Using AI to Strengthen Democracy: Audrey Tang on Taiwan’s Global Role

BY CHEN FAN-YU and WANG CHIA-YING,
COMMONWEALTH MAGAZINE

Taiwan is leveraging digital sovereignty for a future where AI serves not just as a technological tool but as a force to strengthen democracy and counter authoritarian influence.

Audrey Tang, Taiwan’s minister of digital affairs from 2022 to 2024, has lived almost as long as the internet has existed. Born in 1981, Tang has witnessed and participated in the transformative journey of digital technology from its infancy to its current ubiquitous role in global society.

For Tang, the internet is her playground and a cornerstone of her belief that knowledge-sharing and digital democracy can make the world a better place.

So when the internet becomes a breeding ground for misinformation and deepfake videos, social media turns into anti-social media, and generative AI lowers technological barriers, making it easier for authoritarian regimes and bad actors to manipulate AI for attacks, division, and trust destruction, Tang asks: How can AI be used to strengthen democracy?

Taiwan’s first-ever Minister of Digital Affairs is continuing her mission to demonstrate to the world that AI can become the ultimate listening tool. Its ability to scale deliberative democracy contributes to building consensus rather than tearing society apart.

Taiwan can leverage AI to combat misinformation, bypass authoritarian censorship, and play a crucial role in global AI development by using the democratic advantage of open policy discussion, Tang says. According to her, the “Taiwan model” of AI governance is built on two pillars: open-source systems and alignment with democratic values.

Among other efforts, Tang promotes Taiwan’s “Alignment Assemblies” project, utilizing AI technology to hold deliberations with citizens online on AI regulation and ethics, ensuring alignment with democratic values and showcasing Taiwan’s model of AI governance. She advocates for “bridging” algorithms designed to amplify perspectives that connect differing viewpoints rather than those that incite division, thereby strengthening democracy and building consensus.

In 2023, Tang was recognized as one of Time Magazine’s “100 Most Influential People in AI.” She had collaborated with the international non-governmental organization Collective Intelligence Project (CIP) to launch Alignment Assemblies, using open-source AI software to convene citizen deliberations online. These deliberations focused on AI regulation and ethics, ensuring AI aligns with shared democratic values.

While tech elites argue over whether AI development should accelerate or slow down, Tang stresses that the steering wheel must be handed to the people most affected by AI – citizens whose livelihoods, privacy, and access to information are directly shaped by AI-driven policies and technologies.

In early 2023, the Taiwan government launched a project called the “Digital Consensus Forum,” sending out more than 100,000 random invitations via the 111 government hotline and selecting 450 citizens to participate in the six-hour online deliberation via stratified sampling.

This was Taiwan’s largest online mini-public since it began promoting deliberative democracy in 2002, marking a significant milestone in its efforts to foster participatory governance. But how do you enable deep discussions among over 400 participants online?

The project managers divided people into groups of 10, spreading them across dozens of virtual rooms. AI acted as the moderator in these rooms, automatically managing speaking time to guarantee everyone had a fair chance to speak.

According to participants, the AI system’s moderation fostered a balanced and inclusive discussion environment. Tang highlights that this approach effectively reduced the influence of dominant voices, allowing quieter participants to contribute valuable insights. Participants who hadn’t spoken much were gently encouraged to share more, while interruptions or verbal attacks were muted, preventing discussions from being dominated by the loudest voices.

“Many important details come from those who speak softly,” Tang says.

Large language models automatically captured, organized, and summarized everyone’s input in real-time, enabling broad-based listening. These summaries were then used to identify common themes and areas of agreement or contention, helping facilitators and decision-makers to better understand participant viewpoints and refine policy recommendations based on a well-rounded collective insight.

The deliberative polling platform, developed by Stanford University’s Center for Deliberative Democracy, made the discussions of over 400 participants across 40 sessions instantly transparent. Participants could see where conversations converged, where disagreements remained, and what common concerns were shared, all of which improved the quality and depth of the discussion.

Their consensus informed Articles 30 to 32 of the Executive Yuan’s draft “Fraud Crime Harm Prevention Act,” specifically shaping provisions related to digital transparency, citizen protection against fraudulent online schemes, and measures to ensure accountability in AI applications.

In the same year, at the second global Summit for Democracy, Tang utilized the open-source opinion-gathering tool Polis. Polis allows anyone to initiate short surveys, like “Judges should not use AI tools in decision-making,” and others can respond by agreeing or disagreeing, with principles prioritized based on relevance.

Seemingly binary questions often hide complex, nuanced perspectives and diverse opinions. Through dialogue, participants can unpack opposing views and, with the help of AI chatbots, continue the conversation, listening carefully to differing voices while understanding the reasoning behind different stances, and eventually finding consensus.

The results showed that, despite initial differences, no one wanted AI to spiral out of control due to unchecked development. “Everyone agreed that we should invest at least to the point where mistakes can be rectified,” Tang says.

“Technology is a double-edged sword,” Tang noted in a speech at an event organized by Sitra, Finland’s innovation fund, in 2023. “While some emerging technologies benefit authoritarian regimes, others can rejuvenate democracy. So, what’s the most effective solution to combat online harms? The answer lies in ‘Plurality’ – technologies that foster diverse collaboration and increase democratic bandwidth.”

Bridging algorithms

Tang aims to champion the use of bridging algorithms to counteract the divisive nature of current digital spaces.

Algorithms today are designed to propagate content that provokes strong emotional reactions. Open any social media platform, and you’re inundated with sensational headlines, inflammatory remarks, and low-quality discourse – all optimized to capture and hold attention. In contrast, bridging algorithms prioritize content that capture civil discourse, setting a constructive tone that breaks through and provides a refreshing departure from the noise.

This shift has not gone unnoticed. Even X, the social media platform formerly known as Twitter, has incorporated elements of this approach through its “Community Notes” feature. Powered by open-source Plurality technology, Community Notes functions as a form of participatory moderation reminiscent of Polis, an open-data tool.

Volunteers play a critical role, acting as a jury to assess posts on whether they are persuasive, if they offer sufficient context, or whether they might be misleading. They suggest annotations that add clarity and context. If individuals from differing ideological backgrounds agree that an annotation enhances understanding, it is appended to the original post – even if that post is from Elon Musk himself – and it cannot be removed by the original poster.

Community Notes transcend mere fact-checking. Through the power of collaborative diversity, they illuminate multiple perspectives, add depth to public discourse, and bolster collective intelligence. Ethereum co-founder Vitalik Buterin has lauded the feature for its capacity to dismantle echo chambers, showcasing the potential of bridging algorithms to reshape digital communication into a more thoughtful, inclusive space.

“If this concept spreads, we could soon have social media that unites rather than divides,” Tang told Voice of America in an interview.

So, it’s no surprise that after stepping down as Minister of Digital Affairs, Tang joined the Project Liberty Institute, seeking incentives for major platforms to adopt bridging algorithms.

Research suggests that bridging algorithms neither reduce revenue nor usage time while improving user experience. Tang’s team has been in discussions with Google, and YouTube has in the already adopted this technology in the United States.

“Collaborative diversity is about fostering the synergy of diverse opinions,” Tang told Future City, a Taiwanese media platform. “This is my core belief, no matter what position I hold.”

But can a small democratic island like Taiwan really use AI to fight authoritarian regimes? Can AI even enhance democracy?

Tang points to Cofacts, a fact-checking platform built through civic collaboration, which transitioned from relying solely on volunteers to deploying AI language models for large-scale automatic analysis. When misinformation first appears, AI provides instant context and critical thinking. In this way, AI counters AI, and automated defenses using large language models help narrow the human resource gap.

Tang remains optimistic that AI language models, developed in the free and democratic world as open-source tools, will breach authoritarian censorship. These models are small enough to fit on a USB drive, can run on any personal computer, and do not require internet access, enabling access to uncensored information and challenging authoritarian regimes.

Taiwan’s semiconductor expertise can lead to more energy-efficient, powerful chips, accelerating global AI development. Taiwan’s democratic system – where policy discussions are open, and the social costs of such discussions are low – sets it apart from many other nations.

“We shouldn’t sell ourselves short,” Tang says.

– This article first appeared in CommonWealth Magazine in October 2024. It has been reprinted, with editing and updating, with permission from the publisher.