ABSTRACT
Artificial intelligence (AI) is often treated like a black box, which is when scientific and technical work is made invisible by its own success. AI is a field within computer science, with its own subfields of machine learning and deep learning, and that it can draw on Big Data. This chapter explores through a socio-technical approach how diversity—the inclusion of people of various backgrounds—can impact how AI is being built and used by society. The concept of diversity takes into account a wide variety of personal characteristics such as gender, sexual orientation, age, race/ethnicity, physical/mental disabilities and health, socioeconomic background, and religion/spirituality. In general, AI bias can happen in three main stages, i.e., pre, during, and after data processing. Diversity bias can, for example, occur if the data selected for a program or algorithm is not reflecting diverse enough data.
