AI systems are now widely deployed in society. Individuals, corporations, governments, and other organizations are using AI for applications as varied as voice dictation, text synthesis, text-to-video generation, movie recommendations, personal finance, chatbots, credit scoring, screening employment applications, social media propagation and monitoring, face recognition, semi-autonomous cars, and warehouse automation. Many of these systems can be broadly beneficial. However, there are often adverse impacts on people in racialized populations and underserved communities, and on election results and vaccination campaigns.

There are significant ethical and social impacts of AI systems, leading to demands for human-centered AI that is explainable, transparent, and trustworthy. The inputs to an AI agent include the goals and preferences of the agent, but it is not clear whose preferences they are or should be.

Each chapter concludes with a social impact section discussing issues directly relevant to that chapter’s topics. The social impact sections are of two types, sometimes containing both:

• •

  broader impacts of AI, which includes intended or unintended downstream consequences of upstream decisions on the design of the AI system or on the choice of data

• •

  use cases about user-facing applications of AI that have had an impact on society or science, either positive or negative.

Chapter 18 on the social impact of AI considers the effects of AI on the digital economy, work and automation, transportation and sustainability. It highlights the roles of human-centered AI, values, bias, ethics, certification, and regulation.