Overview

Often when we imagine human-AI interaction today, there is a specific kind of way of involving a human: the system makes a first pass at a task, and the human acts as a fallback. This state of human-AI interaction, or humans-as-backup, creates a boxed notion of interaction possibilities: that the only way humans can give feedback to the system is to label data points as examples of what they want and serve as training examples. Building off of Guidelines for Human-AI Interaction, and the AI Design Guide, Meaningful Human Control serves as interaction patterns of re-imagining human-AI interaction to human-AI collaboration, that honors and harnesses the unique capabilities of both people and technology. Building off of Guidelines for human-AI interaction, and the BAG AI Design guidelines, The Meaningful Human Control framework provides designers, researchers, and engineers insightful interaction patterns to assist in re-imagining human-AI collaboration; patterns that honor and harnesses the unique capabilities of both people and technology. We will walkthrough the process, user research, and patterns in practice through a real-world case study of Dynamics 365 Connected Store (AI-powered experiences in retail).