Good local time everyone, my name is Audrey Tang and I am Taiwan’s digital minister and chair of the National Institute of Cyber Security.

It is a tremendous honor and a privilege to deliver the keynote at SITRA’s online event exploring ways of shaping technology to support, not hollow out, democracy. This gathering of the enlightened takes on great significance given its staging in conjunction with International Day of Democracy — an opportunity to double down on delivering compassionate, effective and responsive governance tailored to the needs of the people.

There is no question that democracy is facing unprecedented challenges at every turn from authoritarians seeking to harness the power of AI to destabilize and undermine our free and open societies. These efforts come in all shapes and sizes, but the most insidious ones target our fabric of trust, polluting social media platforms founded on Article 19 of the United Nations Universal Declaration of Human Rights — the right to freedom of opinion and expression.

In 2016, we witnessed a bellwether of sorts in public communication. The venerable Associated Press Style Guide declared the word “Internet” would no longer be capitalized, signifying its pervasive use. Unfortunately, this ubiquity also cleared a path for persuasion industries to engage in large-scale personal data collection and overturn privacy norms, marooning many on desert islands of polarization. Deepfakes — going fully interactive this year — have sullied online interactions, opening the door wider for fraud and manipulation at an unprecedented scale.

Consequently, democracy finds itself at a low ebb. Autocracy is on the rise, utilizing crafted censorship and electronic surveillance systems. The truth of the matter is global internet freedom has declined for the 12th consecutive year. Today, only a billion people live under the umbrella of democratic systems, while more than double toil under authoritarian rule.

It is said that there are two sides to every coin. While certain emerging technologies bolster authoritarians, some may yet be employed to revitalize democracy. How best to combat the pervasiveness of online harms? The solution lies in plurality, or technologies for collaborative diversity, to increase the bandwidth of democracy. Polis is a prime example of this ethos.

Taiwan started its journey with Polis in 2015 when Uber and sharing economy startups took the world by storm. The relentless march of innovation and profit clashed with the livelihood of taxi drivers, triggering legal action and protests. This conflict also raged in Taiwan, and we were fortunate to have Polis to track its route and derive shared wisdom from flame wars.

The beauty of Polis is that participants and opinions are visualized on the same page, as well as how divided the groups are. Despite initially polarized opinions, Polis consistently promotes common points of view to bridge divides. Polis does not just provide a visual mapping of democratic input, it also produces an interactive report, updated in real time.

This has applications for all issues of international import. Twice Taiwan has worked with fellow democracies like Finland through Polis. The first in 2019 amid escalating trade tensions, when some feared Taiwan would be caught in the middle of an epic struggle and sidelined from economic integration. Through the AIT@40 Digital Dialogues, ideas were canvassed across the Indo-Pacific on how to increase defense, economic and people-to-people exchanges between Taiwan and the U.S. These were then swiftly implemented by both sides.

In 2020, Taiwan explored privacy-preserving contact tracing and various pandemic countermeasures, again with avid global participation. It is notable that this was not limited to Polis opinion surveys — solutions were co-created. In recognizing which were aligned with societal values, civic hackers across the world proceeded to animate the ideas via hackathons.

This year, Taiwan is taking Polis to a whole new level. In response to the rapid popularization of AI, the Ministry of Digital Affairs has joined the Collective Intelligence Project to launch Alignment Assemblies. Other partners include OpenAI, Anthropic, the GovLab and GETTING-Plurality research network. The goal is to give the people a chance to contribute to the safe and sustainable development of AI, as well as identify and mitigate risks caused by this emerging technology.

Throughout July, we co-created with the people through Polis, exploring topics on democratizing AI futures and setting the agenda for face-to-face deliberative workshops in August and September. The records will soon be published as open data, facilitating the training of AI language models aligned with Taiwan’s distinct experiences.

To ensure authentic voices are heard in Alignment Assemblies, two features were introduced by polis.tw, which is maintained by the PDIS team at the National Institute of Cyber Security. First, we extended Polis to integrate with trusted external sign-on systems, ensuring the authenticity of participants.

Furthermore, with the help of AI Objectives Institute, the Talk to the City site shown here provides real-time AI-assisted summarization, translation and interaction. This chatroom empowers individuals to engage in nuanced dialogue with diverse opinion clusters, deepening cross-cultural understanding and empathy.

Now, take a look at this picture — every dot represents an opinion and sparkles like a star in the darkness of the background. The holistic reflections form constellations, providing us directions on alignment. For many participants, this is the first time they interacted with AI in the context of deliberation. This occurred not as a solo conversation, but as an evolving group discussion for the collective good.

In terms of AI governance, we believe that a democratic approach — not a technocratic one — is the optimal answer to what is an ethical and political conundrum of global proportions. The bottom line is experts simply cannot catch all use cases nor effects. Just as with climate change, comprehensive statistics are needed on potential harms, including over-reliance, emotional effects and information manipulation.

Unlike climate change, interactions with AI are deeply personal. All indicators must be shared in a thoroughly privacy-preserving manner. It is felt the safest way to operate is to form coalitions of data holders and researchers, mediated through trusted data altruism organizations, to safeguard processing and aggregation of non-personal data on harms and preferences.

Thanks to zero-knowledge proofs, privacy-enhancing technologies and a GDPR-compliant version of Polis pioneered by SITRA, worldwide Alignment Assemblies are well within reach. Step by step, let us build consensus on the boundaries of AI and shepherd its ongoing development, with interconnected local assemblies delivering specific directions for particular political and regulatory contexts.

In closing, I invite all and sundry to participate in Polis conversations of scale, and employ pioneering initiatives in the realms of collective intelligence to rein in societal risks while safeguarding human rights and democracy for generations to come. By drawing on the collaboration, curiosity, courage and vision of the people, we can chart a deliberative and sustainable course for global governance and free the future — together.

Blessings for your attention and time, and congratulations to SITRA for this exciting event. Live long and … prosper!