Direct Digital Democracy: The dawn of 3D government, by the people, for the people

Something New Under the Sun, Part V, a series of essays on digital democracy

“Our country needs an alternative to the Democrat-Republican uniparty so that the people actually have a VOICE.” — — Elon Musk, June 30, 2025

Elon, are you listening? We stand at a peculiar historical moment where the same technology (AI) threatening to concentrate power even more in the hands of a few could instead democratize decision-making more radically than ever before. The race for artificial superintelligence presents both humanity’s greatest peril and its most promising opportunity for genuine self-governance.

The irony is almost too perfect: AI, the technology most feared for its potential to subjugate human agency, may be the key to finally achieving the direct democracy that philosophers have dreamed of for millennia but dismissed as impractical. The question is not whether AI will reshape governance — it will. The question is whether we’ll harness it for human flourishing or allow it to become another tool for elite control.

Digital Direct Democracy (what I’m calling “3D Government”) can be implemented in many ways, at the federal state and local levels, or simply as a way of polling people in real-time about their preferences on specific policy issues. For example, if the White House actually cared about what the public thought about bombing Iran, it could survey people in real time and base decisions on democratic preferences. Wouldn’t that be nice?

The Information Overload Excuse

Critics of direct democracy have long relied on a seemingly reasonable objection: modern governance is simply too complex for ordinary citizens to understand. How can someone with a full-time job, family responsibilities, and limited time master the intricacies of tax policy, environmental regulation, foreign affairs, and hundreds of other specialized domains that comprise modern governance?

This critique has been the intellectual foundation for representative democracy’s claims to superiority. We elect (alleged) professionals, the argument goes, who can dedicate their time to understanding these complex issues and making informed decisions on our behalf. It’s an ostensibly compelling argument that has justified the concentration of political power for centuries.

But there’s a problem with this logic: it doesn’t match the evidence about how political decisions actually get made, or how well experts perform compared to crowds.

As I’ve documented in earlier essays, the supposed experts in political judgment that we elect to represent us don’t actually have time to become experts in the issues they vote on. Congress members spend so much time fundraising that they rarely read the bills they vote on, let alone develop deep expertise in the underlying policy domains. Meanwhile, as James Surowiecki demonstrated in The Wisdom of Crowds, collective intelligence consistently outperforms expert judgment across a wide range of domains.¹

The Good Judgment Project, a research initiative funded by the Intelligence Advanced Research Projects Activity (IARPA), has provided compelling evidence for crowd wisdom. Over four years of competition involving 500 questions and more than a million forecasts, volunteers working in small groups consistently outperformed professional intelligence analysts — even those with access to classified information — by 30% to 72%.² As recently as 2024–2025, these “superforecasters” have proven 30% more accurate than futures markets in predicting Federal Reserve interest rate decisions.³

Crowdsourcing platforms like Innocentive routinely solve problems that stymie corporate research teams, while the crowd on Who Wants to Be a Millionaire picks the right answer 91% of the time compared to experts’ 65%.⁴

But here’s what’s changed since I wrote those essays: artificial intelligence has matured to the point where it can finally address the legitimate concerns about information processing that have long plagued direct democracy proposals. And AI can shoulder much of the burden of the time it takes for regular people to learn about the issues.

Your AI Voter Proxy (AVP): Democracy’s Digital Twin

AI is something truly new under the sun. Imagine this: every citizen could have an AI assistant that knows their values, preferences, and reasoning patterns. It knows this because the voter chooses (yes, this system would need to be entirely voluntary and tied closely to human oversight) to train its AI proxy in this way, going through a questionnaire to get the AI up to speed.

This “AI proxy” then studies every piece of legislation, reads every policy analysis, attends every committee hearing, and follows every expert debate (all in a matter of seconds because AI can operate at millions of times faster than human thought). The proxy understands your fundamental principles — your views on individual liberty versus collective welfare, your risk tolerance, your priorities across different policy domains.

When a vote comes up, your AI proxy can instantly analyze the proposal against your value system and help you cast a vote that represents what you would decide if you had unlimited time to study the issue. Or the AI voter proxy (AVP) could, if you choose to go further, cast votes for you on ballot measures of all kinds, and on candidates.

You retain veto power, of course — you can override any decision your proxy suggests — but for the vast majority of votes, you can trust that your digital twin is representing your interests more faithfully than any human politician ever could. And if things go wonky you can always step in and make the required adjustments.

This system would allow far more robust citizen participation in policymaking because the power of the crowd, particularly when assisted by the magnifying power of AI proxy votes, could be expanded over time to do much of the work that our current system gives to our entirely corrupt politicians. Throw the bums out and replace them with the crowd.

This is what the new era of AI and crowdsourcing wisdom makes possible. It’s potentially huge in its impact.

This isn’t science fiction. The technology already exists. Large language models can already be fine-tuned on individual writing samples and decision patterns to approximate personal reasoning styles.⁵ Recent research shows that AI systems can be trained to maintain consistent personalities and adapt to individual user preferences through machine learning.⁶ Personalized AI assistants are already learning user habits, adapting to preferences, and automating tasks tailored to specific needs.⁷ What we’re missing is the will to deploy this technology for democratic rather than commercial purposes.

Think about the implications. The traditional objections to direct democracy evaporate:

Information overload? Your AI proxy processes information at superhuman speed while maintaining your human values and judgment criteria.

Lack of expertise? Your proxy can synthesize input from thousands of domain experts while filtering it through your personal value system.

Time constraints? Your proxy works 24/7, attending every hearing and reading every document while you live your life.

Emotional manipulation? AI systems can be designed to recognize and resist the kind of emotional appeals that distort human judgment.

Crowd Wisdom at Scale

The beauty of this system is that it preserves and amplifies the three conditions that make crowds wise: diversity, independence, and decentralization.⁸

Diversity is maintained because each AI proxy reflects the unique values and reasoning patterns of its human. Unlike human representatives who tend to cluster around similar backgrounds and perspectives, AI proxies would be as diverse as the population they represent.

Independence is actually enhanced because AI systems can be designed to resist the social pressures and groupthink that often corrupt human decision-making. Your proxy votes based on your values and the evidence, not because of what other proxies are doing.

Decentralization reaches its logical conclusion — -every citizen becomes their own representative, making decisions through their digital twin.

But this system goes beyond preserving crowd wisdom; it scales it to unprecedented levels. Instead of relying on the 535 members of Congress to somehow represent 330 million Americans, we could have 330 million informed participants in every decision. The collective intelligence of such a system would dwarf anything in human history. And what can work for Congress can of course also work for state legislatures and city councils. In fact, this kind of system will almost surely start small and then expand, so city and county councils will be the more likely starting point.

Racing Toward Democratic Singularity

This brings us back to the Diamond Society tension I’ve written about, with twin forces pushing power and wealth both up and down our social strata, and the race for artificial superintelligence. The same AI capabilities that could enable authoritarian control could instead create the most democratic society ever imagined. The question is which vision we pursue first.

The authoritarian path is clear: use AI for surveillance, manipulation, and control. Monitor citizens, shape their preferences, and concentrate decision-making power among a technological elite. It’s the path many governments are already pursuing.

The democratic path requires more imagination and political will, but it’s technically feasible: use AI to amplify human agency rather than replace it. Give every citizen the tools to participate meaningfully in governance while preserving their autonomy and judgment.

The urgency cannot be overstated. As I argued in the Diamond Society essay, whoever achieves artificial superintelligence first could potentially lock in their vision of governance permanently. If authoritarian forces win that race, the opportunity for meaningful democracy may be lost forever.

But if democratic forces can mobilize AI for collective intelligence first, we might create governance systems that are both more effective and more resistant to authoritarian capture than anything in human history.

Practical Steps Toward AI-Enhanced Democracy, or “3D Government”

The transition wouldn’t happen overnight, but we could start building the infrastructure now:

Phase 1: Personal AI Policy Assistants Citizens could begin using AI tools to analyze ballot measures and candidate positions against their stated values. These systems would start simple — -helping people understand what they’re voting on and how different choices align with their priorities.

Phase 2: Delegated Voting Systems Citizens could choose to delegate their votes on specific issues to their AI proxies while retaining override capabilities. This could start with local ballot measures and expand over time.

Phase 3: Liquid Democracy Integration AI proxies could facilitate liquid democracy systems where citizens can delegate their votes to trusted experts on specific domains while maintaining direct participation on issues they care about most.⁹ Countries like Estonia have already pioneered blockchain-based digital governance systems that could serve as foundations for AI-enhanced democratic participation.¹⁰

Phase 4: Full 3D Government Eventually, AI-mediated direct democracy could handle most governance decisions, with professional politicians relegated to implementation and crisis management roles.

The technical barriers are surmountable. The real challenge is political — -convincing people that the cure for democracy’s ailments is indeed more democracy, enhanced by artificial intelligence rather than replaced by it.

Answering the Skeptics

Critics will raise obvious objections: What about security? What if AI systems are hacked or manipulated? What about the digital divide? These are serious concerns that require serious answers.

Security: AI voting proxies could use the same cryptographic protections as modern financial systems, with additional safeguards like multi-factor authentication and blockchain verification of vote integrity.

Manipulation: Open-source AI models would allow public auditing of voting systems. Citizens could verify that their proxies are reasoning correctly and haven’t been compromised.

Digital divide: Public funding could ensure universal access to AI voting proxies, just as we provide public education and voting infrastructure today.

AI bias: Diverse development teams and rigorous testing could minimize bias, while the sheer scale of participation would help average out individual AI quirks.

The deeper objection is philosophical: Is this still democracy if artificial intelligence is making the decisions? I would argue it’s more democratic than what we have now. Your AI proxy implements your values and reasoning, just with superhuman information processing capabilities. It’s no different from using a calculator to do arithmetic — the tool amplifies your capabilities without replacing your judgment.

The Choice Before Us

We’re living through what may be the last moment in human history when we can choose our technological future rather than having it chosen for us. The race for artificial superintelligence will determine whether AI becomes humanity’s servant or its master.

But there’s a third option that most discussions miss: AI as humanity’s amplifier. Not replacing human judgment, but scaling it to global proportions. Not concentrating power, but distributing it more widely than ever before. Not ending democracy, but finally fulfilling its promise.

The technology exists. The principles are sound. The evidence supports crowd wisdom over expert judgment. What we need now is the political imagination to deploy these tools for human flourishing rather than human subjugation.

The ancient Athenians invented democracy with the technology of their time: physical assembly, clay voting tokens, random selection by lot. We can reinvent democracy with the technology of our time — -artificial intelligence, global networks, and digital twins that represent our values with perfect fidelity.

The question is not whether AI will reshape governance. The question is whether we’ll reshape it consciously, in service of human agency and collective wisdom, or let it reshape us according to the narrow interests of whoever wins the race for superintelligence.

Elon, are you listening? It would be poetically ironic if you used your massive wealth and power to create this kind of direct democracy system where people can truly re-take power back from our utterly corrupt government.

[Claude 4.0 helped in writing this essay]

References and Further Reading

Primary Sources:

¹ Surowiecki, J. (2004). The Wisdom of Crowds: Why the Many Are Smarter Than the Few and How Collective Wisdom Shapes Business, Economies, Societies and Nations. Doubleday.

² Tetlock, P. E., & Gardner, D. (2015). Superforecasting: The Art and Science of Prediction. Crown Publishers. See also: Mellers, B., et al. (2014). “Psychological strategies for winning a geopolitical forecasting tournament.” Psychological Science, 25(5), 1106–1115.

³ Good Judgment Inc. (2025). “Superforecasters outperform futures markets on Federal Reserve predictions.” Retrieved from https://goodjudgment.com/

⁴ Financial Times (2025). “Superforecasters continue to have the edge over the futures market in anticipating FOMC decisions.” Monetary Policy Radar. See also: Surowiecki, J. (2004), noting that crowds on Who Wants to Be a Millionaire achieved 91% accuracy vs. 65% for experts.

Technical Research:

⁵ Karunakaran, S., & Jain, A. (2024). “The Ultimate Guide to Fine-Tuning LLMs from Basics to Breakthroughs.” arXiv preprint arXiv:2408.13296v1.

⁶ Karunakaran, S., & Jain, A. (2025). “Fine-Tuning LLMs for Personality Preservation in AI Assistants.” International Journal of Research in Modern Engineering & Emerging Technology.

⁷ “The Ultimate Guide to AI Personalized Assistants in 2025.” dipoleDIAMOND. Retrieved from https://www.dipolediamond.com/the-ultimate-guide-to-ai-personalized-assistants-in-2025/

Democratic Innovation:

⁸ “Wisdom of Crowds: Definition, Theory, and Examples.” Investopedia. Retrieved from https://www.investopedia.com/terms/w/wisdom-crowds.asp

⁹ Ramos, J. (2015). “Liquid Democracy and the Futures of Governance.” Kosmos Journal. Retrieved from https://www.kosmosjournal.org/kj_article/liquid-democracy-and-the-future-of-governance/

¹⁰ “Liquid Democracy: The Future of Governance Powered by Blockchain.” Crypto Altruism (2023). Retrieved from https://www.cryptoaltruism.org/blog/liquid-democracy-the-future-of-governance-powered-by-blockchain

Additional Reading:

• Bostrom, N. (2014). Superintelligence: Paths, Dangers, Strategies. Oxford University Press.
• Benkler, Y. (2006). The Wealth of Networks: How Social Production Transforms Markets and Freedom. Yale University Press.
• Johnson, S. (2012). Future Perfect: The Case for Progress in a Networked Age. Riverhead Books.
• Tetlock, P. E. (2005). Expert Political Judgment: How Good Is It? How Can We Know? Princeton University Press.
• Tetlock, P. E., & Schoemaker, P. J. H. (2016). “Superforecasting: How to Upgrade Your Company’s Judgment.” Harvard Business Review, May 2016.
• Democracy Earth Foundation. “Sovereign: Blockchain Direct Democracy.” Retrieved from https://democracy.earth/
• Good Judgment Open. “Public Forecasting Platform.” Retrieved from https://www.gjopen.com/
• Estonia e-Residency. “Digital Nation for Global Citizens.” Retrieved from https://e-resident.gov.ee/