The Crisis of Accountability: MIT Professor Deb Roy Warns AI-Generated Public Comments Threaten the Foundations of Democratic Discourse

The rapid evolution of generative artificial intelligence has fundamentally altered the landscape of public policymaking, introducing profound challenges to transparency, accountability, and the democratic process. For decades, government agencies and legislative bodies have relied on public comment periods to gauge civic sentiment before enacting new rules or laws. Traditionally, policymakers navigated the routine obstacle of mass-produced, templated feedback campaigns designed to artificially skew public opinion. However, the introduction of large language models (LLMs) has shattered the existing verification paradigms, rendering traditional methods of detecting inauthentic engagement nearly obsolete.

With the advent of highly sophisticated generative AI tools, bad actors can now produce millions of hyper-personalized, contextually nuanced, and unique comments that mimic genuine human thought. Unlike historical spam campaigns that left obvious fingerprints—such as repetitive phrasing and standardized templates—AI-generated submissions seamlessly blend into the administrative record. This technological leap has severely compromised the integrity of civic feedback systems, raising urgent questions regarding who—or what—truly stands behind the words shaping public policy.

The Erosion of the Public Comment Pillar

The fragility of modern civic feedback mechanisms was starkly illustrated long before the widespread public release of ChatGPT. In 2017, the Federal Communications Commission (FCC) solicited public input on a controversial proposal to repeal net neutrality rules, resulting in an unprecedented influx of more than 22 million comments. A subsequent investigation by the New York Attorney General revealed that nearly 18 million of those submissions were fraudulent, driven by automated scripts and stolen identities.

At the time, investigators successfully unmasked the fraud due to the blunt mechanical nature of the scripts. Today, however, large language models have erased those diagnostic markers. Every AI-generated comment can be uniquely articulated, demonstrating advanced syntax, varied vocabulary, and tailored emotional resonance.

This transformation has created a silent crisis within regulatory agencies. Government officials reviewing proposed rules have quietly downgraded the weight and value they assign to public comments. Because the boundary between authentic human sentiment and synthetic noise has collapsed, the foundational promise of democracy—that citizens must have a voice and feel heard before rules are imposed upon them—faces an unprecedented existential threat. When the public perceives that the administrative state is no longer listening because the communication channels are saturated with unaccountable synthetic text, public trust in institutions deteriorates rapidly.

Standing, Answerability, and the Limits of Machine Intelligence

In an essay titled Who stands behind the words?, published in the Bulletin of the Atomic Scientists, Deb Roy, professor of media arts and sciences at MIT and director of the MIT Center for Constructive Communication (CCC), addresses the philosophical and practical dangers of this accountability vacuum. Roy asserts that while generative AI can assist throughout the communication lifecycle—drafting texts, summarizing large volumes of data, and identifying thematic patterns—it fundamentally lacks the capability to anchor the chain of responsibility.

Drawing upon legal and philosophical frameworks, Roy emphasizes two essential pillars of human discourse: standing and answerability. In legal proceedings, standing ensures that parties bringing a case possess a legitimate stake in the outcome. In civic life, this principle is reflected in traditional forums such as New England town meetings, where participants are recognized as community members with vested interests in local governance.

Answerability functions as the logical counterpart to standing. When an individual offers public testimony or lodges a formal accusation, they simultaneously accept the obligation to be cross-examined, defend their assertions with facts, and take personal responsibility for their statements. Large language models possess neither attribute. Operating as software systems with nothing to lose, they lack legal or moral standing, rendering true accountability impossible.

To address this vulnerability, Roy proposes four foundational design principles for future communication infrastructures:

• Attributable authorship, ensuring human identification is anchored to every public statement.
• A decomposed and inspectable process, allowing observers to audit how conclusions or summaries are reached.
• Provenance as infrastructure, treating the origin of information as a core verification requirement.
• Widening standing rather than concentrating it, ensuring that access to civic dialogue remains broadly distributed among real stakeholders rather than monopolized by automated systems.

The core governing rule of these principles is straightforward: autonomy and technological assistance are permitted, but unaccountability is strictly rejected.

Bridging the Temporal Gap Between Technology and Democracy

The rapid pace of technological innovation has introduced a severe friction point between the timelines of Silicon Valley and the deliberative processes of democratic institutions. While technology developers operate under the ethos of rapid deployment and iterative disruption—frequently advocating that democracy must accelerate to keep pace with innovation—civic practitioners maintain that societal structures require stability, deliberation, and trust.

This temporal mismatch threatens to produce breakaway scenarios where critical policy decisions are executed via automated systems before public institutions have time to process, debate, or respond. To counteract this divergence, Roy and his colleagues at the CCC emphasize the necessity of moving "at the speed of trust." Derived from community-level civic work, this philosophy dictates that sustainable technological integration cannot be imposed unilaterally under the guise of institutional efficiency. Instead, digital tools must be deliberately engineered to facilitate genuine human engagement, honoring the social fabric that underpins cooperative governance.

Practical Innovations in Civic and Workplace Dialogue

Founded to combat polarization and restore public trust, the MIT Center for Constructive Communication—alongside its close affiliate, the nonprofit Cortico—has pioneered real-world interventions across schools, civic assemblies, and private organizations. These initiatives demonstrate how technology can be harnessed to enhance human connection without stripping away human accountability.

Among these efforts, the Real Talk initiatives have successfully cultivated structured dialogue across diverse communities nationwide, including campus-wide programs like realtalk@MIT. By establishing small, intimate conversational contexts that encourage authentic sharing, combined with consent-driven technology architectures, the center maps broader thematic patterns while ensuring human participants retain complete control over the output.

Furthermore, the center’s deliberative assemblies have bridged divides in educational settings. For instance, collaborative projects in New York City high schools have brought students, teachers, and administrators together to address complex institutional policies, such as cellphone regulations. To ensure institutional transparency, CCC researchers have developed prototypes like VoiceTrace—a tool created by PhD student Cassandra Overney that links every summarized output statement or policy recommendation directly back to the specific human conversations and voices from which it originated.

Industrial Democracy and the Modern Workplace

As organizations increasingly integrate artificial intelligence into daily workflows, a parallel coordination crisis has emerged within the professional sphere. According to insights gathered through collaborations with entities like the Google Experience Institute, individual employees frequently utilize AI to deeply refine ideas in isolation—a phenomenon described as "tunneling." When these individuals resurface to collaborate with their teams, their divergence in perspectives creates friction, highlighting a pressing human coordination challenge.

Roy notes that workplaces function as vital civic spaces where individuals develop daily cooperative skills and democratic habits. Referencing Supreme Court Justice Louis Brandeis’s historical concept of "industrial democracy," Roy emphasizes that the psychological and procedural tools necessary for conflict resolution, active listening, and procedural justice are identical whether applied to municipal governance or corporate environments.

Ultimately, the preservation of constructive dialogue relies on recognizing the shared humanity of participants. As generative AI continues to reshape the boundaries of public and private communication, the imperative to anchor digital expressions to accountable, living humans remains the single most critical safeguard for the future of democratic society.