Lawmakers Fail To Differentiate Between Human and AI Letters

What's the story?

• A new study by Cornell University revealed that state legislators in the U.S. are unable to distinguish between AI-generated letters and those written by actual constituents.
• The groundbreaking research is raising concerns about the security of the nation's democracy, as it relies directly upon the public having a fair say in what their elected representatives are taking action on. 

Breaking down the study

• The study sent every U.S. state legislator letters composed by both humans and AI chatboxes on six controversial issues — reproductive rights, gun control, policing and crime, tax levels, public health, and education — in both right-wing and left-wing tones.
• Overall, legislators responded to 17.3% of human letters and 15.4% of machine-generated letters, with a mere 2% difference between the two. Politicians were more likely to respond to human-written letters regarding gun violence and health policy and more likely to reply to AI letters about education.
• Deception studies like this are common in the social sciences but raise ethical concerns. After the research was complete, the authors contacted various legislators to reveal what happened and to get their take. 

Implications of the study

• The authors felt the need for the study was pressing given technology's role in major past events, such as the 2016 election, when Russian agents used bots to weaponize social media and manipulate American voters. Susan Lerner, executive director of Common Cause New York, said: 

"As AI becomes more sophisticated, its ability to distort democracy, I think, becomes more obvious and alarming. It's deeply concerning, but I think it's much more active and concerning in the area of social media with bots."

• Sarah Kreps, co-leader of the study and director of the Tech Policy Institute at Cornell, noted that the researchers used ChatGPT-3, the predecessor to GPT-4. The new software would likely create more realistic letters, increasing the likelihood of deceiving lawmakers. She said: 

"For every kind of good use of technology, there are malicious uses. And [legislators] need to be on the lookout and more mindful of how technology now might be misused to disrupt the democratic process."

Growing worry over the future of AI

• Experts are highlighting their concern that once generative language tools like ChatGPT are combined with visual or audio deep fakes, which are improving significantly, democracy will be in even more danger.
• Political watchers asked what would happen if ChatGPT wrote convincing scripts for deep fake voicemails or even TikToks. John Kaehny, executive director of the watchdog group Reinvent Albany, said: 

"I think that could just blow away any of our concepts because you have this whole generation of young voters who are almost post-literate. The power of imagery in politics is just enormous."

• While the study is only a preliminary step, and more research is needed to examine the full effects of AI on democracy, the analysis shows that technology is evolving to influence politics. The study's authors urge lawmakers to be more mindful of how AI can be misused to disrupt the democratic process. 

To hear more, tune into a conversation with Causes' CEO Bart Myers, where he discusses the potential threat AI poses to grassroots advocacy and democracy.

What do you think? Are you concerned about AI's impact on democracy?

-Jamie Epstein 

(Photo credit: Flickr/Arris Web)