Post

JOPRO at the 6th Applied Active Inference Symposium

Jesse Parent and Avery Lim presented work on regulators, world models, and generative models at the Active Inference Institute's 6th Applied Active Inference Symposium, held online on November 12 and 13, 2026.

The Active Inference Institute held its 6th Applied Active Inference Symposium online on November 12 and 13, 2026. The program included pre-recorded talks, livestreamed panels and group sessions, interactive workshops, and keynotes. Participation was free.

Talk: From Minimal Model Systems to World Models

Jesse Parent and Avery Lim presented “From Minimal Model Systems to World Models: Boundary-Relative Distinctions for Factoring Generative Models” as a pre-recorded talk. The work begins from the good regulator idea in cybernetics, in which a regulator’s model is a reduced representation of the system it regulates. It asks how that idea carries over to the generative models used in active inference.

The contribution appears in the symposium’s 2026 Presenter Program, published on Zenodo, as “From Minimal Regulators to World Models.” Both authors are listed with their JOPRO affiliation. The slides are on Zenodo, and the talk has its own publication page.

Abstract. The good regulator theorem holds that a successful regulator must embody a model of the system it regulates, and recent work suggests that even minimal regulators can carry informative world models. We take up a prior structural question: which distinctions should organize the channels through which a model’s content arrives? We compare four modeling starting points, Turing’s unorganized machines, Ashby’s homeostat, autopoiesis, and active inference, for the distinct elements and constraints each makes salient. From this comparison we derive two candidate sets of primitives. Causal Standpoint Routes (CSR) classify a causal relation by whether its source and target lie inside or outside a declared system boundary, and by whether the relation is immediate or mediated relative to declared operations and interfaces, which gives eight route classes. Allostatic Kinds distinguish four directions of coupling between a regulator and its environment and relate each to the essential variables and viable bounds that give it regulatory significance. Unorganized machines offer a particularly direct minimal model for CSR, and the homeostat for Allostatic Kinds. We outline how both could inform the factorization of generative models in active inference, for instance by separating external sources, intermediaries, and sensory evidence, and we propose an empirical program testing whether CSR-guided training or steering improves held-out causal reasoning in language models against matched controls. The correspondences among the four frameworks remain open, and we present the primitives as testable hypotheses.

Cite the talk. Parent, J., & Lim, A. (2026, November 12). From Minimal Model Systems to World Models: Boundary-Relative Distinctions for Factoring Generative Models. JOPRO. Active Inference Symposium. https://doi.org/10.5281/zenodo.24180907

Discussion: World models in natural and artificial intelligence

Bradly Alicea of the Orthogonal Research and Education Lab, a member of the 2026 organizing team, led an interactive workshop on the special issue World models in natural and artificial intelligence. The session summarized the issue and opened it to group discussion.

Further reading