Teaching an AI about empathy — and learning about ourselves

Greg Lyons

At UNICEF Innovation, we’re supporting our partners to help build models that give us a better understanding of empathy. We hear in the news about how artificial intelligence (AI) can be a tool for evil — how robots can take over the world, or how automation will take away jobs. Often that comes from an understanding of a mechanical type of AI. But if AI can understand how human empathy works,we have the potential to build intelligences that help us understand ourselves.

UNICEF deals with a world that is increasingly more fractured and more troubled than ever before, and one of our core missions is to help people understand this complex world and how they can help to fix it. We are asking our friends and supporters to spend some of their time to train a piece of AI that is being built out of MIT’s Scalable Cooperation Lab, so that it can better understand elements of human emotion.

The end result we’re looking for? Models that are publicly available and published, and a deeper understanding that shows us how to learn from AI about ourselves. Understandings such as how can an image help us empathize and connect with the reality of others, or how can AI help us distinguish fake news from real news when talking about a disaster?

Related Content