How to help democracy keep pace with AI

AI for democratic processes so that we can use democratic processes to effectively govern AI.

If our democratic processes are going to keep pace with AI, we need to use advanced technology and research to unlock new capabilities. AI, LMs and ML are well positioned to contribute to deliberative processes by their very nature, and so we have a long list of their potential uses.

This page includes sections for: Products to build and research to do.

Products to build

Some products specifically require harnessing AI whether it's for sensemaking, research or facilitation.

AI Assurance Infra (Bias, Accuracy, etc.)

Urgent AI/ML
Impact Scale
Infrastructure for providing guarantees around bias and accuracy, control over training data usage, comprehensive logging/evaluation, and offline operation capabilities for any integrated AI system.

AI In-Person Facilitator

AI/ML
Impact Scale
AI agent that provides real-time group facilitation for in-person deliberative processes, managing participation equity, topic focus, conflict resolution, and shared understanding across multiple simultaneous breakout groups.

AI Manipulation Detection System

Urgent AI/ML
Impact Scale
Real-time AI-enhanced monitoring and detection system that identifies manipulation attempts, whether AI-generated or human-orchestrated, across all stages of deliberative processes: from recruitment and selection gaming to astroturfing in discussions, coordinated voting patterns, and output tampering. 

AI Online Facilitator

Urgent AI/ML
Impact Scale
AI agent that provides real-time group facilitation for deliberative processes, managing participation equity, topic focus, conflict resolution, and shared understanding across multiple simultaneous breakout groups.

Automated Recruitment Tool

Urgent AI/ML
Impact Scale
AI-powered recruitment system that automates demographic targeting, campaign creation, multi-channel outreach, and participant selection using real-time census data and advanced selection methodologies.

Babelfish

AI/ML
Impact Scale
Real-time translation infrastructure for multi-language deliberative processes that captures all audio and written inputs (microphone, documents, handwritten notes) and provides instantaneous translation to each participant’s selected language through on-screen display or in-ear devices.

Context Mapper

Urgent AI/ML
Impact Scale
Tool automatically generates comprehensive context analysis, process planning inputs, and accessible participant briefing materials with visual aids by processing uploaded information (PDFs, transcripts, emails, recordings) and conducting interviews with stakeholders.

Decision Impact Forecasting and Modeling

Urgent AI/ML
Impact Scale
Interface where participants input policy proposals and receive impact models, showing effects on both targeted and related outcomes with confidence intervals, assumptions, and interactive exploration capabilities.

Opinion Mapper

Urgent AI/ML
Impact Scale
Tool that connects to recording devices and data sources to instantly analyze participant contributions, creating real-time visualizations of views, consensus areas, and divisions on any topic throughout deliberative processes.

Process Design Simulation Sandbox

Urgent AI/ML
Impact Scale
Simulation environment that tests deliberative process designs by running multiple simulated processes, and predicting outcome distributions to show how design choices impact results.

Smart Templates (AI-supported real-time format support)

AI/ML
Impact Scale
System for helping ensure that the outputs of a task are in the right form, e.g., by evaluating content’s fit to a specified template and instruction, and giving feedback as comments, suggested edits, or through chat.

Research to do

Before we can build advanced products, we need fundamental research that demonstrates what can be done by answering any of the AI and ML research questions we have listed here.

How can we design adaptive learning systems that provide personalized learning programs?

Urgent

How can AI support the creation of compelling media experiences that support parascaling?

How can AI support depolarization, and what new problems might it create in low-trust environments?

Can AI generate its own suggested changes and test them to search the latent space for optimal solutions?

Urgent

What design variables in deliberative formats can AI systems reliably identify as leverage points for optimization through automated multi-agent simulation?

To what extent can AI be used to provide reliable live-time fact-checking within deliberations?

Under what conditions can AI-simulated participants maintain democratic legitimacy?

Can AI generate its own suggested changes and test them to search the latent space for optimal solutions?

How can automatic logging of key events improve access for verifiers?

How can we assist or automate onboarding and process troubleshooting to reduce the costs of inclusion?

How much authentic human value is lost at each level of AI involvement (AI note-taker vs. AI facilitator vs. AI co-deliberator) and where is the steepest drop-off in the value-cost curve?

Urgent

If 'doing the work' of synthesizing and clustering is more valuable than having an AI do it, do participants benefit equally from 'doing this work' or does it privilege those with more skills and stamina?

Urgent

Can AI systems identify their own biases and reasoning errors more reliably than individual humans can identify their own cognitive biases when making sense of inputs?

Urgent

To what extent can we clearly communicate the inner workings of AI-augmented deliberative tools?

How to enable AI-provided context that is appropriately comprehensive and sufficiently unbiased?

How can cryptographic mechanisms create locking mechanisms and binding incentive structures?

What are some inputs used in technical alignment approaches that could be produced or enhanced with deliberative processes?

How to develop an AI facilitator that is attentive to power imbalances, adaptive to group dynamics and effective in guiding groups towards successful outcomes?

Urgent

How can digital tools assist human facilitators to more effectively facilitate deliberations?

Urgent

How can we design synthesis and filtering systems that distill massive public input into actionable insights?

What are the effects of AI facilitation on public perceptions, group dynamics and deliberative quality?

Urgent

How can delibtech tools expand the space of policy scenarios and considerations in a transparent and fair way?

Urgent

How can technically binding decisions integrate with AI alignment in gradual ways?

What transparency and consent mechanisms are required for hybrid assemblies?

How reliably can language models trained on deliberative transcripts, stated rationales, and value-elicitation outputs distinguish between implementation decisions that are consistent with versus divergent from the normative commitments embedded in process outputs?

How can individual learning agents identify and pair learning partners for defined objectives (idea crosspollination, depolarization, information gaps)?

Urgent

What machine translation and annotation approaches (comparing human-in-the-loop vs. automated vs. hybrid) maintain semantic accuracy for multilingual data in international or diverse assemblies, particularly for idioms and context-dependent meaning?

How can we solve the technical blockers to effective and truth-worthy multi-agent simulation and modelling?

Urgent

How can AI systems translate, generate and integrate learning materials into diverse formats (text, audio, visual, etc)?

Urgent

Does the integration of deliberative technologies raise fundamentally new transparency challenges to processes and if so, what are they?

Which open standards and API specifications (building on ActivityPub, NDJSON, or deliberation-specific formats) best enable interoperability between different tools while operating within organizations' existing tech stacks and governance constraints?

How do we prevent gaming or manipulation of AI backup systems?

How to develop real-time dashboards that track process health across multiple dimensions?

How can deliberative processes be used to produce formal and verifiable specifications (unit tests, integration tests) for technical systems?

How can we assist or automate the aggregation of deliberative input from diverse participants in real time whilst maintaining nuance around minority perspectives?

How can we develop real-time detection systems for coordinated manipulation attempts during participant recruitment and selection?

Urgent

How can we design responsive information systems that provide accurate context in real-time?

What role can sentiment analysis and emotion recognition play in real-time conflict monitoring?

How can lessons from speculative execution and speculative decoding help increase the availability of deliberative processes through reduced costs?

How can lessons from speculative execution and speculative decoding help increase the availability of deliberative processes through reduced costs?

Urgent

What is the Pareto frontier of speed, accuracy and easy-to-use interactability?

Urgent

What are the key technical blockers (agent behavior calibration, emergent group dynamics modeling, preference faithfulness) to effective and trustworthy multi-agent simulation, and which are tractable with current methods?

What design choices help promote the transparency of deliberative technologies, and what tradeoffs does this raise?

How can we use third party verification of AI systems used in deliberation, using deliberation?

Urgent

Which transcription and annotation approaches (comparing human verbatim, human semantic, hybrid human-AI, or AI-only) best handle cross-talk, non-verbal communication, and emotional valence while maintaining accuracy standards?

How can we translate mathematical bias guarantees from algorithmic settings to real-world human facilitation?

Urgent

What unified data models and schema (using RDF, JSON-LD, or domain-specific approaches) enable structured and unstructured inputs to be harmonized across different deliberative tool ecosystems, without losing fidelity to participants' original contributions?

How can virtual and augmented reality technologies help participants experience the perspectives of animals or future generations?