What problem does it solve? Designing AI-powered interfaces is difficult because non-deterministic outputs cause trust miscalibration and error cost mismanagement. This Skill provides a structured framework to classify trust-risk and error-risk, select the right human-AI collaboration pattern, and design feedback loops and progressive disclosure for AI features. ## Core Features & Use Cases - Trust-Risk and Error-Risk Classification: Applies Yang et al.'s (2020) framework with four assessment criteria each, a rule-based classification algorithm, and higher-risk tie-breakers. - Interaction Pattern Selection: Maps classifications onto a 3x3 matrix producing nine human-AI collaboration patterns, from full human oversight to AI autonomy, with a never-lower-oversight safety rule. - Feedback Loop and Progressive Disclosure Design: Covers all 18 Amershi et al. (2019) guidelines across four phases and Shneiderman's five-stage progressive disclosure plan with advancement and rollback criteria. - Use Case: A team building an AI recommendation engine asks how much autonomy to give it; the Skill classifies trust-risk and error-risk, selects a human-in-the-loop pattern, and produces a feedback loop and progressive disclosure plan. ## Quick Start Ask the parent /user-experience skill to design the AI interaction pattern for your feature, for example: classify the trust-risk and error-risk for our AI-powered recommendation engine and design its feedback loop.
