Chara Podimata

Assistant professor Chara Podimata is collaborating with MIT professors Adam Berinsky and Charles Stewart III on a public experiment, the Election LLM Observatory: a dashboard where the responses of major models to questions about candidates and elections are published and can be examined by journalists, researchers, and ordinary users. "I feel like we haven't taken the right steps to educate people, to make them understand what these chatbots can and can't do," she said. "Technology is changing the world very drastically and very quickly. We have left people behind. We have given them the technology itself, fully accessible, without educating them on how to use it."

"We started hearing from friends around us that they have been using chatbots in order to obtain information about elections, so we started thinking about, 'How does a chatbot that knows some information about you potentially change the information that it gives you with regard to elections?'" said assistant professor Chara Podimata. She created the Election LLM Observatory with MIT professors Adam Berinsky and Charles Stewart III. "I want the public to know that the type of information that each of us gets may be different from what our friend gets for exactly the same question," Podimata said.

Assistant professor Chara Podimata and MIT professors Adam Berinsky and Charles Stewart III have designed a public dashboard called the LLM Election Observatory for tracking how A.I. models are responding to questions about prominent midterm candidates and political issues. "A.I. is becoming part of the way people encounter and make sense of political information, and yet we know relatively little about what that information environment actually looks like, and how it differs for different user demographics, political identities and geographies," said Podimata.

Assistant professor Chara Podimata and co-author wrote: "Large language models (LLMs) like ChatGPT, Claude, and Gemini, among others, are becoming the new vessels (and sometimes, arbiters) of political information. Our research suggests their influence is already rippling through our democracy. These models may appear neutral — politically unbiased, and merely summarizing facts from different sources found in their training data or on the internet. At the same time, they operate as black boxes, designed and trained in ways users can't see."