Audrey Tang, Taiwan’s former digital minister and current digital ambassador, talks to Nate Hagens about a different use of technology: not as a tool for automation and polarization, but as a way to hear many people and find common ground.
Her journey began with the Sunflower Movement in 2014. As she describes it, Taiwanese society was deeply polarized, and parliament was trying to pass a trade agreement with Beijing on the grounds that it was inevitable. The protesters were not just asking for it to be canceled; they proposed an alternative: they peacefully occupied parliament for three weeks while developing tools that allowed half a million people on the streets and many online to discuss the agreement, see how it affected them and submit proposals. Each day, what had been agreed was read out in the plenary session, and the demand became more coherent until the speaker of parliament accepted the demands.
After that, Tang became a “reverse mentor” to the cabinet: instead of clashes on social media, she designed digital public squares for polarized issues, from Uber in 2015 to the pandemic and generative artificial intelligence. She argues that the government’s approval rating returned to above 70% because less obvious points of agreement were systematically being discovered.
One recent example concerns fraudulent deepfake ads. The government sent text messages to 200,000 random numbers from the trusted number 111 and asked how citizens felt about information integrity. From the volunteers, 450 people were selected to form a statistically representative microcosm of the population. They were divided into 45 rooms of ten people and discussed the issue online for almost a day. Each group proposed ideas: treating any ad without the featured person’s signature as fraudulent, holding platforms liable for losses from fraud, or slowing connections to the servers of platforms that fail to comply.
AI facilitated the discussion, not as an avatar pretending to be human, but as a feature of the room: it took minutes in real time, encouraged quieter participants and identified common ground. After the plenary session, more than 85% agreed on a package of measures regardless of party affiliation. The major platforms could not exert pressure because there was no “pro-fraud camp.” The legislation passed very quickly and, according to Tang, such fraudulent ads no longer appear in Taiwan.
Tang explains that trust was built over a decade. Since 2015, citizens have been able to request a response through the national participation platform if they collect 5,000 signatures. This means that even those who did not take part in the initial consultation know there is an institutional process and a way to reopen an issue. She says the model scales when an issue is urgent and does not fall under just one ministry; she mentions the platform in California for wildfire recovery and local consultations such as the one in Bowling Green, Kentucky.
Tang distinguishes “assistive intelligence” from automation. She compares the tools to glasses: they do not replace people, but strengthen relationships between people. Unlike social media, which amplifies the most extreme messages, conversation networks should elevate voices that resonate with the whole room and distribute speaking time.
Before this, the gov zero and Polis projects were testing the same approach. Gov zero created free, open-source alternative versions of government services; the government could incorporate them if it wished. During the pandemic, gov zero’s version of contact tracing, which did not violate privacy, was adopted by the government. Polis allowed thousands of citizens to respond to proposals with “agree,” “disagree” or “pass” and visualized which ideas were accepted across different groups. The Uber consultation produced legislation that allowed Uber to operate as a legal taxi fleet. The same algorithm was adopted by platforms as community notes.
A core principle is using data about feelings, rather than only abstract values. Tang says everyone is an expert on their own feelings, while abstract economic questions alienate people. When people start with feelings, they tend to care about one another’s feelings, and proposals that take care of everyone rise to the top. This connects with an ethics of care and creates positive-sum outcomes, unlike a referendum where 51% feel like winners and 49% feel hurt.
In education, Taiwan changed its curriculum to prioritize civic capabilities: autonomy, curiosity, interaction with different people and building the common good. According to the 2022 ICCS study, 15-year-olds in Taiwan feel among the most capable in the world of influencing society on issues concerning people and the planet, without a decline in STEM performance. Tang says Taiwan is among the least polarized societies compared with similar OECD countries.
Pluralism, which she contrasts with the singularity, does not seek a vertical race of AI self-improvement that could leave humanity behind. It seeks a horizontal spread of capabilities so that every community can guide technology. The clash of differences is treated not as a fire to extinguish, but as energy for co-creation. Tang mentions “defensive democratic decentralization acceleration”: acceleration for decentralization, democracy and defense.
On social media, she describes a “prosocial” feed that ranks posts not by their addictive qualities, but by the cohesion they create among the user’s communities. Instead of each person seeing a different feed, she seeks shared experiences like the Super Bowl. She mentions LinkedIn as an example of more careful content ranking and the Project Liberty Institute, which she advises on a potential change to TikTok. She acknowledges that the economy of dopamine hits remains powerful, but believes there is demand for substance and shared experiences.
Democracy is described as a social technology with bandwidth and latency. Today’s “wicked problems” require more bits of information than one vote every four years, and faster feedback loops. Tang does not describe herself as pessimistic: she believes crises make the need for resilience visible and accelerate the spread of the idea that democracy must improve.
Finally, she discusses the importance of language. By repeating category errors such as “human resources,” we perpetuate distorted maps. She proposes Internet of Beings instead of Internet of Things; shared reality instead of virtual reality; collaborative learning instead of machine learning; and human experience instead of user experience. Her advice to citizens includes a grayscale filter on their phones, using a stylus or keyboard, listening actively without interrupting for five minutes, group facilitation techniques, intergenerational solidarity and organizing around a common purpose. She closes with Leonard Cohen: “There is a crack in everything, and that is how the light gets in.”
Comments