The Almost Intelligent Revolution: Options for Scaling Up Deliberation and Empowering People with AI

Democratic deliberation relies on the "ideal speech situation," where participants engage in rational communication on equal footing. However, institutional settings often employ specialized registers—complex jargon and formal structures—that act as gatekeeping mechanisms. These linguistic barriers systematically exclude individuals who lack the specific cultural or educational capital required to navigate these norms, effectively reinforcing existing power imbalances.

Large Language Models (LLMs) offer a potential solution by acting as intermediaries between citizens and democratic institutions. From a Systemic-Functional Linguistics perspective, LLMs can be used to:

• Simplify institutional communication: By translating dense, bureaucratic texts into accessible language, LLMs make policy documents understandable to a broader public.
• Scaffold argumentation: LLMs can help participants transform personal narratives and lived experiences into structured, formal arguments that are more likely to be recognized and valued within institutional decision-making processes.
• Mediate consensus: Experimental tools like the "Habermas Machine" have demonstrated that AI can synthesize diverse individual opinions into coherent group statements, often scoring higher in clarity and fairness than human-mediated summaries.

While LLMs provide powerful tools for inclusion, they are not neutral. Because they are trained on vast datasets of human text, they inherently reflect the biases and power structures embedded in that data. The author warns that AI-mediated deliberation risks shifting consensus toward majority views, potentially marginalizing dissenting voices. Furthermore, because LLMs lack true communicative intent and are prone to hallucinations, they must be used with caution. Over-reliance on AI could lead to a "sycophantic" environment where models prioritize pleasing users or reinforcing existing norms rather than fostering critical, independent thought.