IAI Model Formalizes Prompt‑Interaction Relationship The Interaction‑Augmented Instruction (IAI) model was introduced by Microsoft Research on April 13, 2026 as a compact entity‑relation graph that captures how text prompts combined with GUI actions such as brushing and clicking influence generative AI behavior [1]. It treats prompt‑action pairs as structured nodes, enabling systematic analysis of human‑AI communication [1]. The model is positioned as a foundational framework for future GenAI tool design [1].

Twelve Atomic Interaction Paradigms Identified Across Tools Researchers examined prior human‑GenAI interfaces and extracted twelve recurring atomic interaction patterns that are composable and reusable [1]. These paradigms include actions like selection, drag‑and‑drop, and multi‑modal annotation, each mapped to specific prompt modifications [1]. The taxonomy allows designers to compare and evaluate interaction choices across platforms [1].

Four Demonstration Scenarios Show Practical Utility The paper presents four distinct scenarios—application refinement, workflow automation, creative brainstorming, and educational tutoring—where the IAI model guides the selection or invention of interaction paradigms [1]. In each case, the model predicts how specific GUI actions will alter AI output, demonstrating descriptive, discriminative, and generative capabilities [1]. These examples illustrate how the framework can accelerate prototype development and user testing [1].

Model Addresses Limitations of Text‑Only Prompts Authors argue that pure text prompts often fail to convey fine‑grained or referential intent, leading to ambiguous AI responses [1]. By integrating precise GUI actions, the IAI model enables users to specify spatial, relational, and iterative constraints that text alone cannot express [1]. This hybrid approach is expected to foster richer, more controllable human‑AI collaboration [1].