I’m excited about that. Thank you so much for your time and thanks for agreeing to meet with me.

I work at a not-for-profit organization based in India called Civis. And we actually work on ensuring that the voices of marginalized communities make their way into India’s lawmaking process. So, I was just sharing that we predominantly work at the national level on different laws and policies, and try and make sure that communities that are impacted by that particular law have the ability to share feedback before the law gets passed.

Nothing about us without us.

Yes, exactly. Exactly. And in India we don’t have, I mean, we have a system where any legislation is open to the public for 30 days.

Yes.

But, at the same time, there’s no way that I, as a citizen, get to know about this until and unless I’m constantly searching government websites or looking for these opportunities to share feedback.

In Taiwan you can subscribe to such consultation notices.

Oh, really?

And we, as a rule, offer 60 days.

Oh, wow. Exactly. Exactly. So, in fact, I was just saying that 30 days is the mandate, but typically we see it somewhere between 15 to 21.

Oh, wow. That’s like the old days in Taiwan, before we entered the cabinet.

Yes, exactly.

It was like 14 days.

Oh, is it?

In 2014, yeah.

Oh, okay. Right. So, we’re trying to strengthen this process by ensuring that community members that are affected by a law find out about it, but also that they understand the law in simple language. Otherwise, it gets very difficult…

Is it just law or regulations also?

It’s predominantly law.

Okay.

Because regulations also go through the same process, but regulators have been at it for much longer. So, they’ve been doing it for many years. Everyone knows to wait for regulators to put out public feedback.

Okay.

And that system is working fairly well. So, we try and focus on laws. And specifically, laws in the domains of environment, public health, urban governance, social justice that actually impact…

So, it’s two different websites in India. The law and regulation discussion…

Yes.

Okay. Because for us, it’s the same website.

Oh, that… We’re trying really hard, in fact, to try and get everything to be on one platform. Right now, though, it’s not two; it’s actually 78 different websites.

Oh, wow.

Each ministry will put it out on their own page.

Oh, yeah. It’s really like the bad old days.

Yeah, that was exactly the situation in Taiwan 10 years ago.

Oh, really?

Yeah.

Yeah. So, and yeah, and then we gather all of the feedback and we share it with the government and try and build dashboards and tools for government officials to be able to make sense of feedback. Because in India, there’s two problems. One is either you don’t get any feedback at all, like two or three people respond in a country of 1.3 billion. Or the other problem is that you get so much feedback that it overwhelms you, because it’s still a very much paper-based process.

So, it has to be printed and the officials have to read it.

Exactly.

So, if it’s a popular regulation or law and you have thousands of incoming messages, it’s like a denial-of-service attack.

Exactly, exactly. It’s exactly like that. And because of that, our government views this process with a lot of resistance. But it doesn’t have to be hard.

I mean we can put in place systems to make this process easy and better for governments as well as for citizens. But yeah, we’re in the early days still and our organization is trying to…

And is it like… Which languages are permitted when it comes to input?

You mentioned marginalized communities. I imagine like more than 10 languages.

22.

Right. 22. Yeah, my friend Abhishek, CEO of…

MeitY.

Yeah, said that that’s their deploying language model. So, that any language you submit, it will be translated even culturally. But I don’t know whether it’s actually used in production.

So, just this year in fact, he was leading this initiative called Bhashini, which does exactly this. It’s a model that is trained specifically on legal language to translate into 22 regional languages. So, we have champions like him who are really keen on the product. And he’s a very good process and we also work in support with government departments, especially his department a lot. So, I think it’s at that stage where we’re growing.

But I was very keen to learn from all of you about how the transition happened in Taiwan, what we can learn from it, who are the forces or people within the system that actually enable that process a little bit more. So, I was just wanted to spend some time to speak with you and to learn a little bit more.

Right. As I mentioned in 2014, the opaqueness of regulatory and legal action was the impetus behind the nonviolent occupy of the parliament in March, and precisely, because people don’t feel that the government is willing to substantially respond to citizen input. And, that’s why I think it’s important to talk about the fact that the government is a cross-strait trade service and trade act deal. So, that people are good for open governments.

Right.

And at the end of that year, all the mayoral candidates that didn’t support open government lose the election. And everyone, including William Lai of Tainan and Ko Wen-je in Taipei won the election. And they are now doing another election on the president now. So, on open government platforms.

This sends a very clear political signal that the issue transcends political party affiliations. This is about a citizen’s will to essentially accept nothing about us without us.

Right.

And so, at the end of 2014, the original premier resigned, and the new premier Mao Chi-kuo along with his vice premier Chang San-cheng said that crowdsourcing, open data, transparency is going to be the national direction. And so, to that end, we, who participated in the Occupy Movement, were hired as reverse mentors, people under 35 that advise the cabinet, on precisely this matter.

So long story short, I think we built alliances with the senior career public service. So, I remember that when I was first introduced to the cabinet end of 2014, Minister Jaclyn Tsai didn’t ask me to talk to other ministers. They asked me to talk to level 12 senior career officials.

There are exactly 120 such people. So, we had three training sessions. And so, they’re the director generals, administrators, right? And so, at once, they understood that we are on their side.

Right.

We’re not working for any political party. We’re working with the government and with the people, not for the people. And so, I think this solves a very important coordination problem. No ministry wants to act first on this unless every other ministry is working on this. So, in the beginning, Minister Jaclyn Tsai and I, and a lot of other people in the g0v movement introduced a scoreboard, So that on the vTaiwan platform that we prototyped back then, you will see how many days on average each ministry responds to people’s voices.

So, the more frequently you respond, the more substantially you respond, you score higher and it’s transparent, like everybody can see the dashboard. So, there’s a friendly competition. And so, each and every ministry related to e-services need to bring at least one law, and later on, at least one important regulation to be deliberated this way concurrently.

So, we handled a lot of things like closely held corporations, crowd sourced funding, Uber, Airbnb and so on. And so, the new government, Dr. Tsai Ing-wen’s administration, expanded the vTaiwan model to become the national one-stop participation service, the JOIN service, which combines, that was regulatory pre-announcement, but JOIN platform also adds e-petitions.

So, 5,000 people with e-collecting can summon a minister or their team to go where people are and listen to people’s ideas and respond point by point exactly as a member of the parliament’s inquiry. So, 5,000 citizens is equal to one MP, essentially, when it comes to demanding a point by point response. And also adding to it participatory budget and also budget visualization, commenting and collaborative meetings. So, if you go to join.gov.tw, all this is part of the platform.

Right. I think that’s so interesting to me because one of the things that we are currently trying to do is also try and understand how do you create those systems of incentive, almost gamified for different ministries. But one of the challenges that we face is that it’s every ministry we need to petition separately to give us information about the consultation, which ends up taking a considerable amount of time.

But I was keen to learn and understand from you a little bit more about other systems for open access of information. How does it work when a ministry might respond to a citizen? What form does that take? And just to understand that process a little bit better.

So, in Taiwan, we have for more than a decade the freedom of information laws. So at least the FOIA requests, there’s a standard request. The problem being the response to FOIA requests are point to point. So, if as a FOIA requester, I send to five ministries, first, none of them know the other four ministries is responding to this person. And also, if I receive this, but I decide not to share it. It’s the press or anyone else that they don’t know either. And so, they will make more FOIA requests. So, this shape creates burden, right? A denial-of-service attack, essentially.

So, the main argument in 2016, when we built the participation officer network, is to convince them that if 5,000 people want something to be answered, to answer them well once actually saves you a lot of time down the line compared to each petitioning through their input or whatever, right? So, that was the main motivation - as a time saver. And no public servant wants to watch the flame wars in discussion boards. It’s a waste of their time.

So, we restructured the discussion board around e-petitions so that there is no way to reply to each other. So, there are two columns. One is the supporting arguments; One is the counter arguments, but there is no way to reply across them. And you can only upload or download. So, at one glance, you can see the best supporting arguments and the best counter arguments. And so, it looks very high quality because the low-quality ones are buried, right?

So, there are a lot of such design ways to essentially harness the flame, the energy of the conflict, the tension without letting it diverge into polarization.

I have so many questions.

Of course.

I’m so interested. I think prior to founding Civis, I worked, volunteered with DemocracyOS, which was based in Argentina. And at that time came across Polis and then your work here. And it’s been many years in the making.

But I think one of the questions that I had, and you spoke about polarity and bringing up opinion. But often times for the citizen to understand that ‘this is important for you to respond to without taking a side’ as an organization becomes really difficult, because everyone is used to shouting headlines and so on. This is going to XYZ to you but actually explaining or…

Like informing citizens.

Right, exactly. Sometimes people don’t want to engage. So, I just wanted to understand, from your perspective and your experience, what have been some of the interesting opportunities to drive engagement from the citizen level so that you have two opinions or more quality input?

Yeah. So, there are, roughly speaking, two ways. And in academic words, it’s self-selection and sortation and because you know all the theories so I would just go straight to the point. Self-selection works best if the agenda is set by the people. And for most government initiatives, they only let people set an agenda if it is either interdepartmental, so every department has a different opinion, or it is emerging, meaning that no existing departments have the same. Everything else they won’t let the citizens set an agenda because they are the competent authority. They are competent in that particular matter.

No, this is public administration 101. So, from the very beginning we focused on things like Uber which is which exists in the intersection of the ministries of transportation, finance and economy. Or, for truly emerging things like Bitcoin, which is the purview of nobody.

Right.

So, for these two, there is a shared clear urgency for clarity. And whenever there is a clear urgency for clarity in the administration, they don’t mind letting the citizens set an agenda. So, just focus on these and ignore everything else if you are focusing on the self-selection model.

So, like nowadays is generative AI because it’s everybody’s business. It’s nobody’s business. So, what are the risks that can be caused by generative AI? For example, over reliance to opaque models - which ministry owns them, right? So, for this shape of things, our Taiwanese administration is perfectly fine for letting citizens set an agenda. So, this is self-selection.

For matters which a statistical representation is needed, for example, in Taiwan, we aim for that in the original design of the public healthcare single payer system or the electronic version of the healthcare system including dentistry coverage and so on. Now, that is truly multi-stakeholder. But if you don’t do a statistical representativeness, people who are very well educated, very well informed, you know, just dominate the conversation, right? So, for and but universal healthcare means truly universal even residents are covered. But they don’t have a vote. So, they don’t have easily an MP that count them as constituents.

Now, we are talking about self-selection when it has universal impact. And so, combining these two, it is conceivable that for something that has a universal social impact and no specific competent authority or overlapping competent authority, to be designed as a like double diamond system, where first the exploration, the problem definition discovery and definition is done in a more self-selected way. But once you have a clear how might we question, then the development and the delivery becomes a more binding referenda or citation and so on, right?

So, we’ve always designed our participation process with this double diamond in mind. And initially, before this ministry started most of our work is in the first diamond, because first we don’t want to compete with MPs on the second diamond. And for most of the topics we explore they really are stakeholder issues. They are not universal impact issues like the health care.

Right. Yeah, that makes a lot of sense. That visualization is very helpful.

I’m also curious about people within the system. So, you mentioned that there were politicians and they ran on that agenda. But at the end of the day, there were administrators who you had to work with and train…

Exactly.

How do you think about finding champions or cultivating champions within the administration? Or how do you think about overcoming any resistance?

Well, we are the resistance.

We are the rebels.

And sometimes you need some of that. But I feel like in our context where rebelling… or not even rebelling where like a combination of assisting and rebelling from the outside when there’s so much we could do if we were on the inside. So, I’m curious about that a little bit.

So, you’re already aware of the plurality mindset. So, let’s just talk about plurality for a bit.

Plurality means collaborative diversity. And things that are diverse you can basically use Dunbar’s number, like 150 people. Taiwan is perhaps here… And the world may be here. And collaboration. You can like… the lowest transaction cost collaboration is maybe just a few bits of, you know, let’s just quote a bargain. So, one dollar to one dollar. And voting. which you can design like for traffic voting. You know, interesting voting system. So, it’s just a step above market. And here you have conversation. And here you have co-creation. And here you have, let’s just call it, intimacy.

Now, the original constraint was that if you wanted these audiences in the world, we only really have a few.

For anyone, really, intimacy is really only shared with a few people. And so, we have this constraint of design of mechanisms. Let’s call it the first horizon. So, mostly, like if you want to hold a conversation, and you try to move beyond this, you hit real logistic problems, where people don’t have shared experience or even shared cultural context. And so, cultural context and what we call contextual confidence is destroyed if you move over the horizon.

And once it’s destroyed, there’s no way to get back. And there’s a backlash. And your horizon actually shortens. So, all the way so far, which I call conservative anarchism. It’s basically keep doing things that are just a conversation with stakeholders, maybe here, like well within the horizon, but then move everybody, including the administration, slightly with more diverse people. Or we keep the same stakeholders, but we move them slightly into deeper deliberation. Like instead of just online deliberation, also face-to-face and hybrid. Like instead of just a human summarizing it, now let the language model summarize it, and so on.

So, the cultural translation tools can work in both directions, right? It can elevate something that is more like a pole into a co-creation. And it can also, through translating 22 languages, turn something that is more like Polis into a conversation with a synthetic avatar of the Polis, right? This is like Talk to the City. And this is freshly open source. So, you can take the survey results and synthesize an avatar, and you can talk to that avatar. It’s like an executive summary that talks to you.

Right.

And you can infuse a lot of existing recordings of your community organizers, and let the people’s voice talk. And once you have a feedback mechanism for that constituent to tune the summarization, maybe they say, you know, I did say that, but not in that tone. Things like that. Then it becomes what we call a people’s AI.

And this is something that we work with Anthropic and OpenAI et al. So, this is a people’s AI. So, the word of plurality is about extending this horizon to the second horizon, but this cannot be done in a one push. Every point needs to be slightly pulled out and right. So, this is like a gradient descent. You have a lot of small groups that start in this scale and try to grow up. Or you have a lot of things that start somewhere here and through plurality and quadratic funding or something, try to scale out, right? So here, maybe we call it decentralization, or just call it word three because it’s shorter. And here we call it civic. So, taken together, this is decentralized civic infrastructure.

Does that make sense?

Yes, absolutely. That makes a lot of sense. I’ve been asking a lot of questions. But yeah, I think this really does make a lot of sense.

In many ways, we are trying to lower the barriers for conversation still. I think as an organization, we’re in that zone. Sometimes it takes the form of, for example, for communities that don’t have internet access, we do phone call-based feedback, IVR-based feedback. For communities which are somewhat conversant but have language barriers, we do a lot of WhatsApp-based feedback. And we’ve just started working on very small pilots of generative AI tools to ask questions from a law. Because oftentimes, what we are telling you is the summary, and may not be adequate for your experience. It might have very specific question.

So, we’re trying to build out tools for asking more questions from the law to understand it better and also working alongside governments on supporting them through this consultation process, almost to illustrate how it can be done and how it can be done in a way that you actually gain something from it rather than it being something where you’re just overwhelmed with too many voices or too many inputs. So, I think that’s a really interesting example.

Yeah. For senior public service, as long as each step is short and you’re well within their existing horizon, they’re happy because they would love to share from slightly more people. Or they would love to have the income signal to be better instead of just noise. Because these two directions are the natural desires of the public administration. It’s just nobody wants to fall outside of that horizon.

Right. Yeah, that’s true. And it’s interesting when you put it that way because there’s so much room to take to move. But oftentimes, from the outside, you might ask for something which is outside the horizon.

Right, exactly. Exactly. Like you’re asking me to jump off a cliff and say that I can fly. How does that work?

Exactly, exactly. But yeah, I think this is really interesting because in a lot of our work, we’ve been asking these questions, especially in a country of our size, representativeness becomes a very big challenge. But also, the quality of the feedback, the relevance of the input that we’re supporting the government with also is something that we’re working towards understanding better and trying to see. Maybe citizens also need a little bit of training on how to give feedback, how to be more maybe public-spirited. So, it’s not just something that benefits you but something that benefits collective commons.

Exactly.

And we’re building out some more modules and training of that kind, but also working on building out an open-source platform for insights from feedback that will help or support government officials in actually making sense of the input. Different languages, different ways of speaking. Different constraints. Some are coming in voice; some are coming via different formats. So, how do you bring that all together? That’s another plan that our team is actually working on.

That’s really good. It seems that you started well within the horizon, which is great.

But I think oftentimes we wonder if we’re… My curiosity is how wide will our arms go? At some point, you need to bring about that systemic initiative within the system because they have the widest arms.

Of course.

And I think that’s part of why everything that happened in Taiwan was so interesting to me because it almost went from an outside-in approach which you don’t see very often.

Yeah. As I mentioned, it doesn’t need to be perfect. It just needs to be better than the current administration, right? So, back in 2014, the trust in the administration was below 10%. So, it’s very easy for us to demonstrate our methods are better than 9%, right?

No, this is not being cynical. This is just political reality, right? So, I mean, like in Taiwan, the central bank works really well, which means that cryptocurrency doesn’t easily take off. Because even if you have Bitcoin Cash or whatever, even if you have stablecoins and so on, the NT dollars is doing just fine.

But in certain countries that just elected a new president where the inflation rate is very high, then Web3 has a thriving ecosystem, because no matter how volatile Ethereum or Bitcoin is, their currency is even worse.

Right. That’s so true.

So, focus on things that the public administration also needs help with. So, this is not like civic tech needs government’s help. This is civic tech can help.

And on that, if you don’t mind, I have just a small question. You have both the petitioning as well as the consultation, do you see that one fares better than the other?

No, they feed into one another.

They feed into one another. Right, right.

So, because consultation is sometimes about hitting the brakes, right, like stop doing this. But e-petition is more like, oh, let’s do that instead. But without the common availability of the budget and the regulatory consultation, there is no community to talk about alternatives. It’s impossible for a single person to think this whole policy context through. But if you have a community where you usually do participatory budgeting or consultation with, then at some point, you will say, actually, I can do better than these planners. And you already know that you have the support of hundreds, if not five thousand.

Right.

So, it’s like training for this. And beyond JOIN, we also have a referendum system, which requires collecting close to 300K signatures on a national issue.

Again, for most of the referendum topics, you see them first taking place in consultation and in e-petition just as a training wheel to get a referendum movement going.

And then you have enough conversation and context to have a meaningful referendum.

Yeah, so all of this is about contextual confidence. The more confident you have around your community’s context, the more partnership you can come into.

Right. That makes a lot of sense. I feel like currently we have petitioning systems that work very well for private organizations. Let’s say, Amazon to reduce plastic in packaging. But government petitions aren’t as effective because if they are not presented politically, let’s say, we petition to an elected official and a significant amount of the population comes in and signs that. If it’s petitioned administratively, it doesn’t have any impetus to action until and unless you get the political establishment looped in. So, we’re seeing somewhat limited success with government petitions.

But at the same time, the consultation window is where we see a lot more conversation happening with stakeholders and with communities.

Yeah. We… Because the joint platform uses a single platform, so we can do this Amazon and Netflix trick of ‘recommend it for you’.

You signed this, maybe you want to sign another. Sometimes in the petition discussion board, you will see that ‘join this Facebook group where we support each other,’ right? Exchanging signatures, which is great.

Yeah. And I think that’s also really fascinating. I was just mentioning that I’ve spent now three days in the country and I’ve really enjoyed it. But one of the most interesting things to me was, I think it was on the MRT or something where we scanned a QR code and it led us to the fan page on Facebook. I feel like it’s so good to see that these platforms are being used in conjunction. Because my personal thought around this is that people will speak where they speak. It’s very difficult to take them to another place and get them to engage.

Definitely. Yes, and I think that our main idea is that you can do most of the community organizing and so on on Facebook or Line or Signal or whatever. It’s only when you want to speak on the record with decision making or at least agenda setting power do you go to the government platform. So, there’s a very clear delineation. But they do both hyperlink a lot.

Right, right. And that also, I’m sure, helps build credibility in what you would say on the petition platform or actually increase the quality because there is a sort of sentiment attached to that.

Yeah, exactly. Exactly. And the Join platform is now celebrating what? Eight years? And, if you look at the Join platform, there is a section on top that says Collaborative Meetings. More than 100 of those were run with complete transcript, all the transparency stuff, and not even one of them went awry. And so, this is our way to communicate to the career public service that we’re well within the horizon. Like if you use this method, although it really only solves maybe 50% of the petitions, but it will never result in violence. And whereas the petitions on the street, you can’t say that, right? So, this is a safe place or a safer place. That is the main message we want to send.

Right, right. And that makes a lot of sense. I think the horizon analogy is a very strong one. Thank you so much for sharing that with me. And I’m truly inspired by everything that’s being done here, and hoping that we can learn from it and grow a similar ecosystem in India as well.

And specifically, if you have thoughts around using non-necessary generative, like deep learning models to facilitate the process, we’re actively working with the Collective Intelligence Project. Divya Siddharth just did a TED talk on this, so feel free to connect to CIP. And also, Wendy here is our point person for the CIP, feel free just to exchange name cards.

Absolutely, will do.

And Zach has been around like… my co-founder since 2016. And so, the process, the whole step, both of them were there. So, feel free to reach out and ask details.

Absolutely. I will definitely do that. I brought a few stickers. It’s what my luggage would permit.

Oh, wow. It’s only fair that I bring stickers. Let me do that.

I just brought a few extra. I thought the entire team might want to pick one up but… Some stickers and a notebook.

Oh, thank you. Thank you so much. I really appreciate that. And thank you for your time. It’s been a pleasure.