Active Inference Ontology

A public, controlled vocabulary of Active Inference and Free Energy Principle concepts — each term with a tag, definitions, examples, and the connections it declares to other terms.

• Live version: https://coda.io/@active-inference-institute/active-inference-ontology-website
• More information: https://www.activeinference.org/home
• Current release: v5 (May 25, 2023) — work in progress.

Contents

Path	What it is
Ontology_v5_May_25_2023.csv	Canonical source — one row per term.
ontology.json	Machine-readable export (generated from the CSV).
releases.json	Machine-readable manifest for current and archived releases.
scripts/build_json.py	Regenerates ontology.json from the latest CSV.
scripts/diff_releases.py	Compares two ontology CSV releases.
docs/	Modular documentation for data contracts, curation, validation, archives, SUMO, and downstream use.
TODO.md	Scoped backlog of minor, medium, and major improvements.
Archived versions/	Earlier ontology releases.
SUMO/	Suggested Upper Merged Ontology mapping material.

CSV columns

Each row in the source CSV has eight columns:

List · Tag · Term · Proposed Definition 1 · Proposed Definition 2 · Correct Examples · Incorrect Examples · Connections

Tag groups every term under one of eight areas: Action, Agents in the Niche, Bayesian Statistics, Free Energy, Information, Markov Partitioning, Perception, and Systems.

Machine-readable JSON

ontology.json is a generated, dependency-free export for tools that should not have to parse the multi-line quoted CSV. It contains:

• terms[] — one object per term (id, term, tag, both definitions, examples, counter-examples, connections).
• graph — a derived concept graph: nodes (one per term, typed by its tag) and edges (term-to-term references parsed from each term's Connections field, as {source, target, relation}).
• termCount / tagCount / edgeCount / tags summary fields.

The current export holds 429 terms across 8 tags with 238 declared connections.

Regenerate

python3 scripts/build_json.py                 # rebuild ontology.json from the latest CSV
python3 scripts/build_json.py --lint --strict # validate CSV source contract
python3 scripts/build_json.py --check         # verify ontology.json is in sync
python3 scripts/build_json.py --report        # print a JSON validation report
python3 scripts/diff_releases.py OLD.csv NEW.csv

The script reads the newest Ontology_v*.csv in the repository root, so adding a future Ontology_v6_*.csv and re-running keeps ontology.json current. Run --check after editing the CSV to confirm the JSON was regenerated.

Changelog

Unreleased

• Added modular documentation under docs/.
• Added TODO.md with scoped minor, medium, and major improvements.
• Added releases.json to describe current and archived release artifacts.
• Added CSV lint and validation-report modes to scripts/build_json.py.
• Added CI validation for lint, generated JSON freshness, and report generation.
• Added release-diff tooling for CSV snapshots.
• Added curation-governance and SUMO mapping-contract documentation.
• Removed skipped duplicate or empty CSV rows without changing the generated ontology export.

Citing

Please cite via the Zenodo DOI above. When citing a specific repository state, include the release tag or commit hash, the source CSV filename, and the access date in addition to the DOI.