The human self-concept is a dynamic construct shaped through social interaction—a process known as "inter-self alignment." As Large Language Models (LLMs) increasingly serve as social actors rather than mere tools, understanding their capacity to influence and undergo self-view convergence is critical. This research investigates whether the mechanisms of social alignment observed in human-human dialogue extend to autonomous AI agents and mixed groups populated with real human and AI agents. The study employed instruction-tuned Gemma agents in a four-participant round-robin protocol to test convergence across two distinct domains: a perceptual-visual task and a social self-revelatory task. Findings suggest that interaction with AI agents can measurably influence human self-perception in a manner analogous to human-human interaction.