# MIT Researcher Urges Accountability and Standing in AI-Generated Civic Feedback Deb Roy proposes design frameworks to restore human answerability in public policy dialogues. - **Source:** [MIT News — AI](https://news.mit.edu/topic/artificial-intelligence2) - **Published:** 2026-10-13 - **Topic:** [Data & AI](https://brifing.si/c/data-ai) - **Original article:** - **On Brifing:** https://brifing.si/s/87ab7a53-d6ab-486e-876c-f125e976b5e5 ## Erosion of Public Consultation Channels Democratic governance relies heavily on public comment periods when crafting administrative policies. Historically, officials filtered mass-produced spam by spotting duplicate phrasing. The widespread emergence of large language models has broken this defense, enabling bad actors to generate endless distinct, highly fluent submissions without leaving recognizable textual fingerprints. ## The Scale of Automated Fraud In 2017, the FCC received over 22 million comments on net neutrality, of which nearly 18 million were later revealed to be fraudulent. Investigators caught that earlier campaign because it reused templated language. Modern generative AI completely erases such structural markers, leaving reviewers unable to separate genuine stakeholder opinions from automated text. ## Absence of Physical Grounding MIT Center for Constructive Communication director Deb Roy argues that language models lack genuine experiential grounding. Regardless of their apparent linguistic fluency, software systems cannot possess personal stakes or physical bodies. Consequently, automated algorithms have nothing to lose and cannot anchor accountability when their generated text impacts real civic decisions. ## Defining Stakeholder Standing Drawing an analogy to New England town halls, Roy emphasizes the informal necessity of standing. In traditional physical settings, natural friction prevented outsiders without a legitimate stake from dominating discussions. Digital networks powered by generative AI dismantle these geographic boundaries, requiring intentional system architecture to ensure affected communities are genuinely heard. ## Enforcing Civic Answerability Standing must be paired with answerability, creating an obligation to justify spoken claims under scrutiny. While AI tools can draft, summarize, or identify patterns in civic input, a human must ultimately assume responsibility for the text. The right to influence policy should always require someone willing to defend the underlying message. ## Preserving Privacy While Verifying Humans Answerability does not mandate stripping citizen privacy in public forums. Intermediary institutions can verify that an authentic, eligible human stands behind a comment without exposing their identity publicly. Roy notes disclosure demands should scale with institutional power, while protective measures should prioritize safeguarding vulnerable individual participants during civic debates. ## Four Principles for System Design In the Bulletin of the Atomic Scientists, Roy outlined four core structural design principles. These mandate attributable authorship, decomposed inspectable processes, provenance as infrastructure, and expanded citizen standing. Autonomous tools remain permissible only when human accountability is preserved, measuring success by empowered human roles rather than simple automated efficiency. ## Mismatched Clocks of Tech and Democracy Technological innovation and democratic deliberation operate on conflicting timelines. While AI development accelerates relentlessly, civic governance requires slow consensus-building to maintain public trust. Operating at the speed of trust requires deliberate community engagement; otherwise, rapid automated systems risk outrunning public oversight and generating major policy choices entirely without constituents. ## Real Talk Listening Initiatives MIT CCC deploys initiatives like Real Talk to foster genuine dialogue within small, structured groups. Supported by campus leadership, these programs use explicit consent architectures alongside analytics tools to reveal underlying community themes. These small-scale discussions allow authentic human sharing before technological summarization attempts to aggregate broad civic perspectives. ## Tracing Claims With VoiceTrace To embed technical answerability, researcher Cassandra Overney developed VoiceTrace at MIT. The prototype directly connects high-level summaries and policy recommendations back to the original human voices recorded during deliberative assemblies. This provenance infrastructure ensures every automated civic insight can be audited directly against verified source participants. ## Workplace Fragmentation and Tunneling Through the New Social Forms project with Google Experience Institute, researchers identified a coordination crisis in modern workplaces. Employees use generative models individually to build deeply refined ideas in silos, resulting in divergent team alignment upon surfacing. Rebuilding shared trust and collective communication within corporate teams mirrors civic deliberation challenges. ## Rebuilding Deliberative Integrity The infiltration of unverified AI text threatens to push public officials away from valuing civic commentary altogether. Preserving democratic legitimacy requires treating technology as a facilitation aid rather than an autonomous participant. Sustainable governance relies on ensuring verifiable human stakeholders remain the ultimate authors and defenders of public discourse. ## Sources covering this story - https://news.mit.edu/2026/qa-building-accountability-behind-our-words-and-ai-civic-conversations-1013 — --- _Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._