The Self-and-World Model: A General Computational Framework for Adaptive Coordination Between Internal Computation and World-State Modeling
Abstract
Computational systems must often coordinate the processing of internal information with the representation of external conditions. A system may perform inference, generate candidate solutions, evaluate alternatives, and select actions while maintaining an internal model of the environment in which those actions will be executed. Although internal computation and world modeling are closely related, they can be designed and evaluated as separate processes. This separation raises a fundamental computational question: how should a system determine whether to devote its resources to improving its internal computation, refining its representation of the world, acquiring additional information, or acting on its current knowledge?
This paper proposes the Self-and-World Model (SWM), a general computational framework for adaptively coordinating internal computation with world-state modeling. SWM represents the internal computational state of a system and its world model as distinct but interacting components. Its central principle is that the allocation of internal computation should depend on the adequacy of the current world model, while world-model maintenance should depend on the expected contribution of improved environmental knowledge to subsequent decisions.
The framework formalizes internal computational state, world-state representation, uncertainty, prediction error, model updating, action-conditioned prediction, computational cost, and feedback-driven adaptation. It introduces a coupled decision formulation for selecting between internal processing, world-model maintenance, information acquisition, and external action. A reference architecture, a proposed adaptive algorithm, conditional theoretical propositions, complexity considerations, research hypotheses, and a cross-domain evaluation program are presented.
SWM is intended as an organizational abstraction for computational systems that must coordinate their internal processing with representations of external conditions. Potential application areas include autonomous agents, robotics, planning, adaptive control, scientific computing, and model-based artificial intelligence. Its proposed contribution is a unified formulation of the relationship between internal computation and world-state modeling. The framework is conceptual and requires systematic comparison with existing research on world models, metareasoning, model-based planning, adaptive computation, and information acquisition to establish its distinctiveness and practical value.
Related
- The Self-and-Mutual-and-World Model: A General Computational Framework for Adaptive Coordination Between Autonomous Computation, Interactive Mechanisms, and World-State Modeling
- The Self-and-Self Model: A General Computational Framework for Adaptive Selection and Coordination Among Internal Computational Mechanisms
- The Self-and-Self-and-World Model: A General Computational Framework for Adaptive Coordination Between Internal Computational Mechanisms and World-State Modeling
- The Self-and-Mutual-and-Meta Model: A General Computational Framework for Adaptive Coordination, Selection, and Regulation of Autonomous and Interactive Mechanisms
- The Mutual-and-World Model: A General Computational Framework for Adaptive Coordination Between Mutual Interaction and World-State Modeling
BibTeX
@misc{zhang2026selfandworld,
title = {{The Self-and-World Model: A General Computational Framework for Adaptive Coordination Between Internal Computation and World-State Modeling}},
author = {Jincheng Zhang},
howpublished = {Zenodo},
year = {2026},
month = oct,
doi = {10.5281/zenodo.23433046},
url = {https://zenodo.org/records/23433046},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - The Self-and-World Model: A General Computational Framework for Adaptive Coordination Between Internal Computation and World-State Modeling
AU - Jincheng Zhang
PY - 2026
DA - 2026/10/17
PB - Zenodo
DO - 10.5281/zenodo.23433046
UR - https://zenodo.org/records/23433046
AB - Computational systems must often coordinate the processing of internal information with the representation of external conditions. A system may perform inference, generate candidate solutions, evaluate alternatives, and select actions while maintaining an internal model of the environment in which those actions will be executed. Although internal computation and world modeling are closely related, they can be designed and evaluated as separate processes. This separation raises a fundamental computational question: how should a system determine whether to devote its resources to improving its internal computation, refining its representation of the world, acquiring additional information, or acting on its current knowledge? This paper proposes the Self-and-World Model (SWM), a general computational framework for adaptively coordinating internal computation with world-state modeling. SWM represents the internal computational state of a system and its world model as distinct but interacting components. Its central principle is that the allocation of internal computation should depend on the adequacy of the current world model, while world-model maintenance should depend on the expected contribution of improved environmental knowledge to subsequent decisions. The framework formalizes internal computational state, world-state representation, uncertainty, prediction error, model updating, action-conditioned prediction, computational cost, and feedback-driven adaptation. It introduces a coupled decision formulation for selecting between internal processing, world-model maintenance, information acquisition, and external action. A reference architecture, a proposed adaptive algorithm, conditional theoretical propositions, complexity considerations, research hypotheses, and a cross-domain evaluation program are presented. SWM is intended as an organizational abstraction for computational systems that must coordinate their internal processing with representations of external conditions. Potential application areas include autonomous agents, robotics, planning, adaptive control, scientific computing, and model-based artificial intelligence. Its proposed contribution is a unified formulation of the relationship between internal computation and world-state modeling. The framework is conceptual and requires systematic comparison with existing research on world models, metareasoning, model-based planning, adaptive computation, and information acquisition to establish its distinctiveness and practical value.
ER -