The Consensus Algorithm

Brian Vogt on How Civic Technology Can Upgrade Democracy

In July 2014, on the streets surrounding Taiwan’s Ministry of Transportation, thousands of enraged taxi drivers formed a massive blockade, protesting the sudden, unregulated arrival of Uber. Drivers marched through the maze of idling yellow cabs holding placards demanding a crackdown on “white-plate cars,” the term for the illegal, unlicensed personal vehicles being dispatched by Uber. In a dramatic display of fury, some drivers set fire to their own commercial operating licenses right outside the ministry’s doors. A paralyzing, zero-sum standoff seemed inevitable, pitting traditional livelihoods against modern tech convenience.

Over the next year, tensions continued to rise. Jaclyn Tsai, Taiwan’s Minister without Portfolio tasked with handling the government’s response, could have retreated behind closed doors or deployed riot police. Instead she took an unprecedented risk and handed the crisis over to civic hackers. Utilizing vTaiwan, a citizen-led initiative powered by an algorithmic consensus-mapping tool called Polis,¹ the government crowdsourced the debate to thousands of Taiwanese. As the algorithm processed the chaotic web of grievances, bypassing the loudest trolls, a surprising conclusion emerged: a silent, overwhelming majority actually agreed on the need for passenger safety, fair taxation, and liability insurance. What began as a volatile street blockade was defused by technology that illuminated shared views. This unorthodox intervention transformed explosive public rage into widely accepted recommendations that informed subsequent regulation of ride-share operators.

A core challenge for democracies is that citizens express their preferences on election day, but often are limited in their ability to engage in policy development between elections. The Polis tool employed by vTaiwan has been used to address a range of contentious public policy issues in Taiwan, including how to regulate online alcohol sales and stop revenge porn. Polis was created in the United States but, somewhat paradoxically, so far hasn’t been used in the United States to the same effect as in Taiwan.

The problem is that, unlike Taiwan, the United States lacks a strong institutional infrastructure to implement consensus recommendations. In other words, the challenge is often not a lack of consensus, it is a governance system that grants multiple stakeholders veto power, combined with political incentives that discourage ceding agenda-setting authority. Proponents of civic technology often point to its effectiveness in building consensus around “common sense” reforms, but sometimes ignore the structural challenges in getting those reforms actually implemented. But with AI increasingly enabling rapid and broad citizen consultation, now is the time for Americans to reconsider how to translate consultation into reform.

How vTaiwan Works

vTaiwan is a decentralized, crowdsourced policy-making process that combines digital tools with face-to-face meetings to help the Taiwanese government resolve complex issues, particularly those related to the digital economy. The process typically moves through four stages: Proposal, Opinion, Reflection, and Legislation. Polis is the algorithmic consensus-mapping tool that is the core engine of vTaiwan, allowing citizens to submit opinions and vote on ideas proposed by other users. Polis drives users towards areas of agreement by grouping like-minded users together and then identifying points of agreement between them, while rewarding users for refining positions that attract support from other groups. Once a clear, supermajority consensus emerges that bridges the divides between groups, community and government stakeholders hold live-streamed, structured meetings to refine areas of agreement into actionable policy recommendations. In Taiwan the government frequently uses those recommendations as the basis for new legislation. Governments and other entities in countries such as New Zealand and the United Kingdom have used Polis and other similar algorithmic consensus mapping tools.

Polis in Bowling Green

There have been some efforts to employ Polis in the United States. In 2025, local officials in Bowling Green, Kentucky deployed Polis to modernize public engagement in their comprehensive 25-year urban planning initiative, “BG2050.” Faced with rapid population growth and concerns about losing its small-town character, the city partnered with the Computational Democracy Project and Google Jigsaw to launch a digital campaign, “What Could BG Be?” Unlike traditional, sparsely attended town hall meetings that often amplify the loudest or most extreme voices, this digital approach crowdsourced input from over 7,700 residents (roughly 10 percent of the city’s population). Citizens submitted nearly 4,000 unique ideas regarding infrastructure, housing, and development, while the Polis algorithm mapped their voting patterns to identify areas of broad, cross-demographic agreement.

What made the Bowling Green experiment particularly innovative was its integration of a new AI analysis tool called Sensemaker, which rapidly digested the massive, month-long online conversation and significantly reduced the time required for analysis. The platform revealed a surprising consensus among the politically and geographically diverse population, including overwhelming agreement on priorities like riverfront redevelopment and managing traffic congestion.

What the results did not do is translate into new laws or regulations. In contrast to vTaiwan, Polis was used for advisory visioning rather than developing legislation. To be fair, the high-consensus data did allow the city to rapidly form specialized community planning groups that developed a vision document to guide the county’s development for the next 25 years, and today, some private developers are altering their construction strategies to match the Polis consensus data. However, although city planning is an important task, the Bowling Green example feels less immediately consequential than some of the policy reforms enacted in Taiwan.

Similarly, other jurisdictions in the US have used civic technology but stopped short of producing regulatory or legislative reforms. The New York City Department of Homeless Services used Polis to map opinions, understand pain points, and identify areas of consensus regarding shelter policies and the Colorado state government partnered with the civic-tech organization RadicalxChange to use Polis to collect citizen input on climate issues. But the key challenge to using consensus-driven digital tools to inform policy decisions in the United States remains turning that consensus into tangible results.

Consensus is the Easy Part

In the United States there are plenty of policy positions that have broad bipartisan support. For example, multiple polls (a tool that now seems quaint compared to Polis) from sources such as Fox News, the McCourtney Institute for Democracy, and NPR show that 80 to 90 percent of Americans support requiring background checks for all gun sales. According to a Gallup survey, 79 percent of Americans favor a requirement that voters provide some form of identification. And according to the Pew Research Center, roughly 72 percent of Americans believe there should be strict limits on the amount of money individuals and organizations can spend on political campaigns. Yet all of these popular reforms remain stuck thanks to the veto power of key political stakeholders and the bias in the electoral system towards political extremes due to gerrymandering, primary elections, and media polarization.

America’s founders wisely created a system of checks and balances that forced lawmakers to slow down, debate, and find broad consensus before acting. However, we must distinguish between mechanisms that force deliberation and mechanisms that enforce paralysis. For example, at the federal level, the veto power of individual Senators is significant. At the state level many governors have line-item veto authority and at the local level unelected zoning boards often hold substantial influence over local development decisions.

By contrast, in Taiwan many of the consensus positions identified by vTaiwan have become law thanks to facilitating structures put in place. The vTaiwan organizers refused to host a debate on the platform unless a specific government ministry agreed to sponsor the issue first. vTaiwan also had the direct backing of the Premier, who forced often stubborn ministries (like Transportation, Finance, and Justice) to sit down at the vTaiwan table and respond to the findings. Finally, the digital consensus identified on Polis was only phase one of the vTaiwan process. Once Polis identified the “hidden consensus,” vTaiwan transitioned into face-to-face, live-streamed meetings. These gatherings, typically including government officials, subject matter experts, industry representatives, civil society, and regular citizens, eventually led to formal reforms.

How to Make it Work

Reformers in the United States should learn from Taiwan’s experience—not just its use of technology, but its method for its method for connectinging consensus to reform, minimizing opportunities for politicians to sidestep issues on which citizens are broadly aligned. To do this, some structural changes are needed, which could include:

1. Reform the Administrative Procedure Act. For any new proposed federal regulation (not legislation) there is a required “Notice and Comment” period. Most states have similar requirements. These processes can be hijacked by organized campaigns that drown out consensus positions. Perhaps it is time to integrate a consensus-building AI tool into that antiquated process. Individual agencies could proactively do this or Congress could mandate it through legislation.

2. Establish a commitment for legislative review. At the state or local level, legislatures could commit to consider all recommendations that originate from a consensus-driven civic technology tool.

3. Modernize the ballot-initiative procedure. Each state has a different procedure for advancing citizen-led ballot initiatives, but they generally involve collecting thousands of signatures. In some cases, deep-pocketed special interests have played a leading role in ensuring that their preferred ballot initiatives are put to a vote. States could design alternative mechanisms to translate consensus recommendations derived from tools like Polis into proposed ballot initiatives.

Technology alone cannot overcome a gridlocked political system. But when paired with institutional reforms that smooth out the pathway between broad consensus and policy reform, tools like Polis, and initiatives like vTaiwan, may prove to be invaluable.

Brian Vogt is currently conducting a Democracy Listening Tour of Conservative America. He has more than two decades of experience strengthening democracy in countries such as Afghanistan, Burma, Indonesia, Pakistan, and Thailand. He also played a key leadership role in the 2021, 2023, and 2024 Summits for Democracy that brought together more than 100 countries to recommit to democratic values and fundamental human rights.

1. A team of Seattle-based programmers designed the open-source Polis software in 2012 and continues to update it under the banner of the Computational Democracy Project, a nonprofit organization.