Q&A: Building accountability behind our words — and AI’s — in civic conversations

In an MIT News interview published October 13, 2026, Deb Roy, an MIT media arts and sciences professor who directs the Center for Constructive Communication, discusses his Bulletin of the Atomic Scientists essay, "Who stands behind the words?" He argues that every piece of public text, however it was produced, needs a person who can answer for it. AI may understand language or not, he says, but it lacks standing because software has nothing to lose.

His lead example is the 2017 Federal Communications Commission docket on repealing net neutrality, which drew more than 22 million comments; the New York attorney general later found nearly 18 million were fake. That fraud left detectable patterns, Roy notes, whereas language models can make each comment fluent and distinct. He believes public comment, a pillar of democracy, has effectively collapsed as reviewers quietly discount it.

Roy says AI can legitimately draft or summarize if a stakeholder adopts the result and will defend it, and that anonymity can survive when institutions verify speakers. His essay offers four design principles: attributable authorship, inspectable processes, provenance as infrastructure and broader standing.

The center and its affiliated nonprofit, Cortico, apply these ideas in community dialogues, including a New York City high school assembly on cellphone policy. PhD student Cassandra Overney's prototype VoiceTrace links summaries and recommendations to the voices behind them, and a project with Google examines workplace teams whose AI-assisted thinking diverges. Roy warns that AI deployment is outpacing democracy's ability to respond, leaving decisions made without the public.