• You were just saying that you’ve just organized this annual conference.

  • We organize it annually, for 13 years already. It’s quite big. We are strong in Europe and in North America. We have important audience there. We are trying to build up our contacts also in Southeast Asia.

  • This is the first time we are here in Taiwan to create partnerships, relationships, to understand better the politics here, to understand the emerging technologies that are coming, also, from Taiwan, and to look on ways how we could cooperate in the future with NGOs, government, or private entities.

  • We also have a think tank. We have a convening power, but we also have a think tank that we are focusing on several political issues like future of Europe. For example, sustaining democracy in the digital age, which is a very important topic right now, especially when democracies are under pressure, especially with growing social media exposure that can be used for manipulation and altering political views of voters.

  • We are also dealing with cybersecurity. We have a very strong defense and security program. Especially in defense, we have very recently unveiled our yearlong research project with recommendations for NATO, how to adapt the new changing environment.

  • Some of our recommendations were made to NATO. We briefed the Secretary General, and the senior staff, and all the ambassadors about especially the technological change, which will have a tremendous impact how war is being fought and how we need to defend ourselves.

  • The original idea or the description of war that it is a violent clash of opposing interests. Maybe we can scratch out the violent because war can move into new domains in cyber, etc., and opens can cause much bigger problems even without the traditional, violent clash.

  • There are a lot of things we are doing. We are increasingly heard in Western Europe and in North America, but not here still. We hosted several times Japanese officials, officials from India. We also had representatives from Taiwan last year. Still, this is a growing relationship for us.

  • That’s from our side. Then we would like to also listen what you are doing in the government. What’s your portfolio? How do you see the future? We’ve heard a lot of good things about you. You started quite early, at 16 years old with a private company. Congratulations for your achievements.

  • ...inspirational person.

  • I’m past retirement probably now...

  • (laughter)

  • ...as the digital minister. I promise to be entertaining, so let’s get the projectors on, all that in just a second. You may want to sit to this side. Do we have extra chairs? You will be there in a moment.

  • How long have you been the digital minister?

  • A year and two months or something like that. It was last October.

  • We have a digital leader, also, in Slovakia who is the vice prime minister. In Europe, when it comes to digital, Estonia is most known. It’s a small country, but they are really...

  • They are very good at marketing.

  • Very good at marketing. We are trying to catch up somehow in Slovakia with the digital legend, as well. By the way, per capita in Bratislava, you have the most innovators now in Europe. We are small, so the capital is just half-million person, but, per capita, we have the most.

  • There is a project going on and a movement to market-center Europe as a Danube Valley, so to create an ecosystem where innovations, digital, etc., could be thriving. We have Vienna, Bratislava, Brno, Budapest, all one area, which could create a really heavyweight, comparing to Western Europe.

  • There is a lot already going on.

  • Where you’re at is the Public Digital Innovation Space, the PDIS, which is like maybe the GDS or the Etalab in France. This directly belongs in the administration itself and doing public digital innovation.

  • I’m the digital minister without a portfolio, meaning there is no digital ministry. We work directly with all the different ministries, together.

  • Across departments?

  • Across departmental. My three main mandates are open government, social enterprise, and also youth affairs.

  • In my open government role, what I’ve been trying to do, even before joining the government, is to spread the use of the so-called open, multi-stakeholder governance model, which is the model that, say, the World Economic Forum uses in Davos, where everybody speak as a stakeholder, instead of...

  • Yeah, in Davos. Instead of representing countries, they represent their stakes, basically. In the Internet governance of course, Davos is somewhat closed.

  • I was also invited to join the WEF more than a year ago, a year and a half ago. They kind of suspended my participation after I joined the cabinet for some reason, mostly political, I guess. [laughs] I look forward to resume when I’m no longer a digital minister of Taiwan.

  • In any case, this is a very different model. It’s a multi-stakeholder model. You have government officials, but they are not above anyone else. They just speak as one of the peers in the peer-to-peer governance.

  • You see in ICANN or in IETF, in the Internet society, for them, UN or any other multilateral entities are also just one of the peers. They are equally important as the technological representatives or the other representatives in Internet governance.

  • Personally, I was raised in this kind political model of the IETF, and then so long working on Internet standards, the World Wide Web, and things like that. My question is, "How much can we leverage this open multi-stakeholder model in a national government? How far can we go as?"

  • As, personally, an anarchist who refused to give commands, how far can we go without issuing a single command? That’s the basic question. To take one of our recent examples, in Taiwan, we have a really good e-tax filing system, where there’s a PKI structure. There’s an e-citizen card, and it’s a copy card now. You can touch or can use the normal IC card.

  • You can automatically download the tax, and it files for you. It’s really quick, like 10 minutes, and your tax is done. It’s really good. However, it only runs this smoothly on Windows. On Mac and Linux, you use Java applet, and it’s considerably less smooth. Also, Oracle just deprecated Java applets, anyway, so it makes the experience kind of bad.

  • Fortunately, we have an e-petition system. Our e-petition system is We the People in the US or many other e-petition systems, with a very key difference. In other e-petition systems, there is a guaranteed response from government once people pass some threshold, but it’s usually from one single ministry.

  • If it is cross-departmental, across ministry or issue, usually one ministry comes forward and say, "OK, this is what we can do. But I can just explain for you the other 80 percent of the problem, and we can’t really solve your problem for you." We see that in all the other countries.

  • Our e-petition system is powered by a cross-functional group of participation officers, or POs. When somebody petitioned, saying, "OK, the tax filing system is explosively user-hostile on Mac..."

  • (laughter)

  • It’s not just the Ministry of Finance’s business anymore. It involves many other ministries. In our monthly meeting, our participation officer in the Ministry of Finance...This officer is a new position that was set up just this year. Every ministry need to have a team, at least one participation officer that directly reports to the CIO, who is usually a deputy minister.

  • It’s a really straight reporting chain, and our POs are all operating in a virtual work space together powered by free software that I set up personally when I joined here, so everybody use the same whiteboard system, chat room system, issue tracking system and everything. We’re a virtual team even though we’re spread across 30 something ministries.

  • The Ministry of Finance participation officer, Yang Jin-Heng, just brought this petition case to our monthly meeting, so basically saying there’s people complaining about tax filing on Mac, and instead of just have the media officer explaining the problem, or the parliamentary officer explaining it to the legislators. It doesn’t really make sense because it’s a systematic failure, really.

  • It’s a technical thing, and it requires more than information technology. It requires service design basically, so it’s outside of their ministry’s expertise.

  • The idea of this participation officer now is that anyone in this PO network can ask all the other ministries for help, for cross-functional discussion, and they did receive help. We run this collaboration workshop every other week or so, so there’s like 25 collaboration workshops now using service design methods.

  • The idea is that we look at 5,000 petitioners, and we invite the ones that are most willing to join us in face-to-face meetings. They just come here in Taipei, or if it’s a petition about a regional issue, we just fly to a remote island or to a southern part of Taiwan and do a deliberation there.

  • The deliberation, of course, because there’s already more than 5,000 people interested, they will want to watch over live stream and over real-time transcript and things like that, and we do provide the service when asked.

  • We do a idea development session, and using user journey and service design methods, we mapped every single point in our tax-filing process. What are the touch points? What are people’s feelings and so on? Instead of just complaining, they become collaborators, volunteers really.

  • We run five workshops like this, and eventually the one who complained the most are discovered to be actually expert UX designers, and expert service designers. Basically, they know more than the contracted IT companies anyway.

  • We systematically say, once their CIOs is OK with it, every Friday, we run this workshop. The next Monday, I bring it to the prime minister and other minister with a portfolio saying, "OK, this is the consensus from the collaboration workshop meeting, so registering all over the nation, so how about we make it a national policy?"

  • The prime minister will say yes or no. If the prime minister says yes, then the Ministry of Finance now gets the official resource saying that, "OK, we have to redesign completely the tax filing system for Mac and Linux for the next year." It’s just like that. It’s a week’s time.

  • Using this mandate, we ran further design workshops, and produced within the span of one month or two a complete redesign of the interface, and also the service flow. The idea is that because it’s radically transparent, everybody knows and expects to know how exactly are we dealing with the petition, and where exactly are we in this direction.

  • We usually use a system called RealtimeBoard to convey this kind of communication. I’ll use another example. You mentioned the disruption of social media on trust and on democracy.

  • Here in Taiwan, we observed that in Facebook about a year ago, there’s a new class of scam that purports to sell some goods, like a USB disk or something, and for people, and pay on delivery. Once they paid on delivery and received the hard disk, they discovered that they write first kilobyte, and then it went broken, so it’s basically counterfeit goods.

  • It’s very cheap. It’s fake, but now they go back to Facebook and discover that there is no sender’s information in the delivery, and then nobody can trace the Facebook account. The customer rep service is actually a bot, so they can’t really locate where and who are responsible of this.

  • For this particular case, there’s the Consumer Protection. There’s the Ministry of Transport, of Finance, of Economy, of the Fair Trade Commission and the Ministry of Interior for Police, and even the Central Bank.

  • Those ministries’ participation officers are all involved in some stage of this discussion, so we invited not just the petitioners, but also stakeholders such as the Association of E-Commerce and the Association of Delivery, and up to five petitioners.

  • What we did was that we used service design methods to identify the problems along the journey, which are in yellow, the suggestions from the petitioners, which are in green, the actual responses from the ministries, which are in orange, and the factual laws and regulations, which are in blue, and the current challenges, which are in red.

  • Together, we just did this collaborative map. Before this, every ministry dealt with this problem on their own, not very effectively. Afterwards, everybody has a map, and it’s sent to over 5,000 petitioners personally in email. They basically subscribe to all the advancement that we do on this stage.

  • They see what you do with it?

  • Right. Then, after the prime minister say, "OK, we can handle it this way," the other thing is that there are two cards that are marked new here that says the E-Commerce Association need to work with Facebook, and basically more advertise the measures that Facebook has to fight such spam bots, and also link the Facebook verification system of companies, of pages with this system.

  • Most importantly, it has to act as a local e-commerce entity. When I visited Facebook and talked to their VPs, I basically showed him this multi-stakeholder idea and says, "OK, so now we’re all doing our part of the job. What about you? Do you want to be socially responsible or not?" The VP of Facebook said, "Of course."

  • They’re also only seeing this for the first time. This bot-based scam is quite new, and then they say it’s due to their incompetence. It’s not due to their MLIs, so they’re very much willing to join to help.

  • Two weeks afterwards, they just joined as the E-Commerce Association in Taiwan as a company, and met face-to-face town hall style with the people who are hit with this problem, and then also did their part of the job.

  • Just like the spam wars, everybody did a little bit to make it more costly to send spam emails, and then collaboratively, we controlled this problem to a place where people know what to do and where to do.

  • This is how Taiwan people expect to participate. We’re the number one in Internet participation willingness in the world. This is very interesting, because Internet development, especially World Wide Web, and democracy in Taiwan basically started at the same year, as 1989, where the press banning was lifted. Martial law was lifted, but also personal computers and early Internet was introduced.

  • We have a lot of televised or Internet deliberations and participations, especially during the Sunflower Movement. A bunch of students occupied the Parliament for 22 days because the members of the Parliament at the time refused to deliberate a trade service agreement with Beijing.

  • The MPs refused to deliberate that, and so students occupied to demonstrate how to deliberate this kind of trade service agreement, as a demonstration in a demo sense. It’s completely peaceful.

  • I was one of the first batch of people who supplied them with communication, with Ethernet, with WiMAX, with all sort of different streaming, so that everybody sees what exactly happens in the Occupy Parliament, in the streets, and nearby. At one time, we have half a million people on the street deliberating this very issue.

  • We develop a lot of information technology that let them enter their company name and see how exactly they will be affected by the trade service agreement so they can have a reasonable deliberation.

  • Every word that was uttered during the Occupy space was court recorded, like stenographed and made a transcript -- that’s how my habit began [laughs] -- and then published for everybody to see the next day, so the discussion could converge.

  • In many other Occupy, the discussion diverged. There’s too many different points. But in this kind of set-up, the conversation converged, because everybody can see what are still at stake or what are still in dispute the previous day. The next day can start where the previous day left off.

  • After 22 days, there’s this set of pretty strong consensus, like six or five consensus items. The head of Parliament actually agree with that, and so then the student retreated. The discussion group met, and then a forum was held by the administration to meet the demand of the student.

  • In that forum, it was proposed that we have an e-petition system like this. The platform and the system you just saw is a direct consequence.

  • Yeah, direct offshoot of the Occupy. After the Occupy, the previous cyberspace minister -- she wasn’t called digital minister back then...

  • Right. She’s also from the private sector. She was head of legal in IBM Asia at one point, so she also speaks the language of civic tech people. She proposed to our g0v movement, which is a bi-monthly gathering of thousands of people that works together to improve the government, what we call "fork the government."

  • The idea is very simple. For every government website that ends in gov.tw, like this, you just change the O to a zero. Then you get into the shadow website that does the same thing, but with open data and with civic tech. You don’t have to go through the 500 page of national budget. Instead, you can look at beautiful tree map or a conversational map of the annual budget.

  • It was then adopted, after the Occupy, by a lot of major cities. Budget.taipei actually used this as their participation budget system. You can see it’s very similar, the difference being, because this is endorsed by the Taipei city, everybody can have a look at any part of the Taipei city budget, and then have a real discussion with the civil servants, directly, bypassing the city council.

  • Is this something...

  • Inspirational, extremely.

  • Then, for people who want more detail, there’s always more detail to see. It’s not just visualization, because it’s what we call a social object. Instead of budget as a whole, I can just talk about the park that’s near me, and then have a conversation with a civil servant in response to that park. The infrastructure budget people can propose genuine new ideas, instead of ideas already under construction.

  • In this g0v movement, we just moved to recreate any government website or service that we don’t like, and then use the same URL, but change the O to a zero, so it still can be discovered.

  • I might have one question. I’m very sorry...

  • Yes, please. Go ahead.

  • Immediately, I had the question, "Are you not allowing the Chinese to see too much from Taiwan?" How you then secure your data, because that data is everything. How you secure these data? Someone in Beijing can see, freely, your budgets, then your everything.

  • I guess that increases trust.

  • As long as they don’t write the data, I’m fine with them reading the data. [laughs]

  • If I may just jump in.

  • We discuss also this issue of, for example, in Central Eastern Europe, Russia meddling in the domestic politics, internally, by spreading propaganda, fake news, by spreading disinformation. The goal of that disinformation campaign is not to win the argument or win the people for your argument, but to increase their distrust in the system.

  • How you increase your resilience as a democracy? You increase trust in the system...

  • ...through transparency. I think the only answer for that is strengthening your institutions. How you strengthen the trust of your institutions? Through more transparency.

  • That’s right. Trust is not like Bitcoin. It’s a relational concept. It only exists between many different sides. At least some party has to trust first. My work in the administration’s mostly to lower the fear, uncertainty, and doubt of the public service when it comes to facing...

  • Very good, a very, very novel thing.

  • ...people directly.

  • Can I ask you? Everywhere where there is a government, there is an issue of also corruption. What’s the perception of corruption here in Taiwan?

  • We’re way beyond where people think the vote would be rigged or it would be corrupt. The main sentiment is that the current laws are maybe too stringent, [laughs] that it stifles innovation because of the anti-corruption laws. But people are generally pretty trusting the public service being not corrupt.

  • That is good. That is one of the key things.

  • That’s also why we’re publishing...Instead of just the Parliament, the administration, and the courts, we have a dedicated system called the Correct Yuan. It’s an auditory body that is separate from the Parliament and the administration.

  • It audits both the Parliament and the administration’s use of budget and also campaign donations. When the legislators were still campaigning, the Correct Yuan would oversee it, independent of the administration and the parties that the administration and the legislation is in.

  • There’s debatable whether there’s a better constitutional design or not, but at least it’s an independent organization, well trusted by the population. They now are moving to publish their audit reports, including the specifics of the campaign donations online, so that not just they can do the auditing, but everybody can do the auditing.

  • This is another g0v project where we can look at the legislators, like the northern legislators and...

  • So you pick a specific legislator?

  • Right. You can just pick any legislator and see their voting records, their positions on the major issues. It’s linked to the Sunshine page, where you can see all their campaign donations and everything.

  • This is obligatory?

  • This is, of course, obligatory.

  • That’s what I wanted to ask you. Did you come into some obstacles, so that some ministries didn’t want to show or present the data, because they need to provide those open data for this to be able to be presented in this way, the user-friendly way.

  • Did you have any obstacles with some ministries trying to avoid this kind of transparency?

  • That’s actually two questions. Let me try to answer the first one. In our website, pdis.tw, it’s actually English website, we have designed the idea of what we call machine-readable, automatically reliable data gathering in the administration. The guiding policy is part of our Digital Nation eight-year program.

  • The underlying validation mechanism -- which is the thing that you just asked -- is that we ask all the systems that produces a user-readable interface, that has the potential to be cross-referenced by other users, either outside or inside the government, to provide, just like as you would provide an accessible A-plus standard for blind people to see.

  • Now they’re also required to provide an API, a machine-accessible version, using the OpenAPI Standard.

  • So it’s in the law that they need to do it in a way that’s...

  • It’s in the procurement regulations. It’s not a law, it’s a regulation.

  • Still, it’s forcing them to do that.

  • It’s basically allowing each ministry to ask their underlying...

  • I see, agencies, and not parties...

  • ...agencies and procurements to do that. It’s an enabling mechanism. We’re not saying that all the systems must be open data. We have a guideline that says, "If the procurement and maintenance cost over the past three years are less than one million euros, give or take a little bit, then it needs to be open data, by default."

  • Otherwise, it needs a write-up of why it doesn’t produce data to the National Development Council. They need to justify, specifically, if it’s not open data.

  • If it’s over or below?

  • It’s below. If it’s below, it’s automatic. If it’s over, usually it’s because the system was built using a lot of taxpayer money, and they will want to recover it in some pay-as-you-use API kind of way, in which case we allow them to write a cost-recovery proposal before they become 100 percent open data.

  • That’s designed for the National Palace Museum, [laughs] when they initially designed that. Now, even the National Palace Museum allowed a low-resolution, like 300 dpi, version of their pictures to be used as open data, while they’re keeping very high-resolution scans, which is very expensive.

  • They, frankly speaking, don’t want to appear on a lot of replicas, because they still want people to go to the museum. [laughs] That’s designed for that case, but usually now it’s open data by default.

  • Also, you may have read that Taiwan is the first place on the Open Data Index of the OKFN. Take it with a grain of salt, because it’s actually not something that’s very innovative, that requires any innovation. It only requires that all the basic data are open. Just let me go to this OKFN Global Open Data Index. Here we go.

  • One of the reasons why we’re the top at this basic data methodology -- here we go -- is because government spending is still red. We’re still working on that. Otherwise, we’re pretty close to a full score. The reason is twofold.

  • First, because of the regulation -- just as any of those systems, if it’s below one million euros, they just need to be developed by default. The second is that we don’t yet have a central data protection agency. The easiest way for two ministries to exchange data is actually through open data.

  • The two facts together meant that, mostly, open data is the easiest route when two ministries want to swap information in some way, because it has to clear the privacy concerns. You know whether the statistics is broad or coarse-grained enough to protect privacy, trade secrets, and whatever.

  • Once this is passed, it’s good enough for other ministries to use. Then, it’s also good enough for the people to use.

  • Also, we adopted a license that is compatible with the Creative Commons Attribution License for our open data license. Basically, any international parties, even if they are not subject to Taiwan’s law, they can just convert it to a CC5 license data set and use it internationally.

  • We get a lot of adoption from the Linux community, from the other community, as well, who will want to use the fonts or the basic vocabulary or whatever, produced in Taiwan. That’s the background of our transparency policy.

  • Now, we’re doing procurement support. We’re also using the Linux Foundation standards, the open API and also Speedex, so that the procurement can be, in addition to providing open API, also reuse as much as possible the open source components as part of their procurement.

  • Again, we’re not biased saying, "You must only procure open source." We’re saying, "You can say open source and open API are the key technical requirements."

  • What’s your goal? When you’re already the number one, what’s your goal? What do you want to achieve in two years’ time?

  • This actually is just like passing the written exam for the driver’s license. It doesn’t say how well you drive, really. [laughs] It’s a very basic measure. What we’re now looking at is social impact.

  • We now have these data. People generally trust that there is data, but we don’t release data at the moment in the cross-ministerial fashion. We respond to popular demand. This is something that is not yet a very well-placed pipeline.

  • For example, people will complain about air pollution in Taiwan. Instead of just having a very concrete map of where and why the air pollution comes -- whether it’s stationary, whether it’s coming from the traffic or from overseas -- we have different ministries collecting these different data in different ways.

  • All the data are there, but it’s difficult for the people to merge them together.

  • To merge it and connect it.

  • To connect it, to link the data.

  • You want to connect it and link the data?

  • That’s exactly right. We start from the air and the water, earthquake disaster recovery data, because there’s no privacy concern. The river will not wake up and say, "You invaded my privacy." [laughs] It’s all public environmental data anyway.

  • We are setting up a four-year program, where we use a national petabyte super-computing cluster to absorb all the data that’s indicated there and normalize its vocabulary using the state-of-the-art XML vocabularies. Then, allow the mission learning people to use the NVIDIA Tesla accelerator GPUs in place.

  • Instead of bringing data to people like the previous generation, we’re now also working on this style, peta style, peta-scale data. It lets the algorithm bring to the data, and then run the algorithms there and publish their statistics.

  • Automatically, of course as part of the pipeline. You just upload a container script or whatever to the super-computing center at the NCHC. Then, you automatically have access to lots and lots of the environmental data without having to download it, because it’s impractical.

  • Just the volume itself is very, very large.

  • When have you started with this policy? When you arrived to this position, or it has been implemented before by previous governments?

  • It’s hard to say because in the previous cabinet my role was, I would say, a private sector consultant, perhaps. I’ve been working as an understudy, you might say, since late 2014. I’ve been virtually in similar positions and co-designed the initial, under-a-million-euro Open Data by default policy. That was in 2015.

  • The very fortunate thing that we have is that during the presidential election and transition, the previous minister before the voting was Simon Chang and he’s a director engineering at Google. He is independent. He doesn’t have any parties.

  • Then, the premier that carries after the election, Dr. Lin Chuan, is also independent. He also has no parties. What they did was something new. Simon instructed all the ministries to using open data and free information to produce a checkpoint document. They upload it to the public Internet for the next cabinet to download.

  • That incidentally also enabled me to study, because I didn’t know it at the time. I joined the cabinet six months down the line. Everybody has access to the basic directions of where the nation is going, the basic data, the raw data. The handoff was really smooth. There’s no fighting, so it’s not on international news.

  • Then, I went to the cabinet, but there was no onboarding stage for me because I’d been working with this information and data for three years now.

  • When you were in this world economic forum working group, or...?

  • Working, I think I was in the Global Futures Council.

  • Global Futures Council?

  • Global Futures Council about the future of computing, I contributed to a report about using AI in politics.

  • Wow. Excellent. I don’t know if I mentioned this artificial intelligence here.

  • No, not really. [laughs]

  • We touched upon it a little bit, because it’s already...

  • One of the recommendations we made to NATO especially is to create a center of excellence for artificial intelligence and applications for defense. We are increasingly seeing that, when it comes to the defense and security where the technology could change and artificial intelligence, in a few years, the balance of powers can change very quickly.

  • US is paying quite a lot of money, but China and Russia are growing in their investments to AI. While AI is much broader issue than just defense, as every good thing can be misused by bad guys or by anybody, we need to be prepared for it.

  • We need to be prepared because if we are not prepared, then we’ll just be caught short. One of our recommendations currently in one of our big initiatives is led by John Allen.

  • John Allen was the adviser of US president. He is also in our advisory report. We proposed to create a center of excellence and create the NATO version of DARPA, the famous program of the US Department of Defense.

  • Just because there are so many things going on in private sector, so many new emerging technologies, you need to scan it and you need to know how to use it for good things. Artificial intelligence is one of the very important things when it comes to governance, when it comes to defense and many issues, health.

  • Definitely. I mentioned also very briefly that our petabyte-scale, super-computer center is actually primarily a GPU cluster optimized for AI research. Our idea is that because President Tsai has in her platform broadband access as human right. Taiwan traditionally has allocated, I think constitutionally, a budget for education.

  • Combining these two together, what we did was that in our new K-12 curriculum, we’re making ICT awareness and media literacy into the core curriculum of all the different fields. Also, using a special budget to ensure that all the remote schools and aborigine, indigenous villages and rural, remote islands all have broadband access to this super-computing facility.

  • Meaning that any child, anywhere in Taiwan, a year afterwards, can use AI or personalized AI as part of their curriculum, without leaving anybody behind.

  • This is fascinating. By the way, we will be also part of the World Economic Forum organizing the panel discussion there. Also, at the Munich Security Conference is our institution.

  • We are now in the process of creating a global council for a future of cyber security, together with EY and Microsoft. We are putting that together. It would be very good to keep in touch if you...

  • I’m interested, whether I’m in this position or some other position. I’m always interested.

  • Let me just ask you because all of these things are very inspirational. You’re the number one. Are you somehow helping other governments to export AI ideas?

  • How does it work? Are governments coming to you, or municipalities, or institutions? Are you helping them? Do you have a process for them, or do you have some lessons learned that could be shared?

  • We got a lot of interest there after we solved the Uber case because Uber is a global epidemic. The fact that we contained it through deliberation is critical.

  • How did you? I don’t know about that.

  • Very quickly then -- just a second -- using AI, incidentally.

  • The idea is that this sharing economy is a meme that says algorithm dispatch better than cars, better than the regulations for cars, so people don’t have to obey regulations, just algorithms. It spreads from drivers to passengers to drivers, just like a common flu maybe.

  • The driver, after driving for three weeks, they decide it’s not a really good idea anyway, but it already affected many people to install this app. There really is very few things that a sovereign nation can do to an app. It’s an idea that counts, that this idea lets people think it’s OK to hire a driver without professional license to carry them for a ride.

  • The idea is essentially saying, "We need to have a thorough discussion because otherwise, people are just looking at the size of this discussion." One side will say, "This is the future, this is our road to self-driving cars and whatever. It’s very cool, very hip."

  • The other side is always saying, "There’s no insurance. There’s no taxation. There’s no professional driver’s license. Nobody caries the liabilities. What do you do?’

  • The discussions tend to be more divergent over time, instead of convergent over time. The basic methodology that we use is called the focused conversation method. It’s invented in Canada. That says, "First, we need to do a stakeholder mapping that absorbs everybody’s facts, and everybody validates those facts."

  • Then, the feelings, we use AI -- which I’ll explain briefly -- to get people’s feelings towards the same facts. Then, we get people’s ideas, and the best ideas are judged by how much of their people’s feelings can this idea take care of? Now, we have a ranking of ideas. The best ideas, as long as they don’t contradict each other, we can just translate it into law and basically have it working.

  • The main problem of this ideal conversation is always the translation problem, because the professional private sectors, governments and people, they usually talk in a professional language, while people on the street talk with another language. As long as these people are being not as open or radically transparent as the people on the social media, it tends to diverge into incompatible groups.

  • Ideas become ideologies. That blinds people to new facts and blind people to each other’s feelings.

  • First, we curate it. Like you just said, just saw on my map a collective fact that everybody can agree with. Then, we run AI deliberation using the system called Polis.

  • The Polis system, I think it’s easier if you get to see the same action. We actually just ran a Polis for a very controversial project, whether we need to use flogging on repeat drunk driving, Singapore style.

  • It’s one of the petitions. About 27,000 people petitioned. Then, we use Polis to try to get people to at least agree on some basic understanding about not just human right versus crime prevention or retaliation, but whether it actually produces measurable outcomes.

  • We produce the facts, and then people can look at the sentiments and answer yes or no on each sentiment. As you answer yes or no, your position would move across the different groups. This has two psychological effects.

  • First, everybody here are your Facebook or Twitter friends, and they’re all over the place. It means that they’re not faceless enemies. They’re not antagonist parties. They’re your friends. You just didn’t talk about this over dinner. It brings some cohesion.

  • Second, it shows the people’s position can change and they do change over time. Instead of just shouting very extreme ideas, actually once you propose yes or no and you get grouped, you get to see people’s common ideas.

  • Even for people who are ideologically very opposed, 88 percent, everybody nearly, agreed that if government publicly discusses this, then everybody knows whether this is a good idea or not. They can at least agree on the necessity of a public discussion and things like that.

  • For the Uber case, after three weeks, they converged on five consensus items about the sharing economy being sharing only if you just incidentally drive to work, drive back and carry somebody else. It’s OK to call it sharing. Otherwise, if you’re going to 50 trips a day, it’s not really sharing economy because it’s just professional driving.

  • The second thing is that professional driver’s license is important. Insurance is important. Taxation is important. Choice is important for the existing taxi companies and co-ops to also operate in the Uber style, so they don’t have to be painted yellow if they’re going only to be held as an app.

  • All these are very commonsense things, but it’s collaboratively decided by thousands of people voting on each other’s idea and competing for consensus. After three weeks, everybody agreed on those ideas. Then, we collect those ideas and present them to the stakeholders.

  • The stakeholders have to show up, because if they don’t show up, they know thousands of people are looking and that they will be villains if they don’t show up to validate the people’s consensus. All the stakeholders has to show up, and then we go through each consensus item, one by one.

  • Check their understanding, whether it’s feasible or not. They basically are all committed to it, and because it’s all live stream, they can’t really take back their words.

  • When we translated this into the ratification in office, 2016, everybody saw it coming. Uber can’t help but just obey the rules. Now, Uber is legally operating using only professional driver-licensed drivers and paying their taxes. On the other hand, the existing taxi companies and co-ops are also operating to this kind of upholding, so it’s a win-win situation that was co-created.

  • After we did this and all the different countries saw that, I only answer questions publicly. That way Uber approve, at the time visited me. It’s not only on the record, it’s on 360 record. Everybody can just be put on VR glasses and re-experience the discussion. That’s Yeh Ning right there. [laughs]

  • The idea is very simple, that when other governments see it, they will want to institutionalize this model to fit their local political climate. I think the first interest I got was from the Paris City.

  • Who is facing very much the same problem with the taxis blocking the Champs-Elysées and things like that.

  • We had it, too, in our country.

  • Then, from there, I went to Madrid and also to Coimbra, and to London. People there are very interested in this idea. I think Corbyn worked this into his platform. Also, the New York City, Ontario, Canada, and Barcelona, of course...I visited Barcelona as a robot during the very hectic days.

  • It’s now spreading, mostly on the city government level. We’re also learning from the Icelandic system of their better...I can’t pronounce that word, Reykjavik or something? It’s their capital.

  • Yeah. There’s this system where in every petition they have a vote on the plus side and vote on the minus side, but they can’t reply to each other. There’s no fighting. It’s like Polis. You can’t reply to anyone else. The only way to refute someone is to propose a better, more resonating idea. We also copied that to our system.

  • We’re in constant contact. We’re always like...

  • Yeah, so that’s the idea.

  • For the petition case, the 25K petition ended up about 2,500 people, so about one-tenth of the petitioners went to the Polis system. For the Uber cases, again, in the thousands. We also look at the demographics. It’s not particular to cities or urban or rural areas. It’s roughly the same everywhere, the same as population density.

  • We’ve been experimenting, then, using more AI in this conversation, because at the moment the AI is only used to analyze the principal component to group people, to cluster people together and pick, and write automated report, which is already very useful.

  • We’re also now looking at AI to synthesize people’s ideas and feelings into personas. We can have a conversation in the Holodeck System for NVDI, and also, we’re also working with the Hololens people in Microsoft so that we have all the advertisers here.

  • To recreate the conversation so that when we’re talking about a park or a stadium, we can actually have a conversation in that hypothetical part or stadium, which is the only thing that is factual anyway when we’re talking about a future building. Otherwise, people have to understand the diagrams of architects, which is very difficult.

  • There’s a lot of experiments with VR. Now since we have 360 recording, we’re also working with the AI labs here to automatically identify the speaker, and produce real-time transcript, and things like that.

  • Wow, real-time transcript.

  • It’s pretty close to deployment actually. It’s now very much near human capability in terms of the accuracies and automated captioning. Now we’re working on a feedback system where we teach it how to place commas, or how to segment sentences, and to have it to have a more accurate vocabulary, and things like that.

  • Exciting, very, very exciting. Congratulations on the job you’re doing. This is tremendous, an inspiration. We’ll bring it back definitely as an ideas. I hope we can somehow come back and learn a little bit more.

  • Yeah, of course. I will email you some supplementary materials.

  • That would be very good.

  • As for limitations, first thing is that it really only works on cases where nobody has any idea how to solve. Like Uber, it’s new to everybody, so nobody will feel that they know the best of how to deal with this situation. The same with many petition cases.

  • That exactly was the Mac tax filing system that the government doesn’t even know how to do this. Of course, they can yield to the crowd-sourced agenda setting, or crowd intelligence as we put it. On cases where the government or the ministers feel that they already know the best answer.

  • Of course, this methodology is less used because it’s, after all, a consultation. It’s not a binding, general referendum, nor should it be because we can’t have a referendum for every single tax filing software case. That doesn’t even make sense.

  • Again, for things that ministries think they already have a very good idea of that we don’t see much use of this methodology. Now impede us because, again as an anarchist, I don’t give commands. I only give advices and support. They come to us when they need this kind of methodology, and if they don’t come, we don’t force them to use this methodology.

  • This is not like open data. Open data is under the Freedom of Information Law. It’s everybody whatever they should do. This kind of AI-based consultation is something that they opt in. There’s nothing forcing them to do this.

  • We end up getting monthly votes from all the participation offices on which cases are good for this kind of treatment. There are many cases where the ministry would say, "But this is on public record." Say, "This is maybe not the best use of this methodology."

  • Now there’s, I think, more cases that are judged by the ministries as not fit for this methodology than the case that they are think, that are fit for this methodology. Out of the 140 or something petition cases, only 25 are voted to use this collaboratively.

  • Why only that much? Because...?

  • There’s two things. One is capacity. For many ministries, to have more than two cases like this simultaneously in the same month would exceed their capacity of planning, and of communication.

  • Basically, they don’t have a large enough team to concurrently do two consultations at a time, especially for smaller ministries. This is true, which is why we’re now working on curriculum for civic participation for service design, and so I try to bring more civil service onboard.

  • At the moment, it’s very much still a demo idea for many smaller ministries. That’s the first thing. The second thing is that if it’s considered a politically sensitive topic, where the risk tolerance is lower...

  • ...then we will need to prove, essentially, that not using methodology will end up in a higher risk. We sometime can prove it, like during the Occupy, it’s clearly the case. For the tax filing stuff, clearly, if we do nothing, the risk will only increase. It will never decrease.

  • For the Uber case also, it would just exponentially increase if we do nothing. For other cases, maybe the ministry will think if we do nothing, the risk will lower over time, in which case they would not use this method.

  • I agree. Very interesting. How big is your team?

  • There’s also 15 interns, which...

  • Yeah. That’s the core team. The participation officers are 45 people, something like that.

  • You said you work with Iceland, Reykjavik, and some other...

  • Yeah. The city council people in Madrid, Barcelona, and also Paris City. There is a team in New York recently, Ontario recently, that’s pretty much it. The Singapore people also, we have some conversation, the GDS from Singapore, and also Etalab.

  • People faces many of the same challenges as we do, and I was, before joined the cabinet, in 6 of the 12 months, I was in Paris. I was literally based in Paris, discussing this with the city government, and so we still have a strong connection with them.

  • Very interesting. Maybe when we will be going on with our projects, like the Council on the Future of Cyber Security and others, we’d be very happy to be in touch, and invite you for some...

  • Of course. Just drop me a line.

  • ...maybe similar consultation. Thank you.

  • The more people know about that, the better, because it can now spread, and it can be useful, useful platform.

  • We got more requests when it appear on things like the "MIT Technology Review."

  • When it’s promoted, it’s...

  • Not the usual government channels, but the tech channels and the AI channels.

  • That’s a great question. I’ll have to clarify a little bit of our methodology. Our methodology is designed with the balance between security and privacy at one hand, and openness. We’re not sacrificing one for the other.

  • For example, the virtual workplace that we use, Sandstorm, it’s entirely free software. It’s open source, but then it’s also got professional cyber security audit, and we actually hired as our cyber security department the elite teams, the one who placed the second in the Capture the Flag of Defcon to attack the system and try to find its vulnerabilities and weaknesses.

  • Because it’s a open source it’s not the simple reverse engineering. They went line by line to look at the loopholes.

  • Once we have this secure underpinning, then we open this innovation for any public servant to write their own software to run on this secure sandbox system. They would write a system to let people order lunchbox together. People in our office us it every week to order lunch together.

  • It’s a simple JavaScript application. The person who wrote this doesn’t even have to care about cyber security because it’s running under a sandbox that’s already hardened.

  • This a layered approach that we’re using. A lot of innovation, transparency, openness on the upper layer, but with utmost importance on cyber security and the guarantee of secure communication on the lower level.

  • Also, as to your second part of your question, many people in this kind of internal meetings, they will be reluctant if I just set up a 360 live streaming that I post to everybody, to everywhere in the world. It’s forceful. It’s violent.

  • If it’s made into a transcription by AI, and then everybody get to correct, because AI still makes mistakes, together for 10 days, and then we publish it. We listen to each other again during the editing process.

  • We also can edit away the parts that we think are mistakes or are better not published. We do get people who get paragraphs and paragraphs out of their transcript after a meeting, which is fine. Then we publish the part that people think are good for the people to see.

  • This also protects the privacy of the public service, because the public service sometimes, after 10 days, decide maybe it’s not a good idea at all, so let’s forget we even had this discussion. The fact that we had this discussion is still important, but maybe not some utterances between the meetings.

  • There is still a compromise in order to get the public service to get into the habit of radical transparency.

  • It would be good if you could somehow showcase this to other digital services.

  • There’s some talks and some materials. I’ll send them to you.

  • Sure, that would be excellent.