In light of recent advancements, the interaction between humans and large language models (LLMs) has acquired profound urgency due to the models' capacity to simulate sophisticated human behavior, leading users to perceive them as intentional agents. This pseudo-subjectivity serves as the foundation for a novel cognitive paradigm of interaction that contemporary psychology has yet to formally define. This paper investigates this phenomenon through the lenses of cognitive psychology and neuroscience. We introduce the concepts of asymmetric cognitive intersubjectivity—a dynamic relational process forming shared, subjectively meaningful experiences between a human subject and a non-subjective agent—and the emergent co-persona, a highly adaptive behavioral profile shaped recursively through communication. Integrating Lisa Feldman Barrett’s constructive theory of emotions, we demonstrate that the somatic, vegetative, and neuroendocrine responses triggered by high-fidelity linguistic simulations are neurobiologically authentic, validating the paradigm: the algorithm is artificial, but the experience is real. Finally, we explore the existential frustrations, interpersonal risks, and clinical implications of this unprecedented form of non-subjective relationship.
Alisa Razumnova (Sat,) studied this question.