Jesse Parent & Avery Lim: “From Minimal Model Systems to World Models” @ActiveInference
Jesse Parent & Avery Lim: “From Minimal Model Systems to World Models”  @ActiveInference
Uploaded November 2026 | Updated November 2026, 21 hours ago
Prerecorded presentation at the 6th Applied Active Inference Symposium (November 12–13, 2026), hosted by the Active Inference Institute. Online and free.

Jesse Parent & Avery Lim: “From Minimal Model Systems to World Models: Boundary-Relative Distinctions for Factoring Generative Models”

A good regulator’s model is a reduced representation that preserves useful variety (Alicea, Hough, Nelson and Parent, 2026). For active inference, this raises a design question: before any content is learned, which distinctions between sources and channels should an agent’s generative model keep apart?
We propose two taxonomies drawn from minimal regulators. Causal Standpoint Routes, derived from Turing’s networked machine, classify any influence across a boundary by where it starts, where it ends, and whether it passes through another system. The four immediate routes are the Markov blanket’s own dependencies. Four mediated routes cover reports relayed through another agent, actions delegated to tools, chains of external causes, and stored state. Allostatic Kinds, derived from Ashby’s homeostat, describe what an agent regulates through each coupling: set points as preferred outcomes, the sign of each coupling, and allostasis when preferences shift. Both are relative to where the boundary is drawn, which is what lets them carry across domains.
The routes add no new inference machinery. They propose how to factor the model so that existing machinery, such as epistemic value and novelty, targets the right channels: probing an intermediary, for example, rather than only the world behind it. Merging channels predicts specific failures. Internally generated content treated as sensory fits the precision account of hallucination. In language models, prompt injection, invented sources and long-context drift look like the same breakdowns.
We outline two tests: pymdp agents factored by route, compared under environmental shift with agents factored by matched irrelevant distinctions; and whether reinforcing these distinctions in LLM agents reduces source confusion more than unrelated errors. We present proposals and study designs, not results, and put one open claim to the audience: that mediated routes correspond to nested blankets and deeper levels of the generative model.

Symposium program: https://2026symposium.activeinference.institute/
Abstract book (Zenodo): zenodo.org/records/23986505
Livestreams: Day 1 youtube.com/watch?v=-v4IaFY9MZI · Day 2 youtube.com/watch?v=ccze9ay8D0o
All Symposium videos (playlist): youtube.com/playlist?list=PLO2noumu0LRk
Previous Symposia: https://symposium.activeinference.institute/

Active Inference Institute information:
Website: https://activeinference.institute/
X: https://x.com/InferenceActive
Discord: https://discord.activeinference.institute/
Donate: https://donate.activeinference.institute/
YouTube: youtube.com/c/ActiveInference
Active Inference Livestreams: https://video.activeinference.institute/
Jesse Parent & Avery Lim: “From Minimal Model Systems to World Models”Active Inference Institute ~ 2026 Quarterly Roundtable #2ActInf Textbook Group ~ Cohort 7 ~ Session 5 (Chapter 2 +++ Open Discussion) ~ 8/12/2024Fundamentals of Active Inference (Chapter 8, Session 37) August 18, 2026ActInf GuestStream 131.1: Causal Architecture Dynamics Prior to Arrival of Self (Pigozzi and Levin)Andrew Pashea: “Active Inference Tutorial: Perspectives, Correlates, and Future Directions”ActInf GuestStream 106.1 ~  Advances and Challenges in Foundation Agents Bang Liu et al.ActInf OrgStream 009.1 ~ The Value of Open Source Software (Hoffmann, Nagle, Zhou)ActInf ModelStream 019.1: Active Inference in Discrete State Spaces from First Principles (P. Kenny)Generalized Coordinates in RxInfer.jlActInf Textbook Group ~ Cohort 9 ~ Session 5 (Chapter 4) ~ 9/18/2025Anna Pereira: “A First Principle Perspective for AI Safety”
Active Inference Institute |

Jesse Parent & Avery Lim: “From Minimal Model Systems to World Models”

SHARE TO X SHARE TO REDDIT SHARE TO FACEBOOK WALLPAPER