Chapter 2. Human-Centered AI

We often measure the success of AI by comparing it to human performance on similar tasks. AI model creators use human accuracy as a benchmark against which model performance is measured. The public at large often expects AI to be perfect, or at minimum always accurate, although we don’t expect this of people. Human-centered AI focuses on pairing humans and AI where neither competes with the other; rather, they amplify and complement each other, enabling or creating optimal results. In this chapter, we define the principles of human-centered AI and explain why it’s essential for building excellent AI healthcare tools and systems of care. This chapter will help with understanding the importance of human-centered AI and how to accomplish it.

Toward Human-Centered AI

A great place to start when thinking about human-centered AI is the world of gaming. AI can be great at playing games like checkers, chess, Jeopardy!, or the ancient strategy game Go. Healthcare is an entirely different domain because of the potential consequences, and because of the absence of a simple set of rules that govern how patients and diseases should behave and how care should be administered. The goal of human-centered AI is not to compete with humans but to enhance human performance and improve collaborations among caregivers. But we have learned that even in competition, such as in games between a computer and the world’s best chess player, machines and people do better when ...

Get AI-First Healthcare now with the O’Reilly learning platform.

O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.