Bounded Agency: An Entropy-Constrained Calculus for Sparse Semantic Agents

• Apeksha Bhuekar

  Author

This paper proposes a formal framework for AI agents that unifies semantic reasoning with resource-aware control. Agents act via sparse policies over structured semantic fields, bounded by entropy and sparsity budgets. We define a typed operational semantics, prove soundness and stability, and derive a sparse free-energy objective with phase transitions. The calculus is categorically structured, maps to unistochastic dynamics, and compiles to executable policies with verified runtime bounds. This framework enables the design of AI systems that are interpretable, resource‑aware, and verifiably constrained, with applications in autonomous systems, decision support, and embodied AI. The result is a foundation for thermodynamically‑plausible agent design.

• PDF

Copyright (c) 2026 International Journal of Intelligent Systems and Data Science

This work is licensed under a Creative Commons Attribution 4.0 International License.