Abstract
This paper introduces and formalizes Noosemia, a novelcognitive-phenomenological phenomenon emerging from human interaction withgenerative AI systems, particularly those enabling dialogic or multimodalexchanges. We propose a multidisciplinary framework to explain how, undercertain conditions, users attribute intentionality, agency, and eveninteriority to these systems - a process grounded not in physical resemblance,but in linguistic performance, epistemic opacity, and emergent technologicalcomplexity. By linking an LLM declination of meaning holism to our technicalnotion of the LLM Contextual Cognitive Field, we clarify how LLMs constructmeaning relationally and how coherence and a simulacrum of agency arise at thehuman-AI interface. The analysis situates noosemia alongside pareidolia,animism, the intentional stance and the uncanny valley, distinguishing itsunique characteristics. We also introduce a-noosemia to describe thephenomenological withdrawal of such projections. The paper concludes withreflections on the broader philosophical, epistemological, and socialimplications of noosemic dynamics and directions for future research.