Open-mindedness is critical for productive dialogue in free societies. Solving complex social problems requires people to engage with diverse perspectives, reflect on assumptions, and build shared understanding. As so much of public dialogue shifts to online platforms, the rise of generative AI introduces both new opportunities and risks.
Emerging research indicates deploying generative AI agents as active participants in online dialogue may help individuals recognize their intellectual blind spots or speak more effectively across social divides. Yet other researchers warn these new technologies will also reinforce biases and encourage dogmatism.
We propose a series of randomized controlled trials on an updated version of the research platform we created via previous support from JTF that will allow us to study how the deployment of AI agents on digital platforms shapes open-mindedness and productive dialogue.
This new project—which will also be supported by Duke University and the federal government— will produce a) 6 research publications; b) an improved platform for scientific research on human-AI interaction that will benefit a broad range of scholars; c) practical guidelines and protocols for stakeholders in academia, government, and the private sector to study how generative AI can be developed and deployed for social good; d) thought pieces that disseminate these findings to a broad public audience; and e) training and research opportunities for three graduate students and one post-doctoral fellow. As both AI technologies and academic research are rapidly evolving, our project is poised to inform scholarly and public understanding of how AI agents can damage or reinforce the interrelation of character development and free societies.
