Learning Unit 5: Human Agency and Oversight
In this learning unit, we will examine human oversight requirements for AI systems. You will learn about different oversight approaches (including real-time and post-hoc interventions) and challenges like automation bias. Methods for communicating system uncertainty are covered and generating explanations to support human decision-making. Through case studies, students will design and evaluate interfaces that maintain meaningful human control over AI systems.
Exercises
Task 1 - Human Agency & Oversight
In this learning unit, we are going to investigate a use case of an AI system where the lack of human agency and oversight lead to severe problems.
Please read the introduction, discussion, and conclusion of the linked article on the Epic Sepsis Model (ESM), an sepsis early warning system that was in use at hundreds of hospitals throughout the US: External Validation of a Widely Implemented Proprietary Sepsis Prediction Model in Hospitalized Patients.
