AI for Public Discourse
The research project “AI for Public Discourse” began in 2022 as a student-led initiative to build an artificial intelligence (AI) chatbot for responding to hate speech. Notre Dame students Nik Swift, Grace Connors and Miriam Bethencourt built a prototype and wrote a paper outlining the project. In year one of the research, the team explored the complex task of hate speech classification and detection. Team members also researched the difficulty of generating effective and contextually sensitive responses using large language models (LLM) from OpenAI and Cohere, a type of AI that can process large amounts of text. A variety of tasks can be performed by LLMs, such as summarizing documents and analyzing the temperature of social media, among others.
The project received seed funding from the Civic Health Project, and the PTAP Lab is now in formal partnership with the Dangerous Speech Project, a non-governmental organization focused on how to respond to toxic online content, and the Notre Dame's Center for Research Computing, which is helping research how best to train AI to respond to toxic online speech.
Using carefully curated datasets and considering potential front-end integrations through a Twitter API bot or a Google Chrome extension, the team experimented with prompting and fine-tuning. Two webinar workshops were hosted, first with researchers to get feedback on research designs and questions related to the project, then with counterspeakers – people who respond to toxic content online – to learn what kinds of tools could benefit them.
In 2025, the PTAP Lab is surveying practitioners and creating a dataset of foundational science articles to further fine-tune LLMs to respond to toxic speech online.
Recent Events:
LLMs & Public Discourse: In February 2025, the Plurality Institute and The Council on Technology and Social Cohesion co-hosted a showcase and workshop bringing together pioneering work using LLMs to improve public discourse. Watch the presentations at the Plurality Institute YouTube channel.
Related Publications:
Research Poster by Kristina Radivojevik and Lisa Schirch for February 2025 workshops on LLMs in Public Discourse