Machines and artificial intelligence agents are increasingly being added to the web of connections of people engaged in collective actions of diverse sorts. More and more, humans are interacting socially with software agents or physical machines. Here, there has been a lot of attention paid to the harm that can be done by artificial intelligence (AI), ranging from reinforcing socially biased interactions, to driving cars into people and injuring them. But work in the Human Nature Lab led by Nicholas Christakis at Yale University has focused on ways that AI acting socially (or “social AI”) can actually help groups of people to help themselves. Christakis’s research addresses the questions: If we drop a few simply programmed bots into a social system, can we foster better behavior in the group? Where in the system should the bots be placed? How should they be programmed to behave?
We have shown that it is possible to use very simple programming — what we call “dumb AI” — to have dramatic effects on collective behavior because, in the social situations we examine, the bots are mixed in with (much smarter) humans in a “hybrid system” comprised of humans and machines interacting on the same plane. We do not have to emulate human intelligence in order to have an impact, but rather we simply must support this intelligence. In other words, our objective is not to develop super-smart AI to replace human cognition; rather, our objective is to develop dumb AI to supplement human interaction.
In this project, we will develop and execute game-theory-driven experimental scenarios using a sophisticated online experiment system to improve our understanding of how groups of interconnected individuals can better solve collective problems of diverse, important sorts. First, we will test humans in a broad range of problem scenarios. Then, we will explore how AI agents (bots), placed within online social groups, can affect the behavior of individuals and groups as they confront these problems. We hope our research will provide valuable insights to these looming social and technological issues as well, informing us about how human beings embedded in these human-machine networks can be best equipped to make decisions that benefit themselves and their communities.
The Using Social AI to Modify Collective Behavior in Realistic Networks project is being led by Nicholas A. Christakis at Yale University (New Haven, US).
Nicholas A. Christakis
Sterling Professor of Social and Natural Science, Internal Medicine & Biomedical Engineering