プルラリティと協働テクノロジー：民主主義の新たなビジョン

In 2019, Uber’s expansion in Taiwan had given millions of riders a cheaper option and thousands of drivers an income. Taxi drivers were protesting in the street and the legislature had no agreed position on how to regulate the service. Other countries had either banned Uber or given in to its lobbying. Taiwan’s government put the question to the public, through a digital platform called vTaiwan, which ran on an AI-assisted consensus tool called Polis.

Over four thousand citizens took part. They submitted statements, voted on each other’s proposals, and watched the system map clusters of agreement among positions that had looked irreconcilable. The format had no debate stage and no reply button, and a proposal only gained visibility once it drew support from more than one cluster, which left little room for trolling. Within weeks a set of regulatory principles had majority support: Uber would operate legally under a new licensing framework, existing taxi drivers would get transition support, and consumer protections would be written into law. The legislation passed with little opposition.

The process was designed by Audrey Tang, Taiwan’s Digital Minister, a former child prodigy and open-source hacker. Tang calls the approach Plurality.

What Plurality Is

Plurality is a framework for governing diverse societies, worked out most fully by Tang and the economist E. Glen Weyl. Their book, Plurality: The Future of Collaborative Technology and Democracy, was written in the open, with contributors, and published under a Creative Commons license.

The starting claim is that diversity is a resource. The book’s argument is that collaboration works best when people who genuinely disagree have a structured way to work together, and that most existing technology is built on the opposite assumption.

Much of the book is spent on what the authors are arguing against. China’s social credit system and centralized AI surveillance get the longest treatment: the state uses data to enforce compliance, and dissent is handled as a malfunction. Meta, Google, and the behavioral-advertising industry get the next: data on individual differences is used to sell attention, and the platforms make more money when users are angry, because angry users keep scrolling. Weyl and Tang are more ambivalent about the wing of Web3 that wants to write governance into smart contracts and replace institutions with protocols. They credit it with useful work on decentralized coordination and blame it for speculative bubbles, governance failures, and for ignoring power imbalances that markets do not fix by themselves.

The Toolkit

The book describes a set of mechanisms. Some have been used in government.

Quadratic voting fixes a flaw in one-person-one-vote: it cannot express how much someone cares. A voter who is passionate about climate policy and indifferent to zoning casts the same vote on both. Under quadratic voting, participants get a budget of “voice credits” to spread across issues, and the price of piling votes onto one issue rises quadratically. One vote costs one credit, two cost four, three cost nine. Colorado’s state legislature used quadratic voting in 2019 to prioritize bills, and the process surfaced bipartisan agreement on issues that had been deadlocked for years. The RadicalxChange Foundation, co-founded by Weyl, has run quadratic voting experiments in Taiwan, in several U.S. municipalities, and in corporate governance.

Polis is the tool used in the Uber case. Participants post short statements and vote agree or disagree on other people’s statements. Machine learning groups the voters into clusters and surfaces the statements that more than one cluster agrees with. Statements that only one cluster likes sink. Taiwan has used Polis through vTaiwan and the Join platform on ride-sharing regulation, telemedicine, online alcohol sales, and dozens of other contested issues, processing input from millions of citizens.

Plural money and community currencies question the idea that a single national currency is the only legitimate medium of exchange. Local currencies, from the long Japanese tradition of chiiki tsuka (地域通貨) to newer blockchain-based systems, let communities build exchange mechanisms for their own needs. Sarubobo Coin in the Hida-Takayama region of Japan circulates as a digital community currency that keeps spending local. The book treats local currencies as a way to keep some economic activity outside global capital markets.

Data coalitions and digital cooperatives sit between the surveillance model, where platforms take your data, and the libertarian model, where you sell it on an open market. In a data coalition, individuals pool their data under collective governance, negotiate its use as a group, and split whatever the data earns. The model comes from credit unions and agricultural cooperatives. Driver’s Seat Cooperative, which lets rideshare drivers pool trip data and sell anonymized insights to city planners, is an early example. Weyl’s term for the underlying principle is “data dignity”: data generated by people should benefit those people.

Decentralized identity and verifiable credentials let people prove things about themselves, such as qualifications, affiliations, or civic participation, without handing personal data to a central platform. Soulbound tokens, community attestations, and decentralized reputation systems are all attempts at this, and several come from people close to the Plurality project.

Taiwan

Tang served as Digital Minister from 2016 to 2024.

vTaiwan processed policy input on more than thirty regulatory issues. The Presidential Hackathon, an annual competition in which citizens propose digital fixes for public problems, produced prototypes that government agencies adopted. The g0v (“gov-zero”) civic tech community, the largest in Asia, built open-source tools for budget transparency, disaster response, and parliamentary monitoring that were later folded into official government infrastructure.

COVID-19 put these systems under strain. Taiwan’s response, widely regarded as among the most effective anywhere, used tools built by civic technologists, and the government put privacy questions to public deliberation rather than settling them internally. The mask rationing system, built by government and civic hackers together in a matter of days, distributed supplies to 23 million people using real-time data and open APIs.

Tang spoke at a previous Tech for Impact Summit. The question raised there, whether any of this works beyond a small, highly connected island, has not been answered.

Japan: Society 5.0 and Digital Democracy

Keidanren set out the Society 5.0 vision in 2016, and it later became national policy. Like Plurality, it puts people at the center of technology. It describes a society in which cyber and physical space are integrated to solve social problems, with human well-being rather than economic output as the stated goal. The Digital Agency, established in 2021, has been rebuilding government technology with citizens in mind.

Japan also has specific obstacles. Digital participation in governance is low. Public deliberation through technology is unfamiliar in a country where agreement has traditionally been built through nemawashi, the patient, informal process of one-on-one conversations that precedes any formal decision. Municipalities are trying digital engagement tools, Kobe’s participatory budgeting platform and Tsukuba’s blockchain-based voting pilot among them, but adoption is scattered.

An aging population and a shrinking workforce mean Japan needs ways to hear from citizens who cannot attend a town hall. The country’s cooperative tradition, from agricultural cooperatives to neighborhood associations, is ground on which data coalitions and community governance could grow. And because Japan is a Western-style liberal democracy inside Asia, anything that works there has a better chance of working in both regions.

Web3 and Its Limits

DAOs, decentralized organizations run by token-holder votes, are the biggest experiments in digital democracy outside national governments. Thousands of them manage billions of dollars in treasury assets, deciding funding, protocol upgrades, and community rules through on-chain votes. Some use quadratic voting. Others have tried conviction voting, rage-quitting, and other governance primitives with no equivalent in traditional institutions.

Where voting power tracks token holdings, governance reproduces wealth inequality. Turnout in most DAOs is very low, often under 10% of eligible holders. Governance attacks, in which someone buys tokens specifically to swing a vote, have cost treasuries real money. And the technical difficulty of on-chain participation shuts out most people who might otherwise join.

Quadratic voting limits the weight of concentrated wealth. Soulbound tokens and decentralized identity can tie governance rights to a person rather than to capital. Polis-style deliberation can run before an on-chain vote so the vote reflects what the community actually thinks rather than the preferences of a small technical elite. Nobody has yet combined Web3 infrastructure with Plurality’s governance design at any scale.

Why This Is Hard

Complexity. Quadratic voting, data coalitions, and deliberation platforms ask a lot of participants. People have to learn unfamiliar mechanisms and understand how their input is processed, at a time when the most successful apps are the ones that ask for nothing.

The digital divide. A governance system that runs mainly online risks excluding people without reliable internet, digital skills, or time. In Japan, where over 29% of the population is over 65 and digital adoption among older citizens lags, this would exclude a lot of people.

Manipulation. Polis and quadratic voting resist certain kinds of gaming, but no system is immune. Coordinated campaigns, fake identities, and AI-generated content can be aimed at a deliberation platform the same way they have been aimed at social media.

Scale. Taiwan has 23 million people. The same principles would have to be tested in India (1.4 billion), in federal systems like the United States, and in supranational bodies like the European Union. Nobody knows whether computational consensus works at that scale. It would probably need AI able to synthesize input from hundreds of millions of people without averaging away the differences the method depends on.

Tang and Weyl’s answer is that these problems need research budgets and field tests, and that so far they have had very little of either.

Join the Conversation

On April 26, 2026, the Tech for Impact Summit brings senior executives, policymakers, and technologists to Tokyo Garden Terrace Kioi Conference to work on the questions that will shape the road to 2050. The theme is “Beyond Boundaries: Building 2050 Together”.

Audrey Tang spoke at a previous Tech for Impact Summit to a room that spans technology, capital, and social impact. This year’s confirmed speakers include Taro Kono (former Minister of Digital Affairs), Charles Hoskinson (Cardano founder), Yoshito Hori (GLOBIS), Kathy Matsui (MPower Partners), Ken Suzuki (SmartNews), Jesper Koll (Monex Group), Sota Watanabe (Astar/Startale), and Ken Shibusawa (Commons Asset Management).

If you run a technology company building collaborative tools, a financial institution rethinking governance, or an organization with a stake in how democracy gets defended, the summit is where this conversation continues.

Explore partnership and membership opportunities →

Watch highlights from previous summits: youtu.be/ujy7ZXflrt4

---

The Tech for Impact Summit is an invitation-only executive gathering taking place April 26, 2026, in Tokyo as a partner event of SusHi Tech Tokyo. Learn more at tech4impactsummit.com.