Amid rising alarm over deepfake videos of politicians and robocalls to voters, as well as social media algorithms that drive people into political news silos, Political Science Asst. Prof. Emma Rodman still sees reason for optimism about AI.
“The worry is that AI and misinformation erode democratic discourse and democratic norms,” she says. “My interest is in the flip side. I think there are a lot of potential ways to use AI to promote democratic engagement.”
Some academics and even social media companies are working on ways to do that, including Rodman, a political theorist. Working with an Honors College student, she is developing a large language model based on writings by political theorists that other political theorists could query to help them think more creatively about fundamental political questions, such as “When freedom and equality conflict, how should we decide what to prioritize?”
AI can’t solve such political dilemmas, she says: Each person will come to a different answer based on their own beliefs, experiences and feelings. But asking an AI model to articulate a wide range of viewpoints could help political theorists approach such questions in new ways.
“No one view is going to be more valid than another, but they’re all going to give me a lot of raw material to start to make those judgments,” she says.
Generative AI also has great potential to help politicians and policymakers, she says, because it’s good at talking through policy questions, although humans ultimately must decide what course to pursue.
For ordinary citizens, Rodman thinks large language models could be created that would empower people to participate more meaningfully in democratic self-governance. For starters, AI is great at summarizing complex policy proposals, court decisions and legislation in everyday language so that people can understand and then act on them, she says.
AI can also be a good conversational partner, she says. Not only is it well-informed, but Rodman believes that a virtual partner could help people ask questions and work through their views on complicated and sensitive topics, such as immigration policy, without fear of a snide remark or angry outburst.
That’s one possible way that large language models could help increase political participation by people who, right now, don’t feel comfortable forming and expressing their views, she says.
“So much of a conversation is talking to ourselves out loud in the presence of another person, trying things out to figure out what we think. And we’re finding that having those conversations with an AI model, you get a lot of the same benefit, but in a low-stakes way,” she says.
Most of all, Rodman is encouraged by her work with students in exploring the potential uses of AI. “Students are uniquely positioned to be really great at thinking outside the box about how to use these models,” she says. “They think about things in ways I don’t, and that’s so valuable.”—KW