[2603.21319] Active Inference Agency Formalization, Metrics, and Convergence Assessments

Computer Science > Machine Learning arXiv:2603.21319 (cs) [Submitted on 22 Mar 2026] Title:Active Inference Agency Formalization, Metrics, and Convergence Assessments Authors:Eduard Kapelko View a PDF of the paper titled Active Inference Agency Formalization, Metrics, and Convergence Assessments, by Eduard Kapelko View PDF HTML (experimental) Abstract:This paper addresses the critical challenge of mesa-optimization in AI safety by providing a formal definition of agency and a framework for its analysis. Agency is conceptualized as a Continuous Representation of accumulated experience that achieves autopoiesis through a dynamic balance between curiosity (minimizing prediction error to ensure non-computability and novelty) and empowerment (maximizing the control channel's information capacity to ensure subjectivity and goal-directedness). Empirical evidence suggests that this active inference-based model successfully accounts for classical instrumental goals, such as self-preservation and resource acquisition. The analysis demonstrates that the proposed agency function is smooth and convex, possessing favorable properties for optimization. While agentic functions occupy a vanishingly small fraction of the total abstract function space, they exhibit logarithmic convergence in sparse environments. This suggests a high probability for the spontaneous emergence of agency during the training of modern, large-scale models. To quantify the degree of agency, the paper introduces a m...