WASHINGTON (December 3, 2025) –Governments say they want public input on AI policy but new research shows they aren’t actually listening.
A new paper, “Lost in Translation: Policymakers Are Not Really Listening to Citizen Concerns About AI,” by Michael Moreno, a research associate, and Susan Ariel Aaronson, a research professor at GW compares public-comment processes in the U.S., Australia, and Colombia and finds that fewer than 1% of citizens participated, and policymakers rarely showed how or if public feedback shaped final decisions.
Despite rising public concern about AI’s impact on jobs, rights, and democracy, the study reveals weak outreach, overly technical materials, short timelines, and almost no feedback loops. The result: a major credibility gap in democratic AI governance.
Here are some findings they discovered in their research:
- Public participation was extremely low across all three countries: Australia received 510 submissions, the U.S. received 326, and Colombia received only 73 all far less than 0.01% of their national populations.
- Most submissions came from academics, firms, and policy insiders, showing that governments failed to reach a diverse or representative cross-section of the public.
- Citizens often lacked the information needed to provide meaningful input, as seen in Colombia releasing its draft roadmap just two days before the consultation and the U.S. posing 52 highly technical sub-questions.
- Policymakers provided little evidence that public feedback shaped their decisions, with the U.S. and Colombia showing no clear changes and Australia offering only limited adjustments.
- None of the three governments moved beyond basic consultation, meaning none created a genuine two-way dialogue or empowered citizens in AI governance.
The authors warn that without meaningful public involvement, governments risk losing legitimacy on one of the most consequential technologies of our time and they propose eight reforms to modernize democratic engagement on AI.
This paper is based on work supported in part by the NSF-NIST Institute for Trustworthy AI, which is supported by the National ScienceFoundation.
To speak with the researchers about these findings please contact Skyler Sales at skylers
gwu [dot] edu.
-GW-