To the honorable members of the Diet, Minister Taira, MP Anno, members of Team Mirai and all our friends in the media, good morning. Thank you for this opportunity.

As you know, we all live in an era of high PPM. I do not mean parts per million of carbon, but polarization per minute. Every time we open social media, we are inundated with content designed to keep us enraged.

A decade ago, a profound change happened online. Platforms switched from showing you what friends posted to a “for you” feed. This shift installed a parasitic AI between people. These AIs maximize attention by maximizing division, discovering a business model of engagement by enragement. Bit by bit, our globally shared reality began to crumble.

But what if, instead of using AI to broadcast polarizing positions, we used it as a tool to listen at scale, helping us find common ground?

In Taiwan, we took this challenge head-on. We helped design digital spaces like Polis. It is deliberately different from the antisocial corners of social media: There are no reply buttons, so trolling is impossible; there are no retweet buttons, so outrage cannot be amplified. People simply see one thought from a fellow citizen. They can agree, disagree or pass, and then they see the next statement. That is it.

The algorithm does something radical: It rewards bridging statements that energize all groups to travel upward to higher ground where more agreement organically exists. In this way, the algorithm creates traction for up-wing discovery across all groups, and even, the joy of co-creation.

Then, we have a real-time visualization showing everyone’s avatars clustering together. People can literally see a group selfie of opinion, where each person is seen and brings their own angle — we are all in the same frame, showing where we stand apart and where we lean together. By listening at scale, we do not just find common ground; we make common ground, discovering surprising areas of consensus that people did not know existed before.

So today, I will spend about 15 minutes showing a few examples of how this works in practice, share the lessons we have learned over the past decade, and finally, offer some practical ideas for leaders here who want to get started.

Ten years ago, our first major test for Polis was when ride-sharing services clashed with our taxi industry. One side championed the innovation and choice Uber brought, while the other fought to protect the livelihoods of taxi drivers. On social media, the debate devolved into a zero-sum battle. But on Polis, the dynamic was entirely different. People offered concrete ideas, others agreed or disagreed, but without the jeering, without the pile-ons.

Within three weeks, we surfaced nine broadly supported, bridging consensus points. For example, one was: allow surge pricing during peak hours but prevent undercutting of meter rates during off-peak hours. Innovation and fairness, both at once. This became the basis for the subsequent regulatory update. The conflict did not disappear — it was transformed. Much like geothermal forces generate and give magma its momentum, this leaves everyone a little bit happier, and nobody deeply unhappy.

This is the spirit of vTaiwan, a process that marries open listening at scale with structured deliberation. While vTaiwan started in civil society, the government soon institutionalized it into the Public Policy Online Participation Platform — carefully branded as Join.gov.tw to signal what it does. On Join, any proposal with 5,000 co-signers triggers a response from the relevant minister. In short: making it easier for everyone to shape government policy that actually matters to them.

Here is an example. In 2017, two opposing petitions each gained 8,000 supporters. One said, let us move Taiwan’s time zone to GMT+9, aligning with Japan. The other said, let us stay at GMT+8. We were not talking about compromise, because a compromise would be GMT+8.5 — moving the clocks by 30 minutes, which is not only costly but makes nobody happy.

So what we did, again, was to find that uncommon ground. We listened to both sides and discovered the real motivation: to express Taiwan’s unique identity. Once this was articulated, participants brainstormed better, cheaper ways to achieve that, like creating a Taiwan Employment Gold Card for global open-source contributors or hosting high-profile innovation events like the Presidential Hackathon. And so the time zone debate dissolved, not because the conflict vanished, but because the underlying common ground found a smarter outlet.

Another example came last year when we saw a lot of deepfake investment scams on social media. These scams used the likenesses of famous figures like NVIDIA CEO Jensen Huang and flooded our feeds. But if you asked people individually, they would tell you they do not want censorship, and they do not want to limit the freedom of speech we have — Taiwan, like Japan, is the freest in all of Asia when it comes to internet freedom.

So, we tried a different approach. Instead of having a few people decide the options, we invited everyone to help set the ballot for something called a deliberative poll. We sent SMS messages to 200,000 random numbers across Taiwan, and thousands volunteered for online deliberation. We used the same polling technology — stratified random sampling — to select around 400 people who mirrored Taiwan’s demographics, like a miniature version of our country.

They met in AI-assisted small groups of 10 to promote equal speaking time and ensure discussions stayed on topic. Then we did cross-group pollination of ideas from these 40-some rooms. Only the bridging ideas that resonated with one another made it to the plenary session. For example, one room said: “All ads should be considered potential scams unless the advertiser signs it with a verifiable credential.” Another room said: “If a platform’s system serves an unsigned scam ad and someone loses millions, the platform must cover the full loss.” And another room said: “For offshore platforms that refuse to comply, we should not censor them directly, but we can slow down their content delivery speed — like their video playback speed. Freedom of speech is preserved, but freedom of reach is moderated.”

We then held a vote, and over 85 percent of the representative sample of 400+ people agreed on the core legislative package. The AI helped us find this uncommon ground and demonstrate to the legislature that no community was deeply unhappy about it. The MPs acted quickly, and the bill was passed in just a few months. The result is that this year, deepfake scam ads have largely disappeared from mainstream feeds in Taiwan. Here, we used AI to facilitate a democratic process, which in turn set guardrails for AI itself.

This approach is not just unique to Taiwan. In Bowling Green, Kentucky, we saw the entire city use Polis to inform its ten-year plan by elevating statements that were endorsed by residents from different neighborhoods and political affiliations.

In California, we partnered with the state government to build the Engaged California platform, using this bridging-first method for statewide consultations to get beyond the YIMBY/NIMBY, or Yes/Not In My Backyard, stalemate on contentious issues like land use and wildfire prevention and recovery.

We have learned two recurring, practical lessons:

First, mayors and legislators are very hungry for innovation. Who does not want a town hall meeting that is effective, nonviolent and leaves everyone happier? But the city staff is sometimes very thinly stretched and risk-averse. So, the way forward is to pick an issue that either saves time without adding risk, or reduces risk without wasting time. That is to say, improve on both dimensions, not as a trade-off. We work with existing vendors, universities, who are already doing polling and focus groups for the government or the legislature. We just tell them, with these generative tools, you can get a better group selfie, but the executive summary you produce at the end is in exactly the same format as a poll or a focus group.

The second lesson is about representation. Of course, you can start by letting everyone who is interested self-select in. That is valuable, but it can also skew the results. So, we designed a double-diamond model. The first diamond, agenda-setting, is open participation. But when you are converging on consensus, you use a stratified random-sampled mini-public. So, for the first phase, you can use something light like SMS verification, but for the second, you need a very rigorous method to ensure statistical representation.

Finally, I want to talk about the use of AI in this context. In Silicon Valley, the debate about AI sometimes gets reduced to two positions: accelerate or stop. But the real question is not whether to hit the gas or brake; the real question is who has got the steering wheel?

Because the method we have described is not about replacing elected leaders like you, the members of the Diet. It is, rather, about equipping you with a clearer roadmap — a high-resolution group selfie of the polity and what your society truly values. Citizens are best at discovering emergent issues and defining shared values, but the second diamond of the double-diamond model — crafting detailed policy and delivering results — still belongs to you, the leaders.

And a final note about the future of AI: I would not recommend going after a super-model that sees all and solves all. In East Asia, we have a different intuition. Indeed, right here in Japan, we have the idea of local Kami, the spirits that look after a place and the relationships within it.

So, it is not a single model to empower an individual superpower, but rather to take care of many local relationships. It is less like a personal tutor and more like a team coach. This is how we should design AIs: as our collaborators, our summarizers, our topic modelers — strengthening our civic muscles instead of letting them atrophy because we outsourced our judgment to them.

When I became Taiwan’s minister without portfolio responsible for digital affairs in 2016, my title in Mandarin, 數位, means both digital and plural. I would like to share with you my job description here. It goes like this:

When we see Internet of Things, let’s make it an Internet of Beings.
When we see virtual reality, let’s make it a shared reality.
When we see machine learning, let’s make it collaborative learning.
When we see user experience, let’s make it about human experience.
Whenever we hear that the singularity is near — let’s remember: the Plurality is here.

Thank you. Live long and … prosper.