Summary

Taiwan's Cyber Special Ambassador Audrey Tang develops AI systems to modernize democracy through broader citizen participation. Tang, who became Digital Minister in 2016, uses technology to transform societal polarization into constructive political solutions—rather than amplifying extremes. She presents concepts such as decentralized digital identities and AI-enabled consensus platforms that make the "surprising common ground" between diverging positions visible. Nearly half of Taiwan's population already actively uses her participatory platform.

People

  • Audrey Tang (Cyber Special Ambassador Taiwan; former Digital Minister)

Topics

  • Digital democracy
  • AI and participation
  • Polarization and consensus
  • Media literacy
  • Cyber resilience

Clarus Lead

The question of how democracies function in the age of AI and social networks is becoming a key point of contention between Western states. While technological platforms worldwide amplify polarization and spread disinformation, Taiwan's approach demonstrates an opposing path: AI as a tool for consensus-building rather than opinion amplification. Tang's systems demonstrate that polarization itself is not a disease but energy – the question lies in directing this energy toward shared solutions rather than extremes. For established democracies like Switzerland, which are still developing digital participation, this approach provides concrete operational models.

Detailed Summary

Tang came to digital policy in 2014 through protest: she developed a digital platform for the Sunflower Movement to document the population's concerns about a Chinese trade agreement. The system distilled thousands of responses into a coherent initiative, which the government later considered. From this success, further systems emerged – the G0v movement created in 2012 civil society parallels to government websites where citizens developed alternative versions.

The central technical concept is the "uncommon ground" principle: rather than sorting users by conviction, the AI platform Polis analyzes thousands of statements and identifies surprising overlaps. A concrete example was the Uber conflict in Taiwan. Taxi drivers and platform partners were in direct opposition – yet Polis showed in real time that 97 % consensus existed on questions of fair insurance and safety. This common ground enabled a new, flexible taxi category that both groups accepted.

Tang distinguishes between "broadcasting" (making complaints public) and "broad listening" (systematic listening). The energy for participation already exists – people are already complaining in social networks. The difference: on Tang's platforms, engagement leads to concrete changes. The Fediverse model (e.g., Threads) is stronger in Taiwan than TikTok (4 %) because it doesn't bind users to a single provider – like a mobile carrier where the number can be transferred.

For media literacy, Tang has shifted the paradigm: instead of only teaching critical consumption, young people are engaged as media producers. They measure air quality together, verify facts, and respond to disinformation with memes. This participation makes them less susceptible to polarization narratives because they evaluate sources and take a critical stance toward narratives.

The "Freedom. We the People" project in the USA surveyed 2,400 people from congressional districts on fundamental values. Surprisingly: 70 % approval even on the strongest polarization issues (e.g., judging people by their abilities). This data shows that polarization is an illusion – network algorithms simply make only extreme comments visible.

For Switzerland, Tang proposes a decentralized digital identity system: citizens should be able to prove attributes (age, place of residence) without disclosing their name or address (partial anonymity). This prevents power imbalances and enables free criticism. Taiwan already uses such wallets for package pickups; Switzerland's planned E-ID could eventually be interoperable.

Taiwan itself is exposed: 3 million cyberattacks daily, targeted polarization campaigns. Tang's approach: understand cyber resilience not as defense but as decentralized robustness. With TSMC, Taiwan develops the E187 standard to make supply chains more independent – no single point of failure. Polarization attacks and disinformation paradoxically make Taiwan freer: people who criticize together are less susceptible to panic.

Key Statements

  • AI as consensus machine: Tang's Polis platform makes surprising overlaps visible, not polarization – 97 % approval in the Uber conflict despite initial polarization.

  • Participation instead of consumption: Digital systems must involve citizens in decisions ("broad listening"), not just collect complaints. Nearly 50 % of Taiwan already actively uses such platforms.

  • Media literacy through production: Young people as media producers (not just consumers) train critical thinking and fact-checking better than classical media criticism.

  • Decentralized infrastructure: Partial digital anonymity + federated network design (Fediverse) reduce dependence on individual providers and enable freer participation.

  • Polarization is energy, not disease: The problem is not passion, but that algorithms only show extremes. This is solved through systemic visualization of consensus.

Critical Questions

  1. [Evidence/Data Quality] How is it ensured that Polis's analysis of thousands of statements is actually representative of the entire population, or does it itself overrepresent self-selective online participants?

  2. [Conflicts of Interest/Incentives] If Tang's systems lead to consensus, is there a risk that minority positions are technically underrepresented, even if the visualization does not explicitly suppress them?

  3. [Causality/Alternatives] Is the success of the "join platform" (50 % usage in Taiwan) attributable to Tang's design or to cultural factors (e.g., trust in authorities, threat from China as a cohesion factor)?

  4. [Feasibility/Risks] How does the decentralized wallet system scale to countries with weak administrative infrastructure or lower digital literacy?

  5. [Evidence] The US example with 97 % approval on insurance questions – were respondents truly representative, or do such projects already attract participation-oriented citizens?

  6. [Causality] Does the ability to jointly verify facts actually lead to less polarization, or does less polarization already exist in the self-selection of these participants?

  7. [Risks] Tang's model relies on AI language models for consensus identification – how robust is the system against adversarial prompts or targeted disinformation in input statements?

  8. [Alternatives] Could a Tang-like system also be misused to legitimize decisions that show technical "consensus" but are politically controversial?

Sources

Primary Source: Polarization releases energy – but you must not direct it to extremes – NZZ, 11.04.2026 | Author: Malin Hunziker

Verification Status: ✓ 11.04.2026

This text was created with the support of an AI model.
Editorial Responsibility: clarus.news | Fact-Check: 11.04.2026