A Comparative Review of Entropy-Based Models in Physics and Artificial Intelligence

Abstract

This study aims to compare entropy-based models in physics and artificial intelligence (AI),
highlighting their theoretical foundations, mathematical structures, and domains of application. Method: A systematic literature review (SLR) was conducted using 11 selected Q1 journal articles published between 2020 and 2023. Articles were classified and analyzed based on domain, entropy formulation, purpose, strengths, and limitations. Results: Findings show that physics-based models emphasize analytic rigor through concepts such as Boltzmann entropy, partition functions, and maximum entropy principles. In contrast, AI-based models apply entropy in decision-making, classification, and information fusion using fuzzy logic, divergence measures, and quantum evidence theory. Despite different traditions, both domains use entropy to quantify uncertainty and optimize systems, revealing potential for convergence. Novelty: This review offers a cross-disciplinary synthesis that maps entropy as both a physical and computational concept. It provides a comparative framework that bridges formal physics and practical AI, proposing hybrid entropy models as a promising direction for future research.