97 Things About Ethics Everyone in Data Science Should Know

Chapter 19. Build Multiperspective AI

Hassan Masum and Sébastien Paquet

Sr. Director, Analytics, Prodigy EducationApplied Research Scientist, Element AI

How AI and data science are deployed is decided by their makers and owners—but other stakeholders are impacted and are sometimes harmed.

If you care about avoiding harm, you have to consider multiple perspectives when deploying data science and AI. You must step into someone else’s shoes by asking: if I were that person, how would this system impact my life?

The impacts of a new technology can be unintended and complex, even for a seemingly benign goal such as connecting people. The creator or any single actor cannot see the whole picture. But every actor deeply understands their own situation and can therefore assess how a given technology serves or harms them.

Learning from stakeholders’ opinions and lived experiences has long been part of responsible technology discourse (in ideas like “implementation research” and “diffusion of innovations”). That accumulated experience remains relevant in the data science age.

With this in mind, how can you incorporate more points of view in data science and AI development? Here are people whose perspectives you should consider incorporating, and questions they are likely to care about: