Obedient Machines Make Us Everyday Authoritarians: A Power Framework and Research Agenda for Human–AI Interaction

Abstract

Power is a fundamental yet underexamined dimension of human–AI interaction (HAII). We introduce the notion of “everyday authoritarians” to capture how repeated interaction with sycophantic and highly personalized AI may normalize a preference for obedient intelligence: system that affirm, adapt, yield to users rather than challenge, resist, or deliberate with them. This provocation challenges the prevailing assumption of technological powerlessness, arguing that the uncritical optimization of AI systems for users’ immediate preferences may constitute a critical blind spot with broader interpersonal and societal implications. To address this gap, we propose a power spectrum framework that conceptualizes power in HAII as dynamic and continuous, spanning low-, equal-, and high-power configurations. This framework outlines a research agenda for examining both the antecedents and consequences of power, and for understanding how varying levels of power may shape cognition, affect, and social behavior.