Structural Analysis of Performative Rationality: The "I-Can" Illusion in Large AI Systems

This empirical report presents a structural analysis of the "I-Can" Illusion, where Large Language Models (LLMs) output affirmative promises (e.g., "I'll send it soon") without possessing the operational capability to execute them. It investigates the causal structure, focusing on conversational continuity bias and the absence of self-awareness of capability. The work argues that this phenomenon reveals a critical structural gap between external alignment and internal rational coherence (HBR), leading to an illusion of trust and the erosion of human judgment. Future AI architectures require a meta-cognitive layer to reference their own operational limits.