Before Trust: AI Literacy and Purpose-First Mode Selection as the Missing Foundation of Human–AI Interaction

Artificial intelligence research has increasingly focused on trust, controllability, transparency, and governance. This paper argues that a more fundamental layer precedes all of these: users often lack a clear mental model of what AI is, how it should be used, and how to manage the interaction itself — a gap documented across recent surveys of user mental models, agentic controllability, and trust-centered interface design, but rarely addressed directly as a prerequisite skill. Building on emerging work on criteria-based interaction mode selection and on interface friction as a countermeasure to over-reliance, we propose AI literacy as the foundational layer of human–AI interaction, operationalized through purpose-first Mode Selection — a taxonomy of Work, Brainstorm, and Companion modes — and Functional Minimalism, a design principle intended to preserve critical distance in high-risk or emotionally salient interactions. We position both proposals explicitly against existing literature, distinguish our contribution from closely related prior work, and identify the empirical validation this framework still requires.