Image

The Community, Democracy and Dialogue initiative at UMass Amherst recently convened a panel in the final weeks before the midterm election to explore the influence of social media, AI and misinformation.

“The Consent of the Governed After 250 Years: Democracy in the Age of AI and Social Media,” which was held on Oct. 14 in the Campus Center Auditorium, gathered experts in history, politics, public policy and computer science to present about and discuss current technologies, including social media and artificial intelligence, and their impact on the democratic process and democracy in general.

Panelist Ethan Zuckerman, associate professor of public policy, communication and information in the Manning College of Information and Computer Sciences, started his panel presentation by outing himself as a “cyber optimist” who, in the 1990s and early 2000s, was hopeful about the internet’s potential to impact democracy in a positive way by distributing new voices through blogs and social media. 

“The hope that we cyber optimists had was that maybe this would change who got listened to, who got to set the priorities, and what problems we took seriously,” said Zuckerman, who is also director of the UMass Initiative for Digital Public Infrastructure. 

It didn’t work out that way, he admitted, but he added that AI now holds similar potential for a positive impact on the future – if we make it happen ethically.

“I can imagine futures in which the current trajectory of AI products gives us significant social benefit. I’m willing to entertain the cure-for-cancer scenario. I’m willing to entertain wonderful batteries connected to solar panels,” he said. “What I have a very hard time imagining is democratic futures being brought about by AI. And the reason for that is that the most capable AIs are being built by a small number of incredibly wealthy companies. Their priorities for achieving artificial general intelligence, making trillions of dollars – that is governing what those companies do.”

Image

What Americans need to decide, Zuckerman said, is “what AI should and should not do, how we should build them, monitor them, audit them, regulate them. And what lines we might demand as a society that AIs do not cross.”

Joining Zuckerman on the panel were Manisha Sinha, the James L. and Shirley A. Draper Chair in American History at the University of Connecticut, and Jonathan Nagler, a professor of politics from New York University. The moderator was Jane Fountain, distinguished professor and director of the School of Public Policy at UMass Amherst, as well as a Manning CICS adjunct distinguished professor and a faculty affiliate in the Department of Political Science. 

In her panel presentation, Sinha laid out the historical narrative that has led the country to this point, beginning with the American Revolution. “Democracy in the United States has always been challenged by forces of economic and political reaction,” Sinha said. That tension between capitalism and democracy has brought about today’s “environment of corporate lawlessness, where we have a regime that is completely bent towards letting them ‘self-regulate,’ which we know is not really possible,” she said.

“Unfortunately, we have a government in power [today] that is not the cure and, rather, has made things worse,” Sinha said. “The fate of our republic now really lies in the hands of ordinary American citizens … The question now iswhether we will be able to rein in some of these people – not to be complete Luddites and reject modern technology, but to see the ways in which it can serve the common good instead of unrestricted individual profit that at times can cost us our democracy and also can cost us our Earth.”

Nagler, in his panel presentation, discussed the impact of social media and AI on citizens’ understanding of politics and their voting behaviors. “The main point I want to make is: There is a lot of stuff on social media,” Nagler said. “We have very little evidence that it is moving people’s opinions a lot.”

One example Nagler cited was the Russian attempt at interference in the 2016 U.S. elections. “They sort of tried and failed,” he said. Russian Internet Research Agency tweets increased during the election year, but most were seen by staunch Republicans. 

“Pretty much anybody working on a campaign would tell you: If your goal is to increase Republican votes shared, don’t bother sending these things to strong Republicans. They’re already voting Republican. So probably not very influential,” Nagler said.

In terms of AI, Nagler’s research has shown that the majority of average users don’t tend to go to AI for political input. Of those who do ask AI agents about politics, most don’t ask for facts, and only 12% asked AI to compare or recommend candidates, he said.

More exploration is needed, said Fountain, the panel moderator. “Objective technologies are neither positive nor negative,” she said. “Yes, if you make an assault rifle, maybe they can say that’s objectively negative, but it has a certain purpose. But information technologies, transportation technologies – you really need to explore them in their use. Both corporate uses and financial uses, as well as individual uses.”

A recording of the complete event can be found below and on YouTube.