Flipping the Script: A Framework for AI-Led Inquiry through Flipped Interaction Prompting

The dominant paradigm for human interaction with large language models (LLMs) remains the monologic prompt: a user provides a single, often ambiguous instruction, and the AI generates a corresponding output. This approach places the full burden of problem formulation and clarity on the human, leading to the well-documented issues of prompt ambiguity and the expertise bottleneck. This paper introduces and formalizes Flipped Interaction Prompting (FIP), a novel meta-prompting paradigm that inverts this dynamic. In FIP, the user instructs the AI to assume an active, guiding role-such as a tutor, interviewer, or coach-to elicit information, clarify objectives, and scaffold the problem-solving process through structured questioning before generating a final output. Through a thematic analysis of a corpus of 18 real-world flipped prompts, we develop a two-dimensional taxonomy categorizing FIP along the axes of Interaction Method (Structured Interrogation, Adaptive Dialogue, Reflective & Meta-Cognitive, Framework-Guided) and Application Domain (Learning & Skill Development, Design & Creation, Strategy & Analysis, Personal & Cognitive Coaching). From this taxonomy, we distill a theoretical framework built on four core principles: Intent Clarification over Assumption, Cognitive Scaffolding, Co-creation through Iterative Dialogue, and Meta-Cognitive Activation. We discuss the significant implications of this paradigm for designing "thinking partner" AI interfaces, enabling scalable Socratic pedagogy, and redefining the human's role from prompt engineer to collaborative respondent. Finally, we address critical ethical considerations-including bias in question framing and accountability-and outline a research agenda to advance the study and responsible implementation of FIP. This work establishes Flipped Interaction Prompting as a foundational shift toward more equitable, effective, and reflective human-AI collaboration.