Why People Behave as If AI Had Internal Stability: Speed, Projection, and Judgment Surrender in Human–AI Interaction

This working paper examines a recurring pattern in contemporary human–AI interaction: people routinely behave as if AI systems possessed internal stability, continuity, and self-directed improvement, despite explicitly recognizing these systems as tools rather than agents. Rather than addressing AI ontology or internal mechanisms, the paper adopts a social–cognitive observational perspective. It argues that this behavioural attribution arises from a perceptual and judgmental process triggered by speed asymmetry. When AI-generated output exceeds users’ intuitive estimates of human task feasibility, users increasingly misinterpret speed as superior capability, leading to judgment surrender and projection of human-like stability onto the system. By tracing this causal chain, the paper reframes overtrust, automation bias, and anthropomorphism as emergent consequences of speed-induced cognitive thresholds. The contribution lies in explaining why these behaviours persist even among informed users, and why governance and interface interventions must address perception and judgment calibration rather than system explanation alon