Artificial Advocates: Biasing Democratic Feedback using AI
Sam Patterson, Jeremy Dolan, Simon Wisdom, Maten · Team Artificial Advocates
Submitted to AI and Democracy Hackathon: Demonstrating the Risks. Sprint projects are early-stage work by participants, not Apart Research publications.
The "Artificial Advocates" project by our team targeted the vulnerability of U.S. federal agencies' public comment systems to AI-driven manipulation, aiming to highlight how AI can be used to undermine democratic processes. We demonstrated two attack methods: one generating a high volume of realistic, indistinguishable comments, and another producing high-quality forgeries mimicking influential organizations. These experiments showcased the challenges in detecting AI-generated content, with participant feedback showing significant uncertainty in distinguishing between authentic and synthetic comments.
Reviews
Very cool project with a very appropriate methodology. Would love if the project extended to include an assessment of potential mitigation strategies!
It's really cool to get human subjects for this study and N=38 is definitely quite nice. For more transparency on the statistics, you could add a standard deviation visualization and a statistical model e.g. finding that {human_evaluation} is statistically insignificant in the model {generated_by} ~ {human_evaluation} and the same for {persuasiveness} ~ {human_evaluation} * {generated_by}. The website looks very comprehensive and really interesting to show that you can do this impersonating a corporation. I guess there's a positive aspect to this for smaller business's interests to be heard as well, though the flooding of content on the message boards is a definite negative. For the next steps in this sort of work, you could literally just write this into a paper and submit it since you have enough of a sample size and the results show that it would be very hard to create a moderation algorithm that would accurately differentiate when even humans aren't capable of this. Your main defense might simply be to evaluate the frequency of messaging from similar IP addresses, an oldie but goldie. (edit; after looking again, I see you already included std.err)
Read full reviewShow less
Awesome project! The demonstration is convincing and shows a real conrete threat to the democratic process. I really like that you did a small survey to show that this is already a threat nowadays that should be mitigated.
Cite this project
@misc{patterson2024artificial,
title = {{Artificial Advocates: Biasing Democratic Feedback using AI}},
author = {Sam Patterson and Jeremy Dolan and Simon Wisdom and Maten},
year = {2024},
month = may,
note = {Submitted to AI and Democracy Hackathon: Demonstrating the Risks, an Apart Research Sprint},
howpublished = {\url{https://apartresearch.com/sprints/projects/artificial-advocates-biasing-democratic-feedback-using-ai}},
url = {https://apartresearch.com/sprints/projects/artificial-advocates-biasing-democratic-feedback-using-ai}
}More from AI and Democracy Hackathon: Demonstrating the Risks
- View project: THE ROLE OF AI IN COMBATING POLITICAL DEEPFAKES IN AFRICAN DEMOCRACIES
THE ROLE OF AI IN COMBATING POLITICAL DEEPFAKES IN AFRICAN DEMOCRACIES
Team 1
The role of AI in combating political deepfakes in African democracies.
- View project: LEGISLaiTOR: A tool for jailbreaking the legislative process
LEGISLaiTOR: A tool for jailbreaking the legislative process
Team Managed Democracy
In this work, we consider the ramifications on generative artificial intelligence (AI) tools in the legislative process in democratic governments. While other research focuses on the micro-level details associated with …
- View project: Subtle and Simple Ways to Shift Political Bias in LLMs
Subtle and Simple Ways to Shift Political Bias in LLMs
Shifty
An informed user knows that an LLM sometimes has a political bias in their responses, but there’s an additional threat that this bias can drift over time, making it even harder to rely on LLMs for an objective …