Human-centered AI means starting with thinking about what people need and then about how AI can help solve the problem. This isn’t just Human + AI, though; here, the order matters. When people come first, you design things that augment people rather than replace them. You create tools that empower people to solve challenging problems. And empowered people can solve all kinds of critical and/or interesting things by teaming up with an intelligent computer.
However, for people to partner with an AI, they must trust the AI. So, Human-centered AI also means trustworthy AI. Trust is many things, but it is always modified by the relationship between the giver and receiver. So, when the giver is a person and the receiver is a human-centered AI, there are several critical components, but the two most important are: to easily know what an AI wants to achieve and to know that an AI cannot cover up a failure.