Navigating the Challenges of AI-Generated Text in Civic Conversations

When policy making takes place, it’s often accompanied by an opportunity for the public to provide their insights and feedback. In the past, lawmakers have had the hefty task of sifting through unrealistic, mass-produced commentary intended to influence their decisions. In recent years, however, artificial intelligence (AI) has introduced a new challenge. Now, there’s a surge of realistic-looking feedback with no way to trace back to a human source.

Artificial Intelligence: A Double-Edged Sword

Deb Roy, a media arts and sciences professor at MIT and the director of MIT’s Center for Constructive Communication (CCC), recently shone a spotlight on this issue. In his essay published in the ‘Bulletin of the Atomic Scientists’, titled “Who stands behind the words?,” Roy insists that even in the new realm of AI-generated text, there’s still a need for accountability and someone to answer for the words used.

Roy’s concern extends to how AI can be used, not misused, in civic discourse. He co-founded Cortico, a related nonprofit, with the aim of facilitating trust-building group conversations on important societal issues. This happens everywhere from civic meetings and schools to private organizations. While useful tools like AI are utilized to help identify trends in commentary, the process is designed in a way that participants wield the power to override AI-based output at each phase. This ensures that at each stage, a human touch is present and someone is responsible for each public comment.

Fake Feedback: Masking Authentic Public Sentiment

Roy’s worry about accountability in AI-generated feedback is not unfounded. Highlighting an instance from 2017, he points out how the Federal Communications Commission was flooded with over 22 million comments, a shocking 18 million of which were found to be fictitious. This is troubling on many levels, primarily because these fake feedback pieces engineered by AI are much more sophisticated: each comment is unique and seems authentic, making it near impossible to distinguish genuine public opinion.

But rather than write off AI entirely, Roy advocates for harnessing its power responsibly. He suggests four guiding principles to ensure accountability in AI-authored text: tangible authorship, a thorough and transparent process, clear tracing back to origin, and spreading the responsibility broadly instead of concentrating it. This would culminate in empowering participants to take ownership of their AI-generated contributions instead of eliminating the human element altogether.

Another key point Roy brings up is the idea of moving ‘at the speed of trust’ as opposed to the ‘speed of tech’. He stresses the importance of earning community trust and aligning the speed of AI developments with democratic procedures. This way, there’s no risk of decisions being made without the inclusion of public opinion.

In essence, Roy underscores the necessity of integrating technology with human oversight, allowing us to enjoy the benefits of AI while remaining cautious of its potential to distort public feedback systems. AI is an undeniably powerful tool, one that can greatly enhance civic engagement when used with a degree of responsibility.

If your company is considering AI automation solutions, you can explore more on how AI can improve your business operations at implementi.ai.