AI-Enhanced Democracy

Creating a more resilient and responsive democratic republic

Created by Author using MidJourney February 2024

Conservatives, Liberals, and Libertarians seem to agree on one thing; the United States’ democratic republic is broken. Polarization of political opinion is high and confidence in government, political parties, officials, and candidates is low. LLMs are already damaging our governing processes. LLM hallucinations and LLM-generated misinformation confuse both citizens and government officials making rational decisions more difficult to achieve. What can we do? Can AI tools, especially LLMs, help the situation instead of making it worse?

My answer is, “Yes, AI can help democracy work better.” AI in general, and LLMs specifically, can be designed to empower citizens, reduce polarization, and make government more efficient and responsive.

Here are five ways that AI systems using LLMs can be designed to improve government.

1. Facilitated Consensus Building: Implement “AI Consensus Building Agents” (Consensus Agents) to monitor and transcribe local neighborhood meetings, then share the detailed feedback and proposals with other neighborhoods and up to the city council and mayoral level. The Consensus Agents can be designed to collect questions from citizens, obtain feedback at each level, and assemble possible solutions to problems. The Consensus Agents could then iterate “feedback => modified solution => evaluation of solution=>feedback” among all of the participating parties for several cycles until solutions converge into either a consensus or an acceptable level of compromise. We could design these systems to continuously improve, doing a better job of converging on a consensus faster and more transparently while generating only an acceptable level of dissent from the participants.
2. Enhanced Transparency: Create “Government Guide Bots” that are connected to every level of government and are instrumented to know everything about government policy, law, operations, staffing, procedures, budgets, and financial performance. Using natural language conversations in all common human languages, citizens could ask the Government Guide Bot any question, request any analysis, and discuss any hypothetical improvement or change. These Government Guide Bots would make the inner workings of government less opaque and would allow the citizen to use government more effectively and to critique government more intelligently.
3. Intelligent Issue Management: At any point in time citizens and businesses and other government agencies are working on various issues toward a mutually satisfactory resolution, or to an adjudicated decision. For example, a building owner working with a construction company may be inquiring as to what permits are required for a proposed project and seeking to obtain those permits. Government operated AI-driven, natural language speaking “Issue Resolution Agents” would be trained on and able to access data such as rules, laws, and policies, as well as geographic information. They would be able converse in natural languages with the citizens, companies, and government agencies regarding their issues. The Issue Resolution Agents would be able to work on each specific issue 24/7 maintaining natural language communication with all parties. Knowledge of escalation rules would help obtain resolution of various resistant threads.
4. Model Building and Comparison of Hypotheticals: Many times, government and citizen initiatives are complex and require thorough analysis and evaluation so a rational decision can be made. For example, a proposed road intersection might have several different designs which would have varying impacts on traffic and neighborhoods and aesthetics as well as different cost levels and cost and revenue profiles over time. “AI Model Building Agents” could help build models of these options, and then support natural language conversations to answer questions and log concerns. Opinions could be gathered and feedback from the appropriate government agencies obtained so the most beneficial and least concerning design can be determined.
5. Advocacy: Various individuals and groups within any large government jurisdiction are chronically underserved. This may be due to racial discrimination, lack of education, use of a non-mainstream language, personal financial weakness (poverty), or myriad other reasons. A natural language speaking “AI Advocate” could be tasked with identifying then assisting these individuals as they work with available government agencies or philanthropic programs, and with the individuals themselves, to help provide all the opportunities available. These AI Advocates would help ensure that AI-driven solutions are accessible to all citizens, including those with disabilities or limited technological literacy.

I am a techno-optimist, and I am hoping that these intelligent conversational AI agents can help us overcome the dysfunctional deadlock that we currently find ourselves in. These AI agents could foster public engagement and support everyone’s participation in the development and implementation of government AI initiatives. This would help ensure legitimacy and trust in the technology we are building.

While we stand up AI oversight agencies to restrict AI’s misuse, let’s also stand-up AI development agencies with the goal of creating AI to make our constitutional democratic republic work better, faster, more efficiently, more fairly, and more intelligently. This is analogous to the FAA (Federal Aviation Administration) and NASA (National Aeronautics and Space Administration), two federal agencies that both regulate aviation and also support the development of new aviation technologies.

The following principles should guide our use of AI and specifically LLM technology to enhance our governance processes:

• Ensure AI systems are designed and monitored to mitigate biases and uphold ethical standards, particularly in decision-making processes.
• Safeguard citizens’ privacy rights by implementing robust data protection measures and transparency protocols.
• Establish mechanisms for accountability and transparency in AI governance to prevent misuse or abuse of power.
• Explore interdisciplinary collaborations and incorporate diverse perspectives in AI development to enhance the effectiveness and inclusivity of these initiatives.

Continuous monitoring, evaluation, and adaptation of these programs, based on citizen and business feedback, are crucial for the success of AI-enhanced governance in fostering a more resilient and responsive democratic system.