Opening Keynote g0v Unconference 2014

More info: g0v summit 2014 hackfoldr

More info: g0v summit unconf 2014 hackfoldr



Hi everybody. This morning’s keynote will have synchronous interpretation where you can access from your laptops. So this is our hackfoldr: . This is the same one as yesterday’s. By the side folder we’ve merged in today’s agenda as well.

If you use a computer with VLC player, please use the earphone.

The icecast interpretation channel is:

[live feedback] [[audience giggles]]

Please take your seats, we’re starting soon.

In addition to interpretation we also have live Hackpad transcripts — appearing on the screen as it happens in real time!

Before we start, we’d like to introduce you to IRC, the preferred tool for g0v to hang out in real time. There’s approximately 200 people at any time on IRC, so if you have an idea or something to call for participants, you can find collaborators on IRC really quickly.

Today we have 5 channels, and you can see what’s happening in each room on IRC and relocate accordingly.

jimmy> Hi, CL asked me to introduce first-timers and IRC-newbies to IRC. It’s a really simple interface for chatting. If you type in your browser, you’ll see a dialog box on the bottom right corner, and use it like Google talk to start discussions.

Today a lot of our old-timers in hackathon will start pitching our ideas here, and for example I’ll look for pm5 [typing pm5:] to send a message that pm5 will see later on.

In this mode of discussion, we can ping each other across rooms. Because this is the first-entry tool, if you’d like something more user-friendly, you can register on with your email, and after login it will ask you the channel to join. Type freenode channel and you’ll see the backlogs as well.

If you’d like to ping people on IRC but has technical difficulties connecting, please ask people around you, or otherwise just connect to

[PA: We need extension cords!] [[extension cords arrive]]

CL: LOTS of people went to yesterday’s venue — A group of people is being diverted back to this location — we’ll wait a few minutes for them to arrive.

We have English<>Chinese live interpretation, the URL is available in the hackfoldr here. If you’d like to connect to the icecast station, please ask people around you for the URL and setting up VLC:

Please use AS_Public as Wifi, it should Just Work (tm).

Good Morning everyone! [VERY LOUD MUSIC]

I can’t adjust the volume... Let me mute the video for now.

Let’s try again. Welcome to the g0v unconference — most of people were here yesterday [[raises hands]] — This is the first time we have a unconference after a conference. We have bimonthly hackathons, but this time we’d like to try a different format.

So I’ll start explaining what will happen today. Yesterday some people have proposed sessions already. You’ll find paper on the whiteboard where you can put in whatever you’d like to discuss, work on, or collaborate, grouped slightly by topic.

We’ll arrange the rooms, clustering together topics that are similar, so you’ll be in the same room talking to each other. Tip for unconf: It’s not a conference, just faciliate conversation and encourage people to participate and maybe come out with something.

We encourage you to find at least one person who takes notes for each session. We have a hackfoldr for the unconf here, and the hackpads are already there — see "Idea Pool" — so you can also put discussions here and related stuff, like a few ones there already.

So, we have 5 different rooms. This room will split into two. We have lunch outside, with an close  room. 108 is toward the entrance.


It’s my great honor to introduce Clay Shirky, NYU professor in ShangHai at the moment, and we WILL have a book signing party.


Thank you all. I talked to a bunch of people last night and this morning, but yesterday sounded like an amazing day. I’m sorry to have missed it, but I’m glad to be here now.

I’d like to do an un-keynote in the spirit of this unconference.

I want to talk about open-ended problems rather than hear some things we already know how to do. I want to talk in particular about the problems that both political movements, broadly participatory political movements, have with scale and the problems that people developing software tools for those groups also have with scale. I’m gonna concentrate in particular on three problems.

  1. Modularity
  2. (Social problems of ) Bridging capital vs Bonding capital
  3. Consensus vs Decisiveness (as a way of running the group)

Let me start with the basic observation which I sketched out on the whiteboard also.

Politics in participatory mode is the act of getting a group of people to come to an agreement. If two people have to agree, they have one agreement between them. If 3 people have to agree, they have to negotiate 3 agreements among the group. If 4 people have to agree, they have to negotiate 6 agreements between the pairs of people. ( 5 have 10 etc.)

If you ever wonder why it seems twice as hard to get a group of 4 people to agree what movie to go see as a group of 3 people, the reason is it is exactly mathematically twice as hard as there is twice as much negotiating to do in the group of 4 as the group of 3.

5 gets you to 10. 10 gets you to 45 (lines). The experience we had in high school, it’s correct — the more people a group have, the harder it is for people to join, because a group means everyone knows everyone, so N people requires N+1 connection.

Brooks’s mythical man-month, observation of IBM’s projects. IBM always have delayed shi pping schedules — not that anyone here has that problem — so IBM will add new people to “speed things up”, but Brooks observes it ALWAYS slows things down.

So it’s Brook’s Law: Adding more people to a late project makes it later.

The reason is of communication overhead of shared participation.

If you’re digging ditches, then it scales linearly. The more coordinated the effort has to be, that harder for growth.

The way we put it in NYU is that scale defeats density and vice versa. You can’t have a tight-knit large group. 6 people goes to dinner, but you can’t have intimate conversation with 60 people. It’s not about people liking each other or not.

Density defeat scale, scale defeat density.

Even worse, that’s true for all imagined densities. You might say fine, “everyone knows half of the people in this room”. Even that density cannot be maintained for scale. For density >0, there is a moderately sized group where it destroys that density.

Groups get more complex faster than they get large.

If you double the group size, you quadruple the overhead of complexity.

This isn’t sociology and this isn’t politics, this is MATH.

So this is a dilemma.

Optimization: First one: Modularity. 

Open Source projects tend to go modular with a plugin architecture. Not when the code base needs it, but when developers do. Whenever they are tired of maintaining this whole thing, that’s the thing that forces the modularity.

Larger code-bases may require modularity, but larger people-bases must become modular.

Many political movements started out as a group of people in a room feeling the same way. That natural impulse to preserve the feeling is so good but you can’t maintain it in scale.

It’s a long-term problem, widely replicated. How do you get a small group of people who can look each other in the eye, to a large group of people who encompass a large range of interests and skills.

There’s a lot of ways to look at the dilemma but no way to get out of it.

Yochai Benkler’s book: Coase’s Penguin identifies three principles:

1. How small is the smallest unit of participation?

When we move from small to larger public, how small is the smallest useful unit. The smaller, it’s easier for recruit. Recruiting for scale is something you’d like to do.

2. How easy is it for people to raise their hand and say I can work on this problem — opaque or complex organization makes it harder for people to find there way in.

In Occupy Wall Street I observed in 2011, I’ll go to the Park with a mental Blue Helmet “Not taking sides”. I got there, saw the people with their political signs — nothing surprises me — except the “shit is fucked up and bullshit” sign.

But walking into the Park and see people not holding signs but just talking to each other — I saw the food tent — not the signs — but just making sure people are fed and asking “how can I help?” I became a casual occupier at that time, and I self-identify as an occupier because it lets me find a little job and do it. I prepare coffee, help the food station, it’s the little units.

Rather than amping everyone, find little units. How easy it is to integrate it into the whole. This is both about politics and software design. Where people makes small contributions to integrate into the whole.

Software has something amazing — better styles for arguing and integration — I hope politics adopt that. If you look at how people are arguing on GitHub, that’s a profound model of arguing that says essentially “show you what you’d like to change and let someone else decide whether to integrate it.”

It’s built-in modularity. Focusing on the integration of volunteers is what any movements need to consider.

The second issue: Political movements must consider bridging social vs bonding social capital.

The difference between the two is this. Think of the list of people who will loan you 100 (US) dollars without asking.

If you are increasing bonding capital, you are increasing the amount of money that those people will loan you. If you are increasing bridging capital, you increase the number of people who would say “yes”. It matters is that the small group when you fell in love with, it is pure bonding capital.

If you look at the drawing here with 10 people and 45 links. Another way is two 5 groups with a bridge between them.

With two liaisons, that creates hierarchy of course, but the counter-risk is that you devolve into this hairball of complexity.

During the occupy movements, there’s multiple encampments all over the world. There’s never going to be a general assembly. So how can we increase the ties across the world?

Our goal was, every occupation needs two groups of two people, each knows two people in two other occupations. That’s the minimum [RAID] allocation that would survive severed connection. It’s the least dense possible arrangement. That’s the bridging capital.

We don’t need to understand or agree, but we do need to communicate.

How to build a movement from bonding to bridging capital is a huge design challenge.

If there’s a dozen people in the room, the answer you get is “That’s Diane” “That’s Sue” “That’s Scott”

40 people, “head of production, Diane”, “chief of …” etc with titles + roles.

Last one: Tensions between consensus and decisiveness.

Small groups tend to operate by consensus. Everyone has a Sense of the meeting, where everyone understands what’s going on. It’s what we know over and again, does not scale.

It’s a tragedy of the human condition. People commit to them but are defeated by scale.

In the 1980’s, a group called Act Up! emerged to fight the big pharmaceuticals and media around HIV and AIDS. (Amazing documentary.) In a matter of 10 days they created this incredible movement. It’s like consensus is great!

10 days later another document. “We need an executive committee.”

It’s spinning out of control! Ultimately in less than 2 weeks, we’ve grown so fast we’re out of consensus range.

It’s a permanent problem for any group — software to get people across that would be fantastic.

To avoid mistake: Especially with people working on data and code for politics.

45 years ago. Internet only had two programs, Telnet and FTP.

It would be great to get more people — scientist, researchers, remote access to data — but for years, nothing happens — there’s no third-party programs.

Early 70s, where Email was ported from MIT, the internet took off (75% of backbone traffic within 90 days).

The most significant thing is not code, not data, but the user. We have to have it. Whenever anything new arrives, we overestimate code and data but underestimate people.

So if you’re working on software that makes decision easier, you need to do the emotional piece, otherwise you’re treating people as machine.

My friend Marianne Manilov pointed out: A purely action-motivated movement can keep around for a year but peters out. Does the movement give a substantial time to take care of each other, is the key to going on.

“Taking care of each other” in the US is a radical idea. Strike Debt and Occupy Sandy are the examples of success that focus on these kinds of models: The intersection between politics and care-taking (the Spanish concept is ’politica affectiva’ - a politics of social relationships and love).

In the spirit of unconference and looking for ways to make progress, I’d say one of the persistent issues dated from the Iranian uprising, is that short term uprising is incredibly available now, but hashing out long-term commitment and communication is still weak.

Density vs Scale, dealing with the Modularity and Consensus and Decisiveness, and Bonding and Bridging capital.

Do not overlook the need of the members who does anything — from coding to occupying — take time commitment to take care of each other, if the movement is going to last.

[[CL does Chinese translation of the transcript above]]

[[[Transcript is over here]]]

Any questions for the speaker?

[我的中文不好,對不起] says Clay

Q from Rich: 

  1. Arguments on GitHub. That’s disagreement by contribution. Instead of just saying no, say "how about this instead." I think it’s a key component, the alternative-proposing part.
  2. With the diagram here. The individual people the dots represent. They can choose how to behave. They have different sort of cultural expectations of negotiation. They are not interchangeable parts. So two groups of five are difference.

The Loomio vision is to teach people how to recognize the density, and the personal needs in relationship to the group’s needs.

There are 16 members in Loomio already, but we’re functioning really effectively, because we put in the effort in how you operate, and developing the ways to scale but looking into the needs of each other, instead of being ego-centric.

Clay: This is simplified diagram. Culture is key. Modularity is a key thing, where I trust the group to do the things over there. But I think the Q of culture is a huge dilemma for us in the US around the movements, is daunting for any liberal people. It’s more difficult in a racially diverse groups. It’s like the worst news for us. In a community who’s diversified, the consensus rate goes down.

So in multi-ethnic groups. we have to create ’artificially’ increased social capital in the context of multi-ethnic groups. It’s also daunting that the two groups who do best with racial integration is Military and Sport teams, because have have clear objectives and an identifiable opponent. It’s difficult, but possible, and picking this up is something we have to do.

GitHub is great. I love that they are called "commits" — It says implicitly that if you are not "committing" I’m not interested.


Argument on GitHub is also issues. The difference between Code issues vs Political issues is testability and focus. I agree what you’re saying around GitHub.

[Clay clarifies: The Iranian event I mentioned is the June uprising.]

Multi-ethnic relation in Malaysia. We have a lot of work on that. The government — I appreciate what they do — is really helping in this regard.


Hi I’m Billy. Currently we find in Taiwan, most of people here does IT work and contribute to g0v. We are the minority however, we make awesome tools but most people don’t know about that. Older people may not actively seek out information, but passively consume newspapers and TV channels, which are obviously manipulated.

Our colleagues and classmates, some of them are not used to this kind of tools either — on their Facebook walls are cats photos and games — but they won’t read the things that may change their bias. What are your suggestions to motivate such people?


First of all, people mostly don’t change their mind, they mostly change their emotions and the mind follows — Like my mental blue helmet — the food tent changed my emotions and made me one of the occupiers.

If you concentrate on media as a way to transmit information, you’re playing in the field most controlled by your opponents. You need to communicate the emotion instead.

[[Tammy translate back to Mandarin for Billy]]

Second thing. Never underestimate the power of a cute :cat:  !

The "Cute Cat Theory" of internet dissidents is no joke! This is widely believed to be true — the tools designed just for political dissent will be ignored, but if you can transmit cute :cat2: on the same channel — it’s a huge advantage. :dog:   :dog2:  

Anthropologist Addition: In mainland China, the street art of political web. When 陳光誠 was arrested people put on sunglasses and staged CGC/KFC art images, which are harder to censor than words — don’t underestimate the ability of ART! [see this amazing book about creative activism:]

Last thing: If you are talking to people who don’t agree with you — say, middle-aged people — Don’t ask yourself how to get them to believe everything you believe — Ask if you can get them to believe anything you believe. Try to get across one thing, so you are not one crazy marginal figure — get them willing to listen to you because you said one sensible thing. If you can say one thing that makes sense — like Occupy Wall Street’s message about income inequality — get the one thing through and work with that.

CL: One more Q...

Q: Do you think the government can be operated in a modular way instead of a hierarchical system? Should the affairs in government encourage people to participate in the governance in small ways?

A: Given the logic of at least democratic governments, any form of citizen participation raises the social capital. Over and again when people study disasters, what makes a place resilient is whether people before the tsunami etc knew each other. It’s not what government does or sirens — those thing are important — but the vital thing is whether people know and trust each other. Any kind of participation is valuable for that.

I understand the concept of 關係, and that’s what matters, says Clay.

To the first half: Don’t think that modular necessarily means non-hierarchical, or vice versa. Hierarchy is a tool. Stamping it out is only useful when it’s part of a larger goal. I think modularity is a good goal, it’s evidently possible to scale out and up with modularity — the real Q is where can we leverage this — like in Icelandic constitutional referendum — it’s often pulled back by people in power. The Q is not about the tools, the Q is where can we find a country, or at least a city, that is willing to experiment with.

CL: I’d like to close this opening with a saying of Clay’s: The most important things came out with a hackathon is not the code or production, but the social capital that we built.

We’ll separate in 5 rooms, so please write-in the topics you’d like to work on or discuss. It’s not about presentation, but actually starting conversations by participants interested. Hackfoldr has the hackpad for each room and we’d like to ask volunteers to keep notes.

There’s food outside — please have food outside, not in this room — so feel free to come to the whiteboard and pitch in, and note the language of your preference.

So, food and planning time!

[[PA: The TutorABC booths has surveys for people who had experience with their equipments of interpretation yesterday.]]

[[Planning sessions ensue]]]





  1. 模組化(Modularity)
  2. 串聯(Bridging)與黏附(Bonding)社會資本(social capital)
  3. 共識(Consensus)與決斷(Decisiveness)




當你有 10 個人的時候,就有 45 條線(需要 45 組共識)。我們高中的經驗是正確的:當一個團體有越多人的時候,就越難讓新的成員加入,這是因為「團體」,代表裡頭的每個成員都認識彼此,所以 N 個人,就需要 N+1 個關係。

Brooks 在《人月神話》一書中提到對 IBM 專案的觀察:IBM 的交付日期總是拖延 --- 我想這裡的人都沒有這個問題 --- 所以 IBM 就僱用了更多人,來「讓事情加速」,但根據 Brooks 的觀察,這樣的做法卻總是讓事情更慢。

這就是 Brooks 定律:在一個已經延遲的專案上,如果你增加更多的人手,反而只會讓專案更加延遲。



我們在紐約大學(NYU) 的理論是這樣(註:講者是 NYU 的教授):「規模」會降低「密度」(Scale defeats density),反之亦然。你沒辦法有一個很大的團體,然後每個人又彼此之間非常熟識。6 個人一起晚餐,彼此之間可以有很熱烈的談話,但你不可能和 60 個人一起晚餐,然後每個人彼此之間都充分談話。這和人們喜不喜歡彼此無關。

更糟的是,這對所有想像中的密度都是成立的。你或許可以說,好,那每個人都認識房間裡一半的人。即使是像這樣想像中的密度,在規模變大時,也會變得難以維持。只要密度 > 0,當團體的規模成長到一定大小的時候,就會把這個密度給摧毀。






開放源碼的專案通常會使用模組化的架構。不是因為程式碼需要這樣設計,而是工程師們需要。當他們對 [必須由一個人] 維持整體的架構感到疲累時,就會趨向使用模組化的協作方式。





Yochai Benkler 的書 《Coase’s Penguin》點出了三個原則:

  1. 參與的最小單位是多小?


2. 新加入的人要知道「我可以做什麼?(我可以解決哪個問題?)」的困難度有多高?—— 不透明或複雜的組織會讓新加入的人更難融入。

2011 年,我去了祖可蒂公園( Zuccotti Park )觀察佔領華爾街運動,心中戴著我的藍色安全帽(指聯合國維和行動),一邊想著「我是來觀察的,所以沒有任何立場」。到了現場,我看到許多群眾,帶著他們的抗議標語——沒有什麼讓我覺得吃驚的,除了那個「shit is fucked up and bullshit」的標語。

但是當我走進公園,我看到人們並沒有拿著標語,他們只是和彼此聊著天——我看到食物攤——而非抗議標語——人們只是確認大家都有東西吃,然後互相詢問「我要怎麼幫忙?」。當下,我也成了一個抗議者,而且我自己覺得是這樣(相較於一開始的「我沒有立場」),因為我找到了一個我可以幫忙的小工作,並且開始幫忙。我準備咖啡,幫忙食物的攤位,這就是一個小的 [可運行的] 單位。

相較於將每個人都超頻增幅(amping everyone),可以改成找出最小可運作的單位,把這樣的小單位整合到整體中。這和政治有關,也和軟體設計有關。人們各自做出小小的貢獻,然後被整合進整體的專案。

軟體開發有個讓人驚豔的特點 —— 一個很好的爭論及整合模式 —— 我希望政治也能變成如此。如果你去看看人們在 GitHub 上是如何吵架的,你會知道,它的模式就是「告訴別人你要如何改變,然後讓其他人決定要不要整合這個改變」。

它是內建的模組化。我想 [政治/社會] 運動也應該專注在「如何整合自願者的貢獻 [成為集體的成果] 」。

第二個議題,政治運動必須考慮「串聯社會資本(bridging social capital)」與「黏附社會資本(bonding social capital)」。


這兩者的不同之處在於:想像一下,有哪些人會借你 100 元美金,而且不問你原因的?


我們再來看看白板上的圖,10 個人的團體,有 45 條連結。另外一個組成的可能方式是:有兩個 5 人的團體,中間有一條關係串聯兩個團體。

從分成兩個 5 人團體的圖來看,雖然建立了階層性,但分離了兩個複雜的人際毛線球,也降低了風險。


我們的目標是,每個佔領現場需要 2 人小組 ×2,每個 2 人小組要認識其他 2 個佔領現場的 2 人小組。這是能夠維繫連結的最小配置,也是最小密度的配置。這就是串聯社會資本的範例。我們不需要彼此理解或是同意彼此任何事情,但我們需要溝通。

如何轉換黏附社會資本的運動形式為串聯社會資本,也是一大設計挑戰。如果一大群人在一個房間,你可能得到的回應為:「這位是 Diane,這位是 Sue,那位是 Scott....」的介紹。40個人的話,就會演變成「Diane-產品總監」、「CxO」... 一堆頭銜和角色。




在1980年代,有一個 Act Up! 小組出現,針對醫療機構和媒體進行HIV、AIDS的相關倡議行動。在短短10天之內,他們建立了驚人的行動,這也像在說「共識」的方式其實不錯。




45 年前,Internet 只有兩種程式,Telnet 和 FTP。這兩種程式的出現,讓科學家、研究者等等可以遠端連線進行存取資料的作業,但是幾年過去了,沒有任何事情發生,也沒有任何第三方的替代程式出現。

然而在 70 年代早期,Email 由 MIT 移植 (Ported) 過來後,Internet 就起飛了 - 90天內佔75%的骨幹傳輸量。



我的朋友 Marianne Manilov 曾經指出:純然的行動導向的運動有機會持續數年,但可以一瞬間就斷掉。運動能夠有足夠的時間來照顧到參與的彼此,是運動能夠持續的關鍵。

而「照顧彼此(Taking care each other)」在美國是一個很基進的想法。專注於「聯結政治和照顧」這種模式的例子,有 Strike DebtOccupy Sandy 等運動。(Intersection between politics and care-taking,西班牙人的概念為 ’politica affectiva’ - 愛與社會關係的政治之意)

關於 Unconference 的精神,以及為了創造一些進展而尋找方向,我想說可以從伊朗革命的進展看到,短期倡議目前發展的非常好,但能夠長期的分散型溝通與投注(運動),仍然不足。

密度 vs 擴展性,決定於:模組化、共識與決斷、聯繫資產和黏附與串聯社會資本。





「我的中文不好,對不起。」Clay 用中文說。

Rich 的提問:

  1. 我想說一說在 GitHub 上的爭論。在貢獻代碼的時候會有分歧。但與其直接說「不」,大家更願意說「要不這樣如何」。我覺得這是一個關鍵點,就是說這種「總是提出實際的替代方案」的特點。
  2. 這圖上所畫的點代表著獨立個人,但他們可以選擇如何舉止。他們在進行協商時有著不同類型的文化預期,並不是一個模子刻出來的。所以這五人的組織,和另外五人的組織,可以是相當不同的。

我們  這套軟體就是用來引導人們意識到密度的重要性,以及處理好個人需求與小組需求的關係。現在 Loomio 團隊已經有了 16 個人,但我們依然高效運作,因為我們在運作上下了功夫。我們發展規模,但也重視相互之間的需求,而非自高自大。



GitHub 很棒。我喜歡它們的「提交」(commits)系統——如果你不先「承諾」(commit)點什麼,那我就不用感興趣。


GitHub 上也有議題(issue)。程式上的議題和政治議題的差別是,前者可以測試、也比較能聚焦。我贊同你剛剛說的。

[[Clay 解釋演講時提到伊朗時,指的是六月起義。]]]



嗨,我是雨蒼。目前我們發現,在台灣有很多在地的 IT 人會對「零時政府」做出貢獻。然而,我們是少數群體,我們做了很多很棒的工具,但是大部份的人並不知道這些。年長者們不會有動力去主動搜尋資訊,只是在被動地消費著報紙和電視頻道,這顯而易見會受人擺佈。






第二點。千萬不要低估了可愛 :cat: 的作用!

「萌貓理論」對網際網路上的異見人士來說並非玩笑!這個理論普遍被認可為真——專門設計來給異議者使用的工具總會被主流忽視,但如果同樣的工具可以用來散佈 :cat2:,那就是一個極大的優勢了。

人類學家補充:中國大陸政治網路的街頭藝術。當陳光誠被捕後,人們戴上眼鏡來扮作陳光誠,並結合肯德基的藝術形象,這樣就讓審查機制更難以審查了——千萬不要低估了藝術的力量![請參閱這本關於創意激進主義的書:] :dog: :dog2:






「我明白中文裡『關係』的概念,這個概念就是重點」,Clay 說。


CL:我想用 Clay 的一段話來總結這段開場:「黑客松的主要產物,並不是會動的程式,而是我們共同建立起的人際關係。」