The Trinity College of Arts & Sciences recently launched the Society-Centered AI Initiative aimed at researching the relationship between the evolution of AI and human behavior.
The initiative, directed by Christopher Bail — professor of sociology, computer science, political science and public policy — addresses both the challenges and opportunities AI presents by focusing on its technological advancement and ethical, social and behavioral impacts.
“We are not simply out to criticize AI, but to think productively about how it can improve,” Bail said. “So striking a balance between the sort of wild optimism and humorism that public debate tends to careen between in public discussions of AI and its impact on society, [and] to really do research on this theme, because there's so much speculation [about the future of AI].”
He shared that the ultimate goal is to create a “large-scale center” at Duke that hires its own faculty and offers courses on the subject. The initiative includes representation from faculty across all of Duke’s schools.
Although the initiative remains in its early stages, Bail shared that he and the team hope to “create better benchmarks for tracking AI’s success and influence,” including whether the technologies can “navigate complex social settings.”
“If we deploy a chat bot on social media or in a healthcare administrative system, can it understand things like human social psychology, social networks, peer influence [and] all of these types of fundamental human processes?” he presented as a potential research question for the team.
Bail also noted that while there are concerns about bias in machine learning, “there's also countervailing evidence that AI can assist humans in overcoming inter-group conflict.” He said that he is optimistic that AI can provide beneficial services, but that the challenge now is disproving “[untested] speculation.”
He further emphasized the group’s goal of addressing AI’s impact on marginalized groups, which he characterized as “often not well represented in the training data that power[s] many of the foundation models.”
“We recognize that though there’s a potential for harm, there's also enormous opportunity, and we need research to inform the middle space between those two,” Bail said.
In the long term, Bail shared that the initiative will involve collaboration with entities outside Duke, including non-profit organizations, industry partners and political actors working on AI regulation.
He also noted that the initiative will offer various opportunities for student involvement, including an event series featuring experts on the topic. Bail also submitted a request to the Arts & Science Council to have a learning Constellation centered around AI and society.
“So many universities are trying to figure out how they fit into the broader ecosystem,” Bail said. “And I think Duke has many irons in the fire in the AI space.”
Get The Chronicle straight to your inbox
Signup for our weekly newsletter. Cancel at any time.