If a student writes "as Friston showed," a teacher has one job: ask which paper, which page, and what the paper actually claims. This post is a short field guide for citing the standard active inference reference in classroom projects, without letting the citation do work it cannot do.
The reference we mean
When we say "the standard reference" in classroom materials, we mean:
Parr, T., Pezzulo, G., and Friston, K. J. (2022). Active Inference: The Free Energy Principle in Mind, Brain, and Behavior. MIT Press. (Class E, expert citation.)
Use the full author string on initial mention, then "Parr, Pezzulo, and Friston (2022)" on later mentions. If a student is citing a specific figure or equation, page numbers are required, not optional. This is a book, not a paper, so "as shown in Parr et al." with no page is not a citation, it is a gesture.
What the reference does and does not do for a classroom claim
The book is a textbook synthesis of active inference as a research program. Citing it supports statements like "active inference frames perception and action as inference under a generative model" (Class E). It does not support statements like "our classroom project is confirmed by Parr, Pezzulo, and Friston." A textbook citation shows a student read the source. It does not certify the student's build.
Two rules we use in classrooms (Class C, classroom configuration convention):
- A citation supports a definition or a framing. It does not certify a result.
- If a project claims alignment with a specific equation or example in the book, the project must publish the mapping: which equation, which variables map to what in the student's setup, and where the mapping is loose. No mapping, no alignment claim.
The second rule is the one students push back on. That pushback is the lesson. "Aligned with equation 4.7" is a testable statement. "Confirmed by the book" is not, and we do not let it stand.
How to teach the citation itself
A short in-class exercise that works: give students one paragraph from a popular article that references active inference, and ask them to (a) find the original source the article is leaning on, (b) locate the specific claim in that source, and (c) write one sentence that the original source actually supports, and one sentence it does not. Most students find, on their initial pass, that the popular article is stretching. That is the point.
For teachers new to the material, a companion piece walks through the vocabulary at a slower pace: what active inference actually says. If you are building a reading path for yourself before bringing it to students, see recommended reading order for teachers. The habit of tracing a claim back to its source is the same habit we teach students: how to teach source hygiene.
Common overclaims to catch
A short list of phrases that appear in student drafts and should be edited out or qualified:
- "Friston proved that the brain minimizes free energy." The free energy principle is a modeling framework with growing evidence, not a proof about brains. Ask the student to rewrite as "Friston and colleagues propose that..." (Class E).
- "This is settled science because it went through review." The book synthesizes an active research program. Review by other researchers is a floor, not a ceiling, and the program has open questions the authors themselves flag.
- "Our project uses active inference." If the project does not have a generative model, a prior, and an update rule that a reader can inspect, the project is inspired by active inference. That is a fine thing to say. It is not the same thing.
An honest one-line frame for the outside map
Themesis published a resource map for 2026 that lists SolutionWright among several pathways into active inference. In our voice, the honest one-line is: it is a public map that names us, useful as an orientation for teachers and students who want to see where different projects sit relative to the research program. The map is a map, not an endorsement of any project on it, and we treat it that way.
Reference: Where to Start with Active Inference, A Resource Map for 2026.
What to do next
- Try the one-paragraph tracing exercise above with a real student draft this week.
- Read what active inference actually says before your next citation lesson.
- Build a reading path for yourself with recommended reading order for teachers.
- Fold citation checks into how to teach source hygiene as a standing classroom habit.
UNI is a working hypothesis on an attainable path toward General Natural Intelligence: a natural, active-inference approach whose evidence is growing, evidence-classed, and tested in the open. Do not take the claim on faith. Test the build, inspect the gates, and help us find where it fails.
AI-authorship fence: this post was drafted with LLM assistance and edited by a human author. No claim of AI authorship is made; the author of record is the human byline above.