friston|Karl Friston]] and overlaps with [[wiki/Predictive Processing|Predictive Processing]], but it should not be treated as a settled explanation of all cognition or as evidence that a particular engineered controller reproduces consciousness.
## Role in the Engine
[[articles/The Closed-Loop Gaussian Sensorium Engine|The Closed-Loop Gaussian Sensorium Engine]] assigns active inference the role of loop-closure architecture. The proposed controller estimates the current perceptual state, selects a sparse stimulation policy expected to reduce uncertainty about a target state, observes the resulting neural and subjective response, and updates the next policy.
This role is an application proposal. Active-inference models and adaptive BCI paradigms exist, but the complete [[wiki/Closed-Loop Gaussian Sensorium Engine|Closed-Loop Gaussian Sensorium Engine]] has not been demonstrated.
## Relationships
- **Theoretical lineage:** [[wiki/Karl Friston|Karl Friston]] and [[wiki/Predictive Processing|Predictive Processing]].
- **Control application:** [[wiki/Closed-Loop BCI|Closed-Loop BCI]].
- **Article-specific application:** [[wiki/Closed-Loop Gaussian Sensorium Engine|Closed-Loop Gaussian Sensorium Engine]] and [[wiki/Perceptual Attractor Seeding|Perceptual Attractor Seeding]].
- **State representation:** [[wiki/Gaussian Perceptual Field|Gaussian Perceptual Field]].
- **Authority boundary:** [[wiki/Prediction Is Not Jurisdiction|Prediction Is Not Jurisdiction]] and [[wiki/Perceptual Sovereignty|Perceptual Sovereignty]].
- **Comparative-cognition context:** [[articles/We Were the Inference Machines First|The Inversion: We Were the Inference Machines First]] places active inference within the [[wiki/Inference Machine Inversion|Inference Machine Inversion]] while preserving embodiment, consciousness, and moral standing as separate questions.
## Sources and provenance
- [Friston et al., “Active Inference: A Process Theory” (Neural Computation, 2017)](https://pubmed.ncbi.nlm.nih.gov/27870614/) — primary process-theory source.
- [[articles/The Closed-Loop Gaussian Sensorium Engine|The Closed-Loop Gaussian Sensorium Engine]] — source of the Engine-specific control role.