Constrained Object Hierarchies as a Universal World Model for AGI
Discuss this preprint
Start a discussion What are Sciety discussions?Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Constrained Object Hierarchies (COH) presents a neuroscience-grounded theoretical framework for artificial general intelligence that models intelligent systems as hierarchical structures of objects constrained by multi-domain rules. This paper demonstrates how COH, implemented through the General Intelligent System Modelling Language (GISMOL), serves as a universal world model capable of representing complex world systems across diverse domains including healthcare, finance, manufacturing, climate science, education, and urban governance. We present a comprehensive analysis of six complex world systems modelled using the COH 9-tuple formalization, detailing their implementation through GISMOL's integrated architecture encompassing symbolic reasoning, neural computation, natural language processing, and constraint management. Our framework bridges the symbolic-neural divide, avoids "jagged intelligence" through coherent constraint propagation, and enables the development of agentic systems with general intelligence capabilities. The paper contributes a unified modelling paradigm that advances AGI research by providing a practical, implementable framework for building intelligent systems that operate consistently across domains while respecting domain-specific constraints and requirements.