Cognitive capacities and autonomous agency are present across all known life, from humans to bacteria. Recent advances in AI highlight a stark contrast: computational systems can solve complex tasks, yet they lack the self-determined purposefulness of natural embodied agents. This motivates our central question: How does agency, the capacity to act on one’s own needs and goals, give rise to intelligence?
We propose a coordinated research program to understand how organizational, informational, and thermodynamic structures jointly enable the emergence of autonomous agency and the semantic information that supports intelligent behavior. Our approach integrates three complementary programmatic foci:
(1) Conceptual and philosophical clarification of agency, autonomy, precariousness, and sense-making;
(2) Mathematical and computational models of organizational closure, viability, and individuality; and
(3) Information-theoretic analysis of how agents generate and use viability-relevant semantic information.
Together, these efforts form the foundation of the Virtual Institute for the Physics of Agentic Intelligence (VIPAI), an international collaboration linking physics, philosophy, complex systems, information theory, and the life sciences. Through in-person and virtual meetings, the project will produce new theoretical frameworks, open modeling tools, and integrative concepts that clarify the conditions under which agentic and intelligent behavior arises in natural and, possibly, in artificial systems.
By unifying conceptual analysis with formal modeling, this project aims to reshape scientific understanding of agency, meaning, and intelligence, and to establish the foundations of a new interdisciplinary field.
